Transparent Venture Capital by Tribe Global Ventures

Speakers compare AI boom to 2008 crisis via debt-driven compute contracts and analyze Canva layoffs.

1h 31m Listen to the original Episode page
Artificial IntelligenceVenture CapitalCloud ComputingMacroeconomicsCorporate StrategyFinancial Crisis

Summary

Speakers argue the AI boom mirrors the 2008 financial crisis, driven by debt-funded 'take or pay' compute contracts that create a future 'reset wall' of costs. They critique Canva's 20% layoff as insufficient 'peacetime' action given flatlined growth and structural product risks. The discussion highlights the fragility of OpenAI's balance sheet compared to Anthropic's, noting Microsoft's withdrawal of support. Finally, they warn that hyperscaler CapEx acceleration is decelerating, signaling a potential market inflection point.

Key takeaways

  1. 1
    More than $2.3 trillion of compute contracts now sit on the books of the four largest American cloud providers as remaining performance obligations and contracted backlog. Signed, celebrated, capitalized into equity prices, and, critically, not yet billing.

    Unidentified speaker (speaker_3) · 11:17

  2. 2
    Personally for Canberra Canva, the way they did the layoff, it was great. But in my eyes, it's still peacetime. And they're offering the $10,000 zero dilution to anyone that wants to start their founders. To me, that's peacetime. And personally, I think it should be wartime for Canva. And you look at all the stats coming out, their growth is nearly flatlined.

    Unidentified speaker (speaker_0) · 30:49

  3. 3
    Speaker 2 asserts that 'almost every layoff is too small' and that management failure is the root cause, arguing that insufficient cuts force repeated reductions that destroy culture and productivity.

    Unidentified speaker (speaker_2) · 31:31

Episode sections

  1. 0:10 – 0:39

    Canva Layoffs and Growth Concerns

    Speaker_0 discusses Canva's recent 20% workforce reduction, questioning if it is sufficient to reverse flatlined growth. They argue the company needs to operate in 'wartime' mode to re-accelerate growth, suggesting the current approach remains too much like 'peacetime.'

  2. 0:39 – 1:03

    Founder Metrics and Go-to-Market Mechanics

    Speaker_1 contrasts experienced founders who understand their go-to-market mechanics and distribution metrics with first-time founders who fixate on product. They identify founders who know their metrics as the 'sweet spot' for investment.

  3. 1:03 – 4:01

    AI Boom vs. 2000 Tech Boom and Financial Crisis Analogy

    Speaker_2 references an article by 'Groundbreaker' (an ex-Bank of America analyst) that compares the current AI boom to the 2008 financial crisis rather than the 2000 tech boom. The analogy posits that AI is a debt-driven real estate boom where the 'real estate' produces tokens.

  4. 4:01 – 8:44

    Commodity Products and Margins in AI

    The speakers discuss a principle from John Hampton's podcast: commodity products with low switching costs and low marginal costs see profits trend to zero. They apply this to frontier AI labs, questioning if high capital costs can sustain margins or if competition will drive them down.

  5. 8:44 – 17:05

    Debt-Driven Structure of AI Compute Contracts

    A transcript of an article by 'Groundbreaker' is played, explaining that the AI boom mirrors the 2008 subprime crisis. It highlights 'take or pay' compute contracts where payments begin at delivery (24-36 months after signing), creating a 'reset wall' of costs that must be met regardless of demand.

  6. 17:05 – 24:35

    Second Derivative Analysis and Valuation Signals

    Speaker_2 emphasizes tracking the 'second derivative' (rate of change of growth) rather than just growth rates. They note that OpenAI's recent funding round showed a decelerating increase in valuation despite accelerating compute commitments, signaling potential risk similar to housing delinquencies rising before prices fall.

  7. 24:35 – 30:03

    AI Agents Eroding Corporate Margins

    The speakers discuss how AI agents can automate tasks like switching utility providers or comparing insurance, eroding corporate rents and margins. They argue that as switching costs drop and agents handle decisions, marketing becomes less relevant for commoditized products, forcing a race to the lowest price.

  8. 30:03 – 30:23

    Canva Workforce Reduction Details

    Speaker_2 reiterates the shock of Canva cutting 22% of its workforce, noting positive comments on their handling of the layoff but embedding it within the broader context of industry changes and competition.

  9. 30:23 – 31:23

    Canva Layoffs and Rate of Change

    Speaker 2 notes Canva's heavy staff cuts and accelerating loss of tech jobs. Speaker 0 argues the layoffs are insufficient, calling the situation 'peacetime' rather than 'wartime' given flatlined growth, and questions if 20% cuts were enough.

  10. 31:23 – 32:34

    Layoff Magnitude and Management Failure

    Speaker 2, citing personal experience, claims layoffs are almost always too small and are a failure of management. They argue that cutting too little kills culture and productivity, causing top talent to leave, whereas cutting hard once is preferable.

  11. 32:34 – 35:22

    Canva's Structural Risks and VC Model

    Speaker 2 expresses concern that Canva's product is no longer in user workflows and questions how it can list without diluting shareholders, noting the VC model relies on paper markups. They highlight the difficulty of managing change loops in bloated organizations.

  12. 35:22 – 36:55

    Heidi Health and AI Competition

    Speaker 0 discusses Heidi Health's recent redundancies and co-founder departures, noting its rapid growth. Speaker 2 warns that AI allows incumbents in the medical profession to build their own tools, creating two waves of competition: direct rivals and counterpositioning by customers.

  13. 36:55 – 39:28

    Career Advice and AI Signaling

    Speaker 2 cites Annie Lau's view that safe careers are unclear, with technical and non-technical roles blurring. Speaker 1 argues being a founder is the only safe career due to autonomy. Speaker 1 also mocks AI researchers smoking to signal belief in AI.

  14. 39:28 – 1:30:46

    LLMs and Personal Themes

    Speaker 0 praises LLMs for automating mundane tasks like editing PowerPoint. Speaker 1 discusses their 'theme of the week': letting go of 'shoulds' to align with reality, mentioning their family structure and plans to buy a boat.

Assets discussed

AssetKindStanceContext
OpenAIprivate companyneutralHow does OpenAI, a company with some $40 billion of run rate revenue, sign $1.4 trillion of compute commitments?
Anthropicprivate companyneutralIn this case, it could be, I don't know, the valuation of OpenAI or the valuation of anthropic.
S&P 500indexbullishS P five hundred is positive 100% of the time after the midterm election since 1950.
NVIDIANVDAequityneutralNvidia's revenue is hyperscalar Scalar CapEx.
Meta PlatformsMETAequityneutralZuckerberg holds dual class control. He has run this exact playbook before and was rewarded with a tripling of the stock. And of all the hyperscalers, Meta has the weakest direct monetization of its AI CapEx, no public cloud to sell the capacity into, which makes its spend the hardest to defend and the easiest to cut.
AlphabetGOOGLequityneutralGoogle will not blink. It builds TPUs at a structural cost advantage and reports a cloud backlog north of $460 billion, so it benefits if rivals retrench.
AmazonAMZNequityneutralAmazon may have its hand forced from the other direction, free cash flow already turning negative under the build.

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

Canva laid off 20% of its workforce.

Unidentified speaker (speaker_0) ·0:10

How to verify: Canva press releases or official company announcements regarding workforce reduction in 2026. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Canva's growth has nearly flatlined.

Unidentified speaker (speaker_0) ·0:10

How to verify: Canva's public financial reports or third-party analytics firms tracking SaaS growth metrics. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

The article comparing the AI boom to the 2008 financial crisis was written in July by an ex-Bank of America analyst known as 'Groundbreaker' on Substack.

Unidentified speaker (speaker_2) ·7:33

How to verify: Search Substack for 'Groundbreaker' and verify publication date and author background. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

More than $2.3 trillion of compute contracts sit on the books of the four largest American cloud providers as remaining performance obligations.

Unidentified speaker (speaker_3) ·11:17

How to verify: SEC filings (10-K) of the four largest US cloud providers (e.g., Oracle, Microsoft, Amazon, Alphabet) for Remaining Performance Obligations disclosures. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

Oracle's RPO grew 363% in a single fiscal year.

Unidentified speaker (speaker_3) ·11:37

How to verify: Oracle SEC filings (10-K or 10-Q) for the relevant fiscal year. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

OpenAI signed $1.4 trillion of compute commitments.

Unidentified speaker (speaker_3) ·12:37

How to verify: OpenAI press releases or SEC filings if publicly traded, or credible financial news reports citing contract values. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

OpenAI raised $11 billion at a $1.2 trillion valuation, down from a previous $30 billion raise at $1.4 trillion.

Unidentified speaker (speaker_2) ·21:23

How to verify: OpenAI funding round announcements or financial news reports from 2026. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

The article by 'Groundbreaker' was written in July.

Unidentified speaker (UNKNOWN) ·7:52

How to verify: Publication date on the Substack article by 'Groundbreaker'. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

The 2008 subprime crisis was driven by the second derivative of house price growth, not just price declines.

Unidentified speaker (UNKNOWN) ·9:17

How to verify: Economic analyses of the 2026 crisis, specifically those focusing on housing price appreciation rates and delinquency correlations. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

AI agents can save consumers money by switching utility providers and insurance, eroding corporate margins.

Unidentified speaker (speaker_2) ·25:14

How to verify: Case studies or user testimonials from AI agent platforms like Grokbot or Instinct demonstrating cost savings. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Canva offered $10,000 zero dilution to anyone wanting to start their own founders.

Unidentified speaker (speaker_0) ·30:49

How to verify: Canva press releases or employee handbooks from the period of the layoff announcement; Canva's official blog or social media channels. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Canva laid off 20% of its staff.

Unidentified speaker (speaker_0) ·31:09

How to verify: Canva's official layoff announcement or press release; LinkedIn profile changes for former employees; news reports from major financial outlets. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Canva's growth is nearly flatlined.

Unidentified speaker (speaker_0) ·30:49

How to verify: Canva's internal metrics (if disclosed), third-party app usage data (e.g., SimilarWeb), or industry reports on design software adoption rates. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Almost every layoff is too small.

Unidentified speaker (speaker_2) ·31:31

How to verify: This is a subjective opinion/claim by the speaker. Verification would involve comparing announced layoff percentages to subsequent restructuring events in the same companies. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Heidi Health announced redundancies and two of three co-founders departed in the same week.

Unidentified speaker (UNKNOWN) ·35:22

How to verify: Heidi Health press releases, LinkedIn profiles of co-founders, and news reports from medical tech outlets. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Heidi Health is growing faster than Canva was from year one.

Unidentified speaker (speaker_0) ·35:34

How to verify: Heidi Health's disclosed growth metrics vs. historical public data on Canva's early growth (if available) or industry benchmarks. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Australia's inflation rate is second only to Brazil among OECD countries.

Unidentified speaker (speaker_2) ·42:46

How to verify: OECD inflation data releases; Reserve Bank of Australia (RBA) statements comparing Australia to other OECD nations. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

The S&P 500 is positive 100% of the time after the midterm election since 1950.

Unidentified speaker (speaker_2) ·44:14

How to verify: Historical S&P 500 data analysis from sources like Yardeni Research, Bloomberg, or Federal Reserve Economic Data (FRED) for election years post-1950. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Federal Reserve total assets went from 3.77 trillion to 8.96 trillion during COVID and are creeping up in 2026.

Unidentified speaker (speaker_2) ·45:31

How to verify: Federal Reserve H.4.1 Statistical Release (Assets and Liabilities of the Federal Reserve System); Federal Reserve Bank of St. Louis FRED data. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

OpenAI's valuation rose from $86 billion in early 2024 to approximately $852 billion by spring 2026.

Unidentified speaker (speaker_3) ·1:01:11

How to verify: OpenAI's term sheets, press releases regarding funding rounds, and public filings if any; news reports from major financial outlets covering the IPO process. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

Anthropic generates $1.70 of revenue for every dollar of compute.

Unidentified speaker (speaker_3) ·1:02:57

How to verify: Check Anthropic's investor presentations or financial disclosures for revenue per compute unit metrics. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Anthropic's liabilities are protected by a $35 billion Apollo/Blackstone TPU facility with Google guaranteeing lease shortfalls and Broadcom guaranteeing residual silicon value.

Unidentified speaker (UNKNOWN) ·1:03:41

How to verify: Review SEC filings or press releases for Apollo/Blackstone TPU facility terms and guarantee structures. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

OpenAI has no path to positive cash flow this decade.

Unidentified speaker (speaker_3) ·1:04:01

How to verify: Analyze OpenAI's financial projections and burn rate disclosures in investor updates or S-1 filings. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

Microsoft stripped away every structural strut in April 2026, ending revenue share and exclusivity for OpenAI.

Unidentified speaker (speaker_3) ·1:04:21

How to verify: Check Microsoft and OpenAI press releases or SEC filings regarding the termination of their strategic partnership terms in April 2026. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

Softbank is trying to raise a $10 billion margin loan against its OpenAI stake.

Unidentified speaker (speaker_3) ·1:04:41

How to verify: Review Softbank's financial disclosures or news reports on its financing activities and OpenAI stake pledges. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify Canva's latest revenue growth rate and user engagement metrics via their official blog or press releases.
  • Check the 'take or pay' contract terms and remaining performance obligations for AWS, Azure, and Google Cloud in their latest 10-K filings.
  • Review OpenAI's S-1 filing (when released) for audited losses, customer concentration, and debt obligations.
  • Compare Anthropic's revenue mix and EBITDA margins against OpenAI's using available financial disclosures or credible third-party reports.
  • Analyze hyperscaler CapEx trends as a percentage of operating cash flow using quarterly earnings reports from Microsoft, Amazon, Alphabet, and Meta.
  • Research the historical performance of the S&P 500 following midterm elections using Federal Reserve Economic Data (FRED) or S&P Dow Jones Indices.
  • Assess the impact of Australia's inflation rate on local interest rates via the Reserve Bank of Australia's monetary policy statements.
  • Investigate CoreWeave's DDTL 5.0 facility terms and OpenAI's customer concentration in their respective SEC filings or press releases.
Full transcriptexpand
0:10
Unidentified speaker (speaker_0): Personally for Canva, the way they did the layoff, it was great, but in my eyes it's still peacetime. But personally, I think it should be wartime for Canberra and you look at all the stats coming Now. out. Their growth is nearly flatlined.
0:24
Unidentified speaker (UNKNOWN): They've
0:24
Unidentified speaker (speaker_0): had to lay off 20. Was that enough? I don't actually know. Do they have there's so many moving things here? On the outside, it still seems way too much peacetime. In my eyes, it's got to be war time for them to re-accelerate growth and get an outcome that they should deserve.
0:39
Unidentified speaker (speaker_1): You can tell founders when they sit down and they know their numbers and they speak straight to distribution. And they've got they know their metrics and they know what's working in their go-to-market. As opposed to say some of the first-time founders that tend to fixate a bit more on the product side quite early and don't really yet know their mechanics. And the sweet spot for us is those that do know the mechanics
0:59
Unidentified speaker (speaker_1): of their go-to-market.
1:03
Unidentified speaker (speaker_2): We, like many others, have often compared this tech boom to the 2000 tech boom. And even in the podcast with Joe Walker and John. there was quite a bit of discussion of what happened in the 2000s. But this uh article actually puts forward that this thing has the
1:22
Unidentified speaker (speaker_2): potential to be very similar to the financial crisis in the analogy of it's actually a debt driven real estate boom. It's just that this real estate produces tokens.
1:38
Unidentified speaker (UNKNOWN): Morning,
1:38
Unidentified speaker (speaker_2): mate.
1:39
Unidentified speaker (speaker_1): Morning.
1:40
Unidentified speaker (UNKNOWN): Are
1:40
Unidentified speaker (speaker_2): you wearing a jumper in the middle of summer?
1:43
Unidentified speaker (UNKNOWN): I'm
1:43
Unidentified speaker (speaker_1): out of Glen Innis and it's it was two degrees overnight, so it's freezing.
1:47
Unidentified speaker (speaker_2): Two two degrees in November. Wow.
1:51
Unidentified speaker (speaker_1): Yeah, it's a different climate down here.
1:53
Unidentified speaker (UNKNOWN): That
1:53
Unidentified speaker (speaker_2): sounds sounds special. Not sure where Jack is. Should we give him a a reminder?
1:59
Unidentified speaker (UNKNOWN): Oh
2:00
Unidentified speaker (speaker_1): he just shot showed a message here who said he'll be two minutes. Mind you, that message was three minutes ago.
2:06
Unidentified speaker (speaker_2): One of us. One of us. I had a really good week last week down in Melbourne with the Canopy team, Nigel and his team running some strategy stuff, so that was cool.
2:17
Unidentified speaker (speaker_1): Yeah, that's awesome. And from the sounds of it, it was a pretty good session.
2:20
Unidentified speaker (speaker_2): Yeah. Lots
2:21
Unidentified speaker (speaker_1): to unpack for them.
2:22
Unidentified speaker (speaker_2): Yeah, no, so much opportunity.
2:25
Unidentified speaker (UNKNOWN): It
2:25
Unidentified speaker (speaker_2): is generally in the buckets of either. people, process, strategy, or structure. It was great to spend
2:31
Unidentified speaker (UNKNOWN): a bit of
2:31
Unidentified speaker (speaker_2): time there. They just got a new office, great vibe.
2:34
Unidentified speaker (speaker_1): That's cool. And last week we had Investfest as well, a couple of days down the Gold Coast.
2:39
Unidentified speaker (speaker_2): Yeah, we did.
2:40
Unidentified speaker (speaker_1): We had some good pitches. Out of the five,
2:43
Unidentified speaker (UNKNOWN): I
2:43
Unidentified speaker (speaker_1): say there's three that we're following up with. So that was really good.
2:46
Unidentified speaker (speaker_2): Yeah, especially one of them that most people would consider to be like a done and dusted old school industry. Yeah, no. Seeing the way that they're gonna change that and modernize it, awesome.
2:59
Unidentified speaker (speaker_1): And it's you can tell founders when they sit down and they know their numbers and they speak straight to distribution and they've got they know their metrics and they know what's working in their go-to-market, as opposed to some of the first-time founders that tend to fixate a bit more on the product side.
3:15
Unidentified speaker (speaker_1): who are quite early and don't really yet know their mechanics. And the sweet spot for us is those that do know the mechanics of their go-to-market.
3:21
Unidentified speaker (speaker_2): Yeah, they certainly showed that repeatable wheel. Good afternoon, Jack.
3:27
Unidentified speaker (UNKNOWN): Hey
3:27
Unidentified speaker (speaker_0): lad, sorry. It's been all hands on deck this morning.
3:31
Unidentified speaker (speaker_2): How the kids?
3:32
Unidentified speaker (UNKNOWN): Yeah,
3:33
Unidentified speaker (speaker_0): everyone's good. Got got got 'em them all out in the car and off to school. So Anna's on our way.
3:37
Unidentified speaker (speaker_1): Oh, that morning rush. That
3:38
Unidentified speaker (speaker_2): feel feeling
3:39
Unidentified speaker (UNKNOWN): of
3:39
Unidentified speaker (speaker_2): relief that then So you've
3:41
Unidentified speaker (UNKNOWN): now
3:42
Unidentified speaker (speaker_2): got six hours of functionalness. Yeah,
3:45
Unidentified speaker (speaker_0): yeah. Sanity.
3:47
Unidentified speaker (speaker_2): Yeah. Let's jump in. I thought I'd just touch on the John Hampton podcast with Joe Walker. Have you had a chance to listen to it?
3:55
Unidentified speaker (speaker_1): Part way through so far. It's it's on the schedule for the drive home today.
3:59
Unidentified speaker (UNKNOWN): Yeah,
3:59
Unidentified speaker (speaker_0): Sam. Great app.
4:01
Unidentified speaker (speaker_2): Yeah. No, it's listeners will be no stranger to us constantly trotting out a line along the lines of things that are a commodity product sub to subject to competition with no switching costs that have a low marginal cost, the marginal profit always trending to zero. And that that came from a podcast of John's a few
4:21
Unidentified speaker (speaker_2): years ago that you found, Jack, and that really helped us analyze. some of the the companies we're regularly looking at. And then also just the concept of being a very small but valuable part of something very large. And that in the last few years I think has helped
4:40
Unidentified speaker (speaker_2): keep us out of quite a lot of crap and get us into a few things that that others didn't like, but we did.
4:48
Unidentified speaker (speaker_0): I'd love to I'd love to back test all the episodes of Tribe Talking to get a count of how many times you've quoted that Don.
4:54
Unidentified speaker (speaker_2): Yeah.
4:57
Unidentified speaker (speaker_2): And even just in life in general.
4:59
Unidentified speaker (speaker_0): Yeah.
5:00
Unidentified speaker (speaker_1): And not just the podcast, but every mission, every offsite. I feel like in everything we do, it comes up.
5:07
Unidentified speaker (UNKNOWN): But
5:07
Unidentified speaker (speaker_0): yeah, something we were saying offline Don is like all episodes that we listen listened to of John are like the way he breaks problems and industries down, and it's just so evident. that this guy's uh first principles thinking, and it's so clear how he's thinking about it and attacking these things.
5:22
Unidentified speaker (UNKNOWN): Yeah,
5:22
Unidentified speaker (speaker_2): 100%. I'll link the episode in our show notes, but I really recommend it for anybody that wants to expand their thinking. But the point and relevance to us, we've been worried about the frontier labs and is there actually a durable competitive advantage there? And will that lead to
5:42
Unidentified speaker (speaker_2): sustained excess margins? And the line that John regularly uses is that it's a very high capital cost with a low marginal cost. And those dynamics can lead to situations where margins just cannot not be maintained and they get competed to zero. And so when I have a look at the
6:01
Unidentified speaker (speaker_2): frontier.
6:03
Unidentified speaker (UNKNOWN): and
6:03
Unidentified speaker (speaker_2): then you look at open weights and you look at the amount of capital that's flowing into it, is there going to be what everybody's expecting? And I'm not sure. I do think he made a great point that I hadn't thought about that actually for the likes of anthropic that actually just the sheer investment in order to get to a model that leapfrogs everything
6:23
Unidentified speaker (speaker_2): else, the sheer volume of capital deployed may actually be the moat. And so if you're one or two and you can just outspend so much that nobody can actually catch up, then maybe that is your moat. I'm just not seeing those dynamics with the open weights. And if you look at
6:43
Unidentified speaker (speaker_2): open weights and how they that gap between them just keeps closing, I'm not sold on the fact that that there are they're actually going to be. able to do that, but could be wrong.
6:52
Unidentified speaker (speaker_1): Yeah, I totally agree on that.
6:54
Unidentified speaker (speaker_2): And then the other thing that came up this week, it was actually Lee Ford who sent it to me. He sent me this long article on the second derivative, why nobody understands the AI boom. Now, you know me, I think everything in life, the answer lives in the second derivative, whether it's companies, family,
7:14
Unidentified speaker (speaker_2): founders, relationships, the whole the whole box and dice. And we, like many others, have often compared this tech boom to the 2000 tech boom. And even in the
7:29
Unidentified speaker (speaker_2): podcast with Joe Walker and John,
7:32
Unidentified speaker (UNKNOWN): there
7:33
Unidentified speaker (speaker_2): was quite a bit of discussion of what happened in the 2000s. But this article, it's actually a an ex-Bank of America analyst. He doesn't actually use his name, but he goes by the name of Groundbreaker and it's on Substack. And it was actually written in July.
7:52
Unidentified speaker (UNKNOWN): So
7:52
Unidentified speaker (speaker_2): it's a couple of months old.
7:54
Unidentified speaker (UNKNOWN): But
7:55
Unidentified speaker (speaker_2): I think it really well lays out it's not just about commodity products, high capital costs. He actually puts forward that. this thing has the potential to be very similar to the financial crisis in the
8:08
Unidentified speaker (UNKNOWN): in
8:08
Unidentified speaker (speaker_2): the analogy of it's actually a debt-driven real estate boom. It's just that this real estate produces tokens.
8:16
Unidentified speaker (UNKNOWN): And
8:16
Unidentified speaker (speaker_2): I really like it because it unpacks things from first principles. Let me just play, I'm gonna put at the end of this podcast. I've gone and taken the article whack. whacked it into 11 labs, and turned it into a little mini podcast that I'm gonna paste at the end. It's 45 minutes. I think it's excellent, but I'm a little bit weird and love
8:36
Unidentified speaker (speaker_2): this technical shit. But let me just play you this three minutes here, and then let's have a chat because I think it's a great analogy.
8:44
Unidentified speaker (speaker_3): Ask a portfolio manager what caused the 2008 mortgage crisis, and you will hear a tidy causal change. chain. Lax underwriting produced loans that should never have been made. Home prices crashed, borrowers found themselves underwater, they defaulted, and the securities written on top of those loans detonated. Prices fell,
9:04
Unidentified speaker (speaker_3): therefore, borrowers defaulted. It has the great virtue of sounding obvious. It is also, as a matter of sequence, wrong. It is the same error the market. is making right now about the AI boom.
9:17
Unidentified speaker (UNKNOWN): The
9:18
Unidentified speaker (speaker_3): subprime machine did not run on prices. It ran on the change in prices, and more precisely, on the change in that change, also known as the second derivative. We have spent the last 18 years describing the 2008 housing financial crisis as a shock, a black swan, a hundred-year hundred year flood,
9:38
Unidentified speaker (speaker_3): a tail event. It was none of those things. Every rate reset was contractually inevitable from the moment of origination. The defaults were not primarily caused by an exogenous macro shock, like a spike in unemployment or a recession that arrived first. They were the scheduled mathematical consequence of loans that assumed
9:58
Unidentified speaker (speaker_3): perpetual appreciation of home prices. The mortgages were built to break because housing. prices could not keep increasing forever. Once house price growth started to slow, the system broke. Be clear, it wasn't house prices started to fall and decrease. No, house price growth just had to slow for the system to
10:17
Unidentified speaker (speaker_3): break. The AI boom has rebuilt this exact structure, and the market is once again underriding underwriting the teaser. It has a reset wall of its own. a schedule of dated, contractual, non-negotiable payment shocks, hiding inside the trillions of dollars of compute contracts signed by OpenAI and other Frontier Labs since
10:37
Unidentified speaker (speaker_3): 2024. The take or pay compute contract, the instrument at the center of the AI buildout, has a structural feature that almost no one prices. Its payments do not begin at signing, they begin at delivery. A lab signs a multi year capacity commitment today, but the payments do not start until the
10:57
Unidentified speaker (speaker_3): data center is energized, the capacity is accepted, and the contractual ramp schedule commences. An interval set not by finance but by construction. Siting, powering, and filling a gigawatt scale campus takes 24 to 36 months from signature, mirroring the two to three year teaser of a subprime ARM.
11:17
Unidentified speaker (speaker_3): More than $2.3 trillion of compute contracts now sit on the books of the four largest American cloud providers as remaining performance obligations and contracted backlog. Signed, celebrated, capitalized into equity prices, and, critically, not yet billing. During the teaser period, everyone
11:37
Unidentified speaker (speaker_3): wins. The seller reports backlog growth that compounds at rates no operating business has ever sustained. Oracle's RPO grew 363% in a single fiscal year. The buyer, a Frontier lab burning cash at historic rates, books no expense because the capacity does not yet exist. The market
11:57
Unidentified speaker (speaker_3): capitalizes the booked number as if it were revenue and ignores the billed build number as if it were a technical. technicality. And then, on a schedule fixed at signing, booked compute becomes build compute. The take or pay clock starts. From that day forward, the Frontier Labs and the Hyperscalers incur those costs regardless of utilization.
12:17
Unidentified speaker (speaker_3): The invoice is a function of the contract, not of demand. That is the reset. The parallel to 2006 is exact, and it explains the single most cited absurdity of this cycle. How does OpenAI, a company with some $40 billion of run rate revenue, sign $1.4 trillion of
12:37
Unidentified speaker (speaker_3): compute commitments?
12:39
Unidentified speaker (UNKNOWN): The
12:39
Unidentified speaker (speaker_3): same way a household with $60,000 of income signed a $600,000 mortgage. Because the terms at signing do not require the payment yet, and because everyone at the table, borrower, lender, and the market, believes the growth will arrive before the payment does.
12:56
Unidentified speaker (speaker_2): The the article isn't trying to be exact in terms of, hey, the world's going to fall in, or hey, on this particular date it's all going to. But it is trying to explain how derivatives work and it is trying to explain a series of things that would need to be in place for effectively the Neo Labs. and the Frontier
13:16
Unidentified speaker (speaker_2): models to be able to fly out of this. And if there's a situation where those things don't end up being true, and they're pointing to the second derivative, it paints like a series of stage gates to watch out for
13:31
Unidentified speaker (UNKNOWN): and
13:31
Unidentified speaker (speaker_2): to be able to track. And what's really interesting is that it in the full in the full article, you'll see that even the subprime.
13:39
Unidentified speaker (speaker_2): crisis and everything that was just laid out there, it all it just explains it quite simply. But it explains where in 2006 and 2007 people were writing about this. It was in the media. It was but like the market ignored it. The the vast majority of people ignored it, and then the thing blew
13:59
Unidentified speaker (speaker_2): up. And they're putting forward that uh a similar thing. is happening happening here. Now, why is it relevant to us? We don't operate in a vacuum. Capital is a function of what's happening in the environment. And we do have to be looking out into the future of what is the season we're going to be going into, what
14:19
Unidentified speaker (speaker_2): is the regime that we're going to be going into. And the one that we're currently in at the moment is this
14:26
Unidentified speaker (UNKNOWN): just
14:27
Unidentified speaker (speaker_2): unbelievable growth rates that are going to go forever. Everybody's planning on AI is just all is going to deliver value. But in my experience, the universe conspires.
14:37
Unidentified speaker (UNKNOWN): And
14:38
Unidentified speaker (speaker_2): as John points out in the podcast and did a really good job of people conflate like world-changing and society-changing society changing technology with the ability to actually produce a profit. And one of the examples. he used quite well is solar panels. Anybody that went into solar panel manufacturing, which is changing the world and continues to do
14:58
Unidentified speaker (speaker_2): so and continues to be an amazing technology producing power. But at the panel level, it's a catastrophic destruction of capital.
15:07
Unidentified speaker (UNKNOWN): I
15:07
Unidentified speaker (speaker_1): love this article, Don. This is a great, bloody great find. I love just the way it unpacks true causality of something. And
15:14
Unidentified speaker (UNKNOWN): just to
15:14
Unidentified speaker (speaker_1): your point there about the reason why this matters to us, I see this same type of it's like a rising tide causes us to ignore the signals of future consequence. And we see this play out with founders taking on debt, with valuations that they believe that they will grow into, with capital deployments by
15:34
Unidentified speaker (speaker_1): investors, with like assuming that growth will eventually. equate to profit at some point. And the line that you always say, Don, not all growth is healthy. And for me, there's that this type of thinking where you actually go back to what is true causality, apply a bit of first principles thinking and the second and third order
15:54
Unidentified speaker (speaker_1): consequences of decisions. I think this is something that founders need to adopt and investors need to adopt a across all decision making. There's one line in this one I really like, which it was the invoice is a function of the contract, not of demand. And
16:08
Unidentified speaker (UNKNOWN): I
16:08
Unidentified speaker (speaker_1): love that thinking around like the the cost base that people are taking on that is like disconnected to actually the demand side of their business.
16:18
Unidentified speaker (UNKNOWN): I
16:18
Unidentified speaker (speaker_1): I feel like I need to read this one two or three times to fully take in everything that's in there because it it is there's a lot in this.
16:25
Unidentified speaker (speaker_2): The forty 45 five minute audio version at the end of this podcast, I've listened to three times just to get my head around it. And each time I keep learning some things. But what I've been I've just been concerned about the debt being used in like essentially pre-revenue Neolabs. And then if you look at even at the frontier levels or other token
16:45
Unidentified speaker (speaker_2): generating companies, like the levels of debt that are being taken on there, and then even the hyperspection. hyperscalers. Like we're in this phase where they're taking on debt. And the thing that fucks everything is always debt upon debt. And I haven't been able to really understand all of it. But the way this article, I'll just share my screen here
17:05
Unidentified speaker (speaker_2): for the online listeners or watchers. So what he does is he takes this this chart and just shows that on on the bottom line is delinquencies in the housing crisis in the US. And when they started to tick up was when the rate of change of house
17:24
Unidentified speaker (speaker_2): price growth started to dip down. And he just points out that even though the rate of change is coming negative, as long as it's above zero, So the housing price appreciation is still positive.
17:37
Unidentified speaker (UNKNOWN): But
17:37
Unidentified speaker (speaker_2): that's when you start started to. to see the delinquency started to tick up. And so in there is this chart just demonstrating the actual level of something. So in this in that case, it was house price. In this case, it could be, I don't know, the valuation of OpenAI or the valuation of anthropic. And then there's the
17:57
Unidentified speaker (speaker_2): velocity, which is either the growth rate is either positive or it's negative. But then there's the rate of change. And there's just a point where the rate of change has turned negative, but it's still above zero, which means that your velocity is still positive. So things still look good. And there's still these
18:17
Unidentified speaker (speaker_2): announcements of how positive everything is, but it's that rate of change, that second derivative where it shits the bed. And with our portfolio companies, whenever they send their monthly or quarterly updates, I just drag them into Claude. And the thing that I track on all of them is just the second derivative. And it has led to a little bit of
18:37
Unidentified speaker (speaker_2): conflict with a couple of portfolio where I'm the negative dickhead all the time because I'm preach screeching about no, you're decelerating now. Yes, but we grew. We're still growing. But yes, we're decelerating.
18:51
Unidentified speaker (speaker_2): And that's that doesn't really matter because there'll be points in time where you accelerate and de exc and decelerate. But then if you're wanting to sell as an example, or if you're in the middle of a sales process and you're decelerating, that's a fucking bad place to be. But just not a lot of the population track that second
19:11
Unidentified speaker (speaker_2): derivative.
19:12
Unidentified speaker (speaker_1): And the ability to see signals that others don't. much earlier that could be potentially fatal. And this article, it made me think about that plane crash. Was that last week? It was last week. Adam
19:24
Unidentified speaker (speaker_2): Robinson, fuck that effect. That that took
19:26
Unidentified speaker (UNKNOWN): the wind
19:26
Unidentified speaker (speaker_2): out of my sails.
19:28
Unidentified speaker (speaker_1): I can't stop thinking about it. And you go reading everything on it, there's some choices that occurred. It sounds like they went through precipitation that was wet on the wing, which then led to icing, which then, as they went to climb, the the differential air pressure likely pulled off the elevator.
19:47
Unidentified speaker (UNKNOWN): Now,
19:47
Unidentified speaker (speaker_1): there's a point at which they're flying thinking everything is fine, but actually it's all the the damage has actually already occurred. It just hasn't been realized yet. And I think that's what we're seeing in And these types of things. There's this potential for a catastrophic event. There's some signals. When it all unfolds, there'll be likely causality pointed to that actually isn't the
20:07
Unidentified speaker (speaker_1): true causality. And so it just reminded me of that. Like
20:11
Unidentified speaker (speaker_2): the second derivative, right? Like it just second derivative is always where things are at. Even the second derivative of why a founder does what the why a founder is behaving a certain way. It's never the outside thing, it's the underlying thing. Even in relationships, you have. your arguments and your fights, but the second derivative is generally what's
20:31
Unidentified speaker (speaker_2): around things like care and closeness and respect and power and control. Those are the second derivative areas. And that's where you should be trying to place your efforts. And that's the pattern
20:43
Unidentified speaker (UNKNOWN): that
20:43
Unidentified speaker (speaker_2): I see. And I think why that's important is that one of the little bellwethers that if you listen to the full 45 minutes at the end of this episode, is that OpenAI, he separates OpenAI from anthropic and explains why anthropic is a much stronger beast and why OpenAI is a much weaker beast.
21:03
Unidentified speaker (speaker_2): And one of the sort of signposts in the future is the rate of change of the valuation that's required to actually fund these things. And one of the bellwethers is when OpenAI gets into a position where it can't complete these, can't complete some of the raises. And did you see they went out for
21:23
Unidentified speaker (speaker_2): 30 billion at 1.4 trillion and they ended up with 11 billion at 1.2 trillion? Now, it's still 1.2 trillion, and maybe that number's wrong and I need to check it, but it's still a bit above the last foul, if I'm not mistaken. We haven't gone, we're not doing down rounds yet. But when you look at
21:43
Unidentified speaker (speaker_2): the the last successive rounds, the last five rounds, it's a decelerating increase. And then when you then have an accelerating compute commits, something's going to give. Now, the bull case, I think, would be that we are just growing, like we haven't even touched.
22:02
Unidentified speaker (speaker_2): touched demand yet. And the demand that's going to be happening in this space, it just means that we're going to be able to deal with all of this. Don't worry about
22:13
Unidentified speaker (UNKNOWN): it. But
22:14
Unidentified speaker (speaker_2): we're also dealing with a technology where the token cost is continuing to fall regularly. And in But most technologies, the cost and the time required to create something always drops. And it always drops significantly. And then we've also got what we're seeing with some of the neo clouds in the data
22:34
Unidentified speaker (speaker_2): centers, where at least whether it's regulatory or NIMBYism or whatever it might be around the world, we've got power shortages, we've got people not wanting to have it in their backyard. There's at least going to be a short short-term term blip. Like all of these things happening at once is just a little little risky. Now, it was more ars than class. I
22:54
Unidentified speaker (speaker_2): just got a little bit suspicious a little while ago. So I luckily had a few uh puts on the SP that gave me a few dollars the other day for some ice cream for the kids. But yeah, you can't really predict the future, but you're just trying to look for markers of where we may be and what regime we're ending, we're getting into.
23:12
Unidentified speaker (UNKNOWN): I
23:12
Unidentified speaker (speaker_0): suppose the other thing you we haven't really spoken about. the assumption as well is more compute equals more intelligence. Like it's still the assumption that we keep drawing money to then the intelligence of the models keep getting better and better and the rate of change of the models keep growing in the same way. But
23:27
Unidentified speaker (UNKNOWN): I think
23:28
Unidentified speaker (speaker_0): that's a huge assumption there. If the rate of change stops and the tech doesn't turn out to be AGI, which we're all moving towards, then there's a huge kick there right? And that's not price into all of the demand.
23:41
Unidentified speaker (speaker_1): That's a good point. Like we haven't seen the S curve S yet.
23:44
Unidentified speaker (UNKNOWN): But
23:44
Unidentified speaker (speaker_1): presumably at some point it will. You're right, that's not being factored in. They
23:48
Unidentified speaker (UNKNOWN): just assume
23:48
Unidentified speaker (speaker_1): there'll be another compute multiplier.
23:50
Unidentified speaker (speaker_2): But the idea that there is the even the S curve, one of the most dangerous sayings is always it's different this time. There is no S-curve S curve this time. It's AGI.
24:01
Unidentified speaker (speaker_0): And like I actually don't know. That's the hard part. Like, we don't know. That's the hard part.
24:05
Unidentified speaker (speaker_2): Yeah.
24:06
Unidentified speaker (speaker_1): Growing
24:07
Unidentified speaker (UNKNOWN): up with
24:07
Unidentified speaker (speaker_1): Renee Rifkin, it's never different this time. So I've I've got that ingrained that it's gonna it's gonna happen. It's gonna it's gonna hit the S-curve.
24:15
Unidentified speaker (speaker_2): Yeah, you gotta play on the field. But the bull case is that we are so early in the demand, and I had my parents over on the weekend and I set dad up with instinct. And mum and dad don't have any of this stuff. They they have chat GPT, but it's it's fairly random. But when you have a look at a higher percentage of token usage is coming from agents,
24:35
Unidentified speaker (speaker_2): yet almost nobody is using these things yet. But at the same time, is it actually going to be expensive Frontier Lab tokens or will it be open weight? But Jackie Villings of Air AirTree, Tree, I think her post was relevant. We don't tell enough stories of how AI is good for consumers. How about this? On the tram this morning, I got
24:55
Unidentified speaker (speaker_2): my Grokbot to check if it could save me money on my gas and electricity. It went through my emails, found my current rate, checked it against what's available. In five minutes, it found the rate of over a thousand dollars cheaper a year. Within the hour, I'd switch. Big corporates rely on friction to keep existing customers on worse rates and increasing profits.
25:14
Unidentified speaker (speaker_2): Agents can work through this friction tirelessly, eroding corporate rents and increasing disposable incomes for consumers.
25:23
Unidentified speaker (speaker_2): One, like that use case is
25:28
Unidentified speaker (speaker_2): exactly what we're gonna start to see roll out. And even with Gropbot now, its ability to have my credit card and actually start doing things is really impressive. Then there's the potential erosion of those margins. And suddenly when switching costs come down because you've got this thing that can just
25:48
Unidentified speaker (speaker_2): tirelessly work in the background. There's just there's so many different things floating on the horizon that our economy and our world is going to change. We're going to see token usage go through the roof, which I think is true, but will it actually be high expense frontier models versus just using an open weight
26:08
Unidentified speaker (speaker_2): model?
26:08
Unidentified speaker (UNKNOWN): Yeah,
26:08
Unidentified speaker (speaker_0): I saw Jackie's post over the weekend, ran my bills through
26:11
Unidentified speaker (UNKNOWN): it
26:11
Unidentified speaker (speaker_0): and got to save 700 bucks. So I'm stoked.
26:15
Unidentified speaker (speaker_2): Yeah. As soon as I saw that post, I'm like, fuck yeah, let's just get the agents going. What
26:20
Unidentified speaker (UNKNOWN): it
26:20
Unidentified speaker (speaker_2): has highlighted is the amount of subscriptions that I'm on that I'm not getting value. from. And it's an embarrassment.
26:29
Unidentified speaker (speaker_0): Can you tell the listeners are you still getting those things delivered to your house for the endurance athletes? Oh,
26:34
Unidentified speaker (speaker_1): the gels.
26:35
Unidentified speaker (speaker_2): Yeah, there was that point in time when I just couldn't work out how to turn off the auto payment of these magic gels. I
26:43
Unidentified speaker (UNKNOWN): think
26:44
Unidentified speaker (speaker_0): Don, you're the you are the great ICP for many subscription businesses. Forgets
26:48
Unidentified speaker (UNKNOWN): about
26:49
Unidentified speaker (speaker_0): what he's paid
26:49
Unidentified speaker (UNKNOWN): for.
26:50
Unidentified speaker (speaker_2): It says
26:51
Unidentified speaker (speaker_0): he's going to change it, but six months down the track he's still paid.
26:54
Unidentified speaker (speaker_2): Yeah. And there's there's more and more companies building their own harnesses for their own coding tools, making their own cowork, lone claw code, just connecting into open router. And as that becomes easier, again, I don't know if I'm going to ever need for a lot of things that I do the intelligence of hardcore
27:13
Unidentified speaker (speaker_2): frontier model tokens that can cure cancer for cancelling my various subscriptions, that can run on a much dumber model.
27:23
Unidentified speaker (speaker_0): Something it has sort of made me aware of. Like since using instinct and even using the gruckbot to compare your insurance, margin going forward has to be squeezed because now it's not the human made. making the decision. So for example, I'm using instinct after you said
27:43
Unidentified speaker (speaker_0): sign up, Don, and I've got tracking what we like, the big presents we want to buy for Christmas for the kids. So it's just like I just got the notification then saying all things unchanged. And it's just going to track between now and closer to Christmas. So hopefully Black Friday. And then as soon as the sale comes, bang, that's when I know like I'll go onto the website and buy it. Where before like there
28:03
Unidentified speaker (speaker_0): was no way I would have had the patience to check every day. So I think like that's just something to keep in mind, right? Like it's not the human making the decision. And the thing that Salmon Beard gave the an not the analogy, the example of creatine. Like he used to always buy the A grade creatine from Germany, but Claude found pharmaceutical grade for one tenth of the price but had to buy three kilos.
28:22
Unidentified speaker (speaker_2): Yeah, you're so right. Like the all that marketing guff and fancy photo photos and branding and all of that shit. It's again, it's gonna be another massive change to the way the economy and humans operate. And so many people are not anywhere near using this technology. But also a lot of the stuff
28:42
Unidentified speaker (speaker_2): you just described there can be done with a almost free open weight model where the marginal cost is effectively the electric the electricity and the GPU that it's running on.
28:52
Unidentified speaker (UNKNOWN): So
28:53
Unidentified speaker (speaker_0): I yeah my worry is I don't think many brands have are preparing for this. Because like they I don't think they have the margin to actually sustain the compression. So if you're marketing on the fact that you've got all these fancy little things that humans like when they do the purchasing decision where an agent doesn't And care. So I yeah, I'm just interested to see where this
29:12
Unidentified speaker (speaker_0): ends and how brands have to adapt. It's got to be lowest competitor wins.
29:17
Unidentified speaker (speaker_1): 100%. Like you see the role of marketing in B2C in the future. Like marketing, doesn't it just become irrelevant other than building credibility for an agent to trust?
29:29
Unidentified speaker (UNKNOWN): Like,
29:29
Unidentified speaker (speaker_1): what else is the role of marketing in that environment where, like, you've got Jack, like you've just got an agent just monitoring, and then you
29:36
Unidentified speaker (UNKNOWN): just
29:36
Unidentified speaker (speaker_1): buy on the cheapest price for what is otherwise a competitive. commoditized product. Like,
29:40
Unidentified speaker (speaker_0): yeah.
29:41
Unidentified speaker (UNKNOWN): It
29:41
Unidentified speaker (speaker_1): was, I just you made me think the the event on the Gold Coast last week, it was weird to have a panel chatting about AI, different founders, and then having Simon Beard in the audience. I felt like that should have been flipped around because his insights would have been phenomenal. Not to take anything away from the founders that spoke, and that they all spoke very intelligently.
30:00
Unidentified speaker (UNKNOWN): But
30:00
Unidentified speaker (speaker_1): I would have loved to have heard from Simon Beard at that event.
30:03
Unidentified speaker (speaker_2): There's a lot of change happening, and people that just aren't the in thing anymore. And last week Canva came out, and I think it was fair to say it was a shock, the sheer size of it, of cutting 22% of their workforce. How they've gone and handled it and what they're offering is got a lot of positive comments, but embedded in
30:23
Unidentified speaker (speaker_2): there's a few things that you can tell in trying to get people to. move back to the office or move country that has his the offer, but they know that a lot of people are not going to be able to take it up. And then Heidi followed suit as well, even though it's still on the up, it's gone and cut staff heavily. Yeah. It's it if you think of rate of change,
30:43
Unidentified speaker (speaker_2): this accelerating loss of tech jobs is going to be very interesting.
30:49
Unidentified speaker (speaker_0): Personally for Canberra Canva, the way they did the layoff, it was great. But in my eyes, it's still peacetime. And they're offering the $10,000 zero dilution to anyone that wants to start their founders. To me, that's peacetime. And personally, I think it should be wartime for Canva. And you look at all the stats coming out, their growth is nearly flatlined.
31:09
Unidentified speaker (speaker_0): They've had to lay off 20. Was that enough? I don't actually know. Do they there's so many moving things here and on the edge. outside, it still seems way too much peacetime. But in my eyes, it's got to be war time for them to reaccelerate growth and
31:21
Unidentified speaker (UNKNOWN): get
31:21
Unidentified speaker (speaker_0): an outcome that they should deserve.
31:23
Unidentified speaker (speaker_2): Having had to unfortunately preside over a number of staff reductions, the first thing I thought was even though 22%
31:31
Unidentified speaker (speaker_2): seems like a big number, my initial reaction is it's never enough. And then Jason Lebkin came out with a post that said, this is brutal. to say, but almost every investor that's been doing this for a while has seen this and will agree. Almost every layoff is too small. Layoffs suck. They're the failure of management, period. Period, management's fault,
31:51
Unidentified speaker (speaker_2): period. But almost every time they're too small. And certainly the ones that I had to preside over because I had my head up my ass, they were too small. And the problem is if you don't cut once and cut hard. and you end up in a situation where you've got to continue to do it, you kill culture, you kill productivity. The best people fuck off and leave anyway, and you
32:11
Unidentified speaker (speaker_2): then spiral into the ground.
32:13
Unidentified speaker (speaker_1): Totally agree. It was interesting. Listen some of our longtime listeners who were going through our back catalogue recently were saying we actually predicted a lot of what's unfolding with Canva some time, some time ago,
32:26
Unidentified speaker (UNKNOWN): which goes
32:26
Unidentified speaker (speaker_1): back. to the canary and the coal mine type stuff where you start seeing patterns and you start seeing these things that are likely to play out.
32:34
Unidentified speaker (speaker_2): There's an inevitability to the organizational lifecycle aspects. So there's not it's not really a magic trick. It's just having some frameworks and some tools. My concern for them is the entire VC industry is predicated on significant amounts of r capital return from the paper markups.
32:54
Unidentified speaker (speaker_2): I don't personally see how they can ever list unless they bending spoon themselves. But then that's a fraction of their vow. And the product itself, even if you bring in a much better model than say Claude Clawed and OpenAI and whatever else, their product is no longer in people's workflows.
33:14
Unidentified speaker (speaker_2): And so unless they can solve that structural issue, I'm quite nervous for them.
33:19
Unidentified speaker (speaker_0): Yeah, agreed. Like we, yeah, just to reiterate, like we're 100% aligned and backing Canberra even though we have no exposure. Bardons, one or two shares.
33:29
Unidentified speaker (speaker_2): Multiple shares, thank you, Jack. Multiple shares in paper.
33:32
Unidentified speaker (speaker_1): I was gonna say it's not one or two, it's multiple. Michael.
33:36
Unidentified speaker (speaker_0): But like we need it, like our ecosystem needs it. this. And like Blackbird needs it, the super funds, we all need this.
33:42
Unidentified speaker (UNKNOWN): The
33:43
Unidentified speaker (speaker_0): the issue is, like, I don't have a solution for him, but I still think Mel Cliff and the other founder, they're amazing founders. So yeah, I'm hop I hope they find the but hope.
33:55
Unidentified speaker (speaker_2): Pray for miracles, but don't rely on 'em. on them.
33:58
Unidentified speaker (speaker_0): Agreed. But I just don't want to be portrayed as like we're sitting in the cheap sheet seats just piling on camera.
34:03
Unidentified speaker (UNKNOWN): No, that's
34:04
Unidentified speaker (speaker_0): not the case. Like we're actually behind them.
34:06
Unidentified speaker (speaker_2): But one of the key points that was made, and there's been made by another a number of companies, is just that what happens when you get bloated and you have different layers of management all in different locations, that ability to manage the change loop just grinds to a halt. The ability to actually identify problems, make decisions around them, and get them in.
34:27
Unidentified speaker (speaker_2): implemented during wartime is just really difficult when you've got a rate of change of growth decelerating. And so I think that's maybe why Heidi and other growing successful in Inverter Commerce companies are still doing it, is because you just got to be so careful that you don't grow bloated and that you are able to keep
34:47
Unidentified speaker (speaker_2): that change loop in the center, which is just change continues to happen. like what has happened with Canva. That's going to create problems and opportunities that they've got to capitalize on. And it's fair to say they just haven't been able to deal with a number of the changes that are being thrown at them. And I bet you one of them is too many layers of
35:06
Unidentified speaker (speaker_2): management spread all around different areas. And then there's no incentive for those people because why would they want to? be Turkeys voting for Christmas? And so sometimes you've got to come in, chop off a few heads, make yourself more efficient so you can start dealing with that change loop.
35:22
Unidentified speaker (UNKNOWN): The
35:23
Unidentified speaker (speaker_0): other point as well is that for Heidi, they announced the redundancies, but they also announced that two of the original three co-founders departed all in the same week. That
35:33
Unidentified speaker (speaker_2): was interesting.
35:34
Unidentified speaker (speaker_0): Yeah. So I just wonder is there something else going on that we don't know about yet? Because if this is the company, like by the reports, is growing faster than Canva was from year one. And like it's a phenomenal business every time I go to the doctors, every time I give permission to use Heidi.
35:51
Unidentified speaker (speaker_2): Yes. But I wonder if you think of rate of change, peop insiders tend to know when their rate of change is now falling off a cliff. Now you can still be positive,
36:01
Unidentified speaker (UNKNOWN): but
36:02
Unidentified speaker (speaker_2): it's going to be very gonna interesting, Jack, to your point, in the months to come. Because what we're seeing with this AI boom is not only lots of competitors at the Heidi level, but some of our portfolio are suffering that the companies, like the systems that Heidi wants to integrate with or that are the incumbent Incumbents
36:21
Unidentified speaker (speaker_2): installed base inside a lot of the medical profession, they can now use AI coding. in their internal teams to build their own tools. And so suddenly you actually have two waves. One of like for like competition, and two,
36:35
Unidentified speaker (speaker_2): the incumbents are no longer in Hamilton Helmer's terms. There's the counterpositioned. Counterpositioning is becoming a lot easier or harder, depending on what you want to describe it. And a number of our portfolio are founding there customers who used to be so slow and laggard ish now saying, Oh, actually, yeah, we're gonna we're gonna
36:55
Unidentified speaker (speaker_2): have a crack at this ourselves. There was a good little post I thought from Annie Lau over in San Fran SF. I honestly don't know what a safe career looks like anymore, she says. Canva, Atlassian, Meta, Google, companies These people spend years trying to get into are cutting jobs too. If you're entering the workforce, the advice is confusing. Learn to
37:15
Unidentified speaker (speaker_2): code, learn AI, be more creative, develop into personal skills. Meanwhile, technical people are learning to sell and understand customers. Non technical people are learning to build. Everyone's trying to learn what the other person knows. I keep wondering what will happen when we meet in the middle. I worry about juniors. If we
37:35
Unidentified speaker (speaker_2): automate work that people used to learn on, how do they get the experience? we're asking for? The advice I feel most comfortable giving is to create your own options, build something useful, understand who needs it, learn how to get it into their hands. I think that's I think that's right. And we, I think, are trying to do that ourselves with the building of our own products and trying to
37:55
Unidentified speaker (speaker_2): stay closer to our LPs and deliver things to add value. But as someone with a three-year-old, three year old of six-year-old six year and old soon to be a seven year old, yeah, um fuck knows what they're gonna do.
38:08
Unidentified speaker (speaker_1): I've always held that view though, that the only safe career option is being a founder because then you have autonomy in a locus of control. As an employee, you could be terminated anytime, laid off or whatever, and you're in the hands of someone else's leadership. It is interesting with the boys, like my I'm just pushing them to build anything. And
38:28
Unidentified speaker (speaker_1): so far it seems to be working, but will even that be relevant? I don't know. It's bloody hard for kids today.
38:33
Unidentified speaker (speaker_2): I think I wonder if these conversations have happened at every revolution
38:38
Unidentified speaker (UNKNOWN): this one just
38:39
Unidentified speaker (speaker_2): having quicker like whether you're a founder or an investor, as long as you've got some way to keep inside the capital machine and and deploy.
38:47
Unidentified speaker (speaker_1): Yep.
38:48
Unidentified speaker (speaker_2): But now some people in S SF are bullish on AI and they want to what's that term? Virtue signal how strongly they support AI. I saw this, which just made me chuckle at how fucking stupid humans are.
39:02
Unidentified speaker (speaker_1): So many AI researchers have started smoking as a way of signaling to other researchers how much they believe in AI. I believe
39:09
Unidentified speaker (UNKNOWN): so much.
39:09
Unidentified speaker (speaker_3): That makes no sense. What do you mean?
39:11
Unidentified speaker (UNKNOWN): I believe so
39:11
Unidentified speaker (speaker_1): much in AI that it's going to cure cancer. So I will I I'm showing you my chips are all in. So there is kind of this mass psychosis social delusion around it.
39:20
Unidentified speaker (speaker_3): Uh huh.
39:20
Unidentified speaker (speaker_1): So
39:21
Unidentified speaker (speaker_2): yeah, you're so bullish on AI that you've taken up smoking. That makes sense.
39:25
Unidentified speaker (speaker_1): Oh fuck, that's random. All
39:27
Unidentified speaker (UNKNOWN): right,
39:27
Unidentified speaker (speaker_2): Jack, theme of the week.
39:28
Unidentified speaker (UNKNOWN): I
39:28
Unidentified speaker (speaker_0): suppose part of the conversation today has been LLM's cannibalizing business models. It's been a little bit doom and gloom, but a conversation I had this morning in the gym with someone who works in MS. MA who's pivoted to property. And we were just talking about like how great LLMs are at doing the mundane bullshit, e.g., editing
39:48
Unidentified speaker (speaker_0): PowerPoints and formatting word docs. And we just both agreed and said, What a waste of time that was. And thank God that's been taken over and we no longer have to do that.
39:57
Unidentified speaker (speaker_1): That is true. My my theme of the week is around is want should. and being honest about the structure of your life in terms of then what expectations are just going to be unmet. And so for us four kids, a puppy, two businesses, it's like
40:18
Unidentified speaker (speaker_1): Jesus
40:18
Unidentified speaker (UNKNOWN): Aaron, you've
40:19
Unidentified speaker (speaker_2): gone dark. What's it yeah
40:20
Unidentified speaker (speaker_1): but Jack,
40:21
Unidentified speaker (UNKNOWN): do we need to do
40:22
Unidentified speaker (speaker_2): a work a a welfare check here?
40:25
Unidentified speaker (UNKNOWN): No,
40:25
Unidentified speaker (speaker_1): but it's actually positive when you recognize recognise this is our structure. And therefore, if I let
40:30
Unidentified speaker (UNKNOWN): these things I
40:31
Unidentified speaker (speaker_2): recognise and I've given up
40:32
Unidentified speaker (speaker_0): Aaron, can you say what is the is? Is it one business, two kids?
40:38
Unidentified speaker (speaker_1): It's just Maren and I. No, but it's more it's not giving up on the structure. I love the structure, but some of the narrative I've been holding on to around expectations of these family weekends and all of our kids being together and other pieces and having time for certain things. That's just not reality right now in this phase. So it is a good relief when
40:58
Unidentified speaker (speaker_1): you actually let go of all the shoulds that you're shooting all over yourself and go, here's our structure. In this structure, what are realistic expectations? And it's actually quite a relief.
41:09
Unidentified speaker (UNKNOWN): My
41:09
Unidentified speaker (speaker_2): little clawed scheduled schedule task, which every day in the morning checks the tides, checks rain, checks the wind, and then based on that is supposed to send me a little message if in the next three days there's an ideal window. And the next three days are an ideal window. However, uh
41:26
Unidentified speaker (speaker_1): Well, after
41:27
Unidentified speaker (UNKNOWN): our
41:27
Unidentified speaker (speaker_1): last
41:28
Unidentified speaker (UNKNOWN): that
41:28
Unidentified speaker (speaker_2): ain't happening
41:29
Unidentified speaker (speaker_1): out after our last boat outing, I've managed to convince Moran Muran that we should get a boat. So that's now in the plan.
41:37
Unidentified speaker (speaker_2): Instinct currently is tracking a couple. of auctions. I've
41:41
Unidentified speaker (speaker_1): got some
41:44
Unidentified speaker (UNKNOWN): Yeah.
41:44
Unidentified speaker (speaker_2): The problem is that the is bucket jack is where the money lives. That's the problem.
41:49
Unidentified speaker (speaker_1): Yep. Yep. Yeah.
41:51
Unidentified speaker (speaker_2): Hi listeners. Unfortunately there was an error with the Riverside recording system. So we've actually lost a chunk where Jack explained to us what was happening with Jensen Wang and the trouble that he found himself in. and my theme of the week. So it then just transitions to charts, which for new listeners, charts are just
42:11
Unidentified speaker (speaker_2): shitty charts where we try and find interesting data sets that have been converted into something visual, often poorly explained.
42:19
Unidentified speaker (UNKNOWN): And
42:20
Unidentified speaker (speaker_2): unfortunately, there's no soundtrack for Jack for the rest of the episode, so you'll have to catch him at the next one.
42:26
Unidentified speaker (UNKNOWN): So here's
42:26
Unidentified speaker (speaker_2): the midterms in one picture. So this is move towards Democrats or move towards Republicans. Have a look at how blue this entire country went, except for a couple of a few states. Like that that is a whitewash. Like that is unbelievable. So this is the gap between Between Australia the Australian f
42:46
Unidentified speaker (speaker_2): inflation rate and our inflation target, because I think it was the RBA governor came out and said, our target's higher than the rest of the world. That's why we're higher. So we are second only to Brazil, and we're just above Spain at one point one point two percent above the inflation
43:06
Unidentified speaker (speaker_2): target. But a number of people just keep pointing out that a lot of other OECD countries still have our inflation rate pain, but don't have our interest rate. problem. And that they continue to point back to government spending. This isn't really a chart, but I thought it was funny. Some people can't process things until they talk about them. Some people can't talk about
43:26
Unidentified speaker (speaker_2): things until they've processed it. Then those two people marry each
43:29
Unidentified speaker (UNKNOWN): other.
43:30
Unidentified speaker (UNKNOWN): Oh
43:30
Unidentified speaker (speaker_1): my god, this is so true.
43:35
Unidentified speaker (speaker_2): Yeah. Just hold on. I've got a couple more. This is the cumulative house price by location since respective peaks. And we're all coming down fairly, uh fairly hard. The worst being 10%, the least being four point seven percent. So we're still thanks to our good mates
43:54
Unidentified speaker (speaker_2): Albo. S P five hundred is positive 100% of the time after the midterm election since 1950. So this is the other interesting thing, right? So we've got all this shit happening, but since 1950, according to this, whether it's Republicans or Democrats, US stocks are higher six and twelve months after every
44:14
Unidentified speaker (speaker_2): midterm since 1950. If that holds, it must mean our AI trade has to hold for at least this period of time. Percentage of overbought and oversold stocks. stock, so oversold and overbought continues to be like a set of jaws. So that's going to be interesting to see how
44:34
Unidentified speaker (speaker_2): that continues. And for those at home, this is probably a pretty shitty chart because it just oscillates. I'm not sure what informational value is in there. This one is the K-shaped economy. I think this is what has driven partly the blue wave.
44:51
Unidentified speaker (speaker_2): The red line here is the capital share of income, of gross domestic income. And this is going back, it's a bit hard to read that, but this is going back many decades. And then labor share. And since COVID, the capital share has just continued to spike and labor share continues to drop
45:11
Unidentified speaker (speaker_2): down and just continues to go at record lows. whereas capital continues to go record highs. Now, the reality of AI, I think that gets worse, right? Because that is something that's going to shift. And that is when the pitchforks come out. Skip this one. I think this is interesting. This is the Federal Reserve total assets since
45:31
Unidentified speaker (speaker_2): COVID. And it dipped down, then COVID came along, and it went from three 3.77 point seven seven trillion up to eight 8.96 point nine six trillion. And it's been going down, but it's started to creep up a little bit in 2026. And it's just going to be interesting that if we do have a little bit of a
45:50
Unidentified speaker (speaker_2): bubble bursting moment, is the Fed gonna come in like it offers does with a lot of its balance sheet work? So this is a chart going back to 2017 for demand in skills for curiosity, adaptability, resilience, willingness to learn, and AI fluency. And AI fluency is three
46:10
Unidentified speaker (speaker_2): times curiosity. So it's the demand for curiosity is a skill. So don't be curious, just use the fucking LLM, will
46:17
Unidentified speaker (UNKNOWN): you?
46:18
Unidentified speaker (speaker_2): Although willingness to learn is three times up. And this is the final chart. So this is going back to 1926. Republican and Democrat and the growth of a dollar in the SP 500. And the point of this chart is that the market doesn't really give a shit who the president is
46:37
Unidentified speaker (speaker_2): or what the party is. The SP 500 has been a compounding beast. Alright, that's our charts. What are we looking forward to?
46:46
Unidentified speaker (UNKNOWN): I'm
46:46
Unidentified speaker (speaker_1): looking forward to Adelaide on Wednesday flying down to uh deliver the D. debrief event from the
46:52
Unidentified speaker (UNKNOWN): US
46:52
Unidentified speaker (speaker_1): mission. Catch up with Jack, catch up with the State Gov down there, a lot of the others in the ecosystem. So that'll be fun. Yeah.
46:58
Unidentified speaker (UNKNOWN): I'm a
46:59
Unidentified speaker (speaker_2): bit myopic at the moment. Looking forward to getting our Victoria tender in. Thank you to the awesome Lee Ford and his amazing work behind the scenes at the moment, getting us to the Greek.
47:08
Unidentified speaker (UNKNOWN): Yeah,
47:09
Unidentified speaker (speaker_1): that's been a huge effort.
47:10
Unidentified speaker (UNKNOWN): All
47:10
Unidentified speaker (speaker_2): Alright, right, lads, I've I'm got gonna to jump to this meeting that I keep getting text messages for.
47:14
Unidentified speaker (speaker_1): All Alright. right. Enjoy.
47:15
Unidentified speaker (speaker_2): All right. See
47:16
Unidentified speaker (UNKNOWN): you
47:16
Unidentified speaker (speaker_1): guys.
47:16
Unidentified speaker (speaker_3): Groundbreaker Substack, July 2nd, 2026.
47:21
Unidentified speaker (UNKNOWN): The
47:22
Unidentified speaker (speaker_3): second derivative. Why no one understands the AI boom.
47:27
Unidentified speaker (speaker_3): The market misremembers 2008. That same blind spot sits at the center of the AI boom.
47:36
Unidentified speaker (speaker_3): Part one. 1. Ask a portfolio manager. what caused the 2008 mortgage crisis, and you will hear a tidy causal chain. LAX underwriting produced loans that should never have been made. Home prices crashed, borrowers found themselves underwater, they defaulted, and the securities written on top of those loans
47:55
Unidentified speaker (speaker_3): detonated. Prices fell, therefore, borrowers defaulted. It has the great virtue of sounding obvious. It is also, as a matter of sequence, wrong.
48:06
Unidentified speaker (UNKNOWN): It
48:06
Unidentified speaker (speaker_3): is the same error the market is making right now about the AI boom.
48:10
Unidentified speaker (UNKNOWN): The
48:11
Unidentified speaker (speaker_3): SubPime machine did not run on prices. It ran on the change in prices, and more precisely, on the change in that change.
48:20
Unidentified speaker (UNKNOWN): The
48:20
Unidentified speaker (speaker_3): canonical product of the era, the 228 and 327 hybrid adjustable rate mortgage, or ARI. ARM for short, was not designed to be repaid on its stated terms. It was designed to be refinanced.
48:34
Unidentified speaker (UNKNOWN): A
48:34
Unidentified speaker (speaker_3): borrower took a low teaser rate for two or three years. That's the two in the 228. Two years as a teaser, 28 as a variable, 30 years in total. The implicit underwriting assumption, shared by originator and borrower alike, was that the loan. would never reach its reset. It would never go past this first
48:54
Unidentified speaker (speaker_3): two years and they would refinance. Rising home values would manufacture equity, the borrower would refinance into a fresh teaser, and the clock would start again. The structure was a treadmill, and the treadmill was powered by appreciation in house prices. It worked spectacularly while it worked. Nearly four in five subprime hybrid arms.
49:15
Unidentified speaker (speaker_3): originated in 2003, had been refinanced away by the end of 2006.
49:20
Unidentified speaker (UNKNOWN): Now
49:21
Unidentified speaker (speaker_3): watch the timing. National home price appreciation did not crash in 2006. It decelerated.
49:28
Unidentified speaker (UNKNOWN): The
49:28
Unidentified speaker (speaker_3): year over year rate of gain, which had run in the mid to high teens through 2004 and into early 2005, it began bleeding off. Still positive, still printing great. green, but slowing. Prices were higher than they had ever been. And yet, with prices at their peak and still rising, subprime delinquencies
49:48
Unidentified speaker (speaker_3): inflected upward. This is why the popular causal story is, in economist Didier Sornet's phrase, right mechanically, but wrong because it takes the fall in house prices as exogenous. As though the decline simply arrived once. one day, a meteor from outside the system. It did not arrive from outside.
50:08
Unidentified speaker (speaker_3): The deceleration was endogenous to the structure. The structure required ever accelerating prices to keep refinancing its way out of its own reset schedule, and no series accelerates forever. The second derivative, the rate of change, was always going to roll over.
50:25
Unidentified speaker (UNKNOWN): When
50:25
Unidentified speaker (speaker_3): it did, the first derivative followed it down through zero, negative equity spread from the margin inward, and the defaults the market insisted were caused by falling prices had in fact begun a year earlier, when prices were still rising but had stopped rising faster. Part 2. A short theory of derivatives.
50:46
Unidentified speaker (speaker_3): Let the relevant quantity be s. Three numbers describe it. The level is s itself, how big the thing is. The first derivative is the velocity, s prime, how fast it is growing. The second derivative is the acceleration, s double prime, whether that growth is itself speeding up or slowing
51:06
Unidentified speaker (speaker_3): down. Markets are instrumented to observe the first two and almost entirely blind to the third. Cell Sell-side side models forecast levels. Momentum strategies trade the first derivative. Virtually nobody builds a position around the second derivative.
51:21
Unidentified speaker (UNKNOWN): Yet
51:22
Unidentified speaker (speaker_3): the second derivative is precisely where information about regime change lives, for a structural reason. When financing embeds a growth assumption, a reset that presumes refinancing, a covenant that presumes rising cash flow, a commitment size sized to presumed expansion. The assumption is satisfied not by the level
51:41
Unidentified speaker (speaker_3): being high, but by growth being sustained. Sustained growth at a declining rate still satisfies the headline. Revenue grew 40%, while quietly violating the embedded premise. But the incremental capacity we committed to assumed it would grow 70%. The gap between what the headline shows and what the structure
52:01
Unidentified speaker (speaker_3): needs opens. silently. There is a window, call it borrowed time, between the moment the second derivative rolls over and the moment the first derivative crosses zero. During that window, everything looks fine. Revenue is at record highs. Growth is still positive. The press releases are
52:21
Unidentified speaker (speaker_3): triumphant. And the machine is already broken. It simply has not been told yet. Part 3. The AI boom is a credit-driven real estate cycle. Open any AI bull or bear note and observe what it argues about.
52:36
Unidentified speaker (UNKNOWN): It
52:37
Unidentified speaker (speaker_3): argues about levels, how many billions of revenue, how many gigawatts, how large the total addressable market, and about the first derivative. Is growth 200% or 150%? Is enterprise inflection? inflecting? Are tokens per minute rising?
52:53
Unidentified speaker (UNKNOWN): The
52:53
Unidentified speaker (speaker_3): bears say the levels are unsustainable. The bulls say the growth justifies them. Both camps are staring at S and S prime. Neither is watching S double prime. And the entire financing architecture erected over the past 24 months is a bet on S double prime, on the acceleration of demand continuing.
53:14
Unidentified speaker (speaker_3): Dressed up, as a bet on the level. This matters because the dominant analogy in everyone's head is the wrong one. Is AI a bubble like dot-com? Is the question being asked, and it produces dot-com answers. Maybe the leaders survive and the laggards wash out. Maybe multiples compress. Maybe we get a
53:34
Unidentified speaker (speaker_3): 50% drawdown and a recovery. That framing is a category error. The year 2000 was an equity multiple event. Too much optimism priced into stocks with thin balance sheets and almost no debt. The pain was real, but it was equity pain. And equity is patient capital that can simply be
53:54
Unidentified speaker (speaker_3): marked down and held. The AI build out is structurally different. It is increasingly financed not by selling overpriced equity, but by contracting future cash flows and borrowing against hardware. Take or pay capacity deals, GPU collateralized term loans, off-balance sheet vehicles, asset back
54:13
Unidentified speaker (speaker_3): notes sold to insurers. That is not the architecture of 2000. That is the architecture of 2008. There is a deeper misclassification beneath the 2000 versus 2008 question. and it is the one that does the real damage. The market is pricing AI as a technology cycle when its actual
54:33
Unidentified speaker (speaker_3): anatomy is that of a credit driven real estate cycle, which is precisely why the 2008 mechanics apply, and the two break for entirely different reasons. Technology cycles are driven by innovation and adoption. Their risks are obsolescence and competition. They live or die on whether the product is
54:53
Unidentified speaker (speaker_3): wanted, and they can derate slowly as the future is repriced. Real estate cycles are mechanical: leverage, hard assets, occupancy, debt-financed construction at scale, commercial leases disguised as take or pay contracts, and long construction lags that guarantee supply arrives after
55:13
Unidentified speaker (speaker_3): demand has turned. Walk down the AI build-out, and every feature is a property development in disguise. A data center on entitled land, financed with debt against the structure, and leased to tenants on take or pay terms. This is not a software business that happens to own servers. It is a real estate business that happens to
55:32
Unidentified speaker (speaker_3): compute. Real estate cycles break the same way every single time.
55:37
Unidentified speaker (UNKNOWN): Not
55:37
Unidentified speaker (speaker_3): when demand collapses. it rarely does, but when the rate of demand growth decelerates against the fixed supply the boom has just finished building.
55:47
Unidentified speaker (UNKNOWN): The
55:47
Unidentified speaker (speaker_3): second derivative again, in the one asset class where it has been studied for a century.
55:52
Unidentified speaker (UNKNOWN): And
55:53
Unidentified speaker (speaker_3): the credit machine does not derate gently. It refinances or it seizes. The instruments that seize take or pay leases, GPU collapse. collateralized term loans, asset backed notes, are each negatively convex. Within this machine, the financing mechanics diverge. The PurePlay
56:12
Unidentified speaker (speaker_3): NeoClouds borrow non recourse debt against a specific tenant's take or pay and collapse into seizure when the tenant can't pay.
56:19
Unidentified speaker (UNKNOWN): The
56:19
Unidentified speaker (speaker_3): hyperscalers fund with corporate bonds and operating cash flow. For them, a tenant default means impairment and margin compression.
56:27
Unidentified speaker (speaker_3): not seizure. Oracle sits between them, corporate funded but dangerously concentrated.
56:33
Unidentified speaker (UNKNOWN): The
56:34
Unidentified speaker (speaker_3): distinction matters because the seizure and the impairment are two different wounds, inflicted by the same deceleration, and both are hiding inside the same $2.1 trillion backlog. But to price the credit quality of that book, we must look past the total headline RPO. to the core structural exposure. Across the big
56:54
Unidentified speaker (speaker_3): four platforms, the RPO that Wall Street values as forward revenue is, in reality, a concentrated credit exposure to a handful of cash burning frontier model labs and specialized AI tenants.
57:07
Unidentified speaker (UNKNOWN): The
57:07
Unidentified speaker (speaker_3): Hyperscaler has, in economic substance, extended a concentrated infrastructure credit facility to tenants with no independence. independent operating income.
57:16
Unidentified speaker (UNKNOWN): If
57:16
Unidentified speaker (speaker_3): those tenants default, the backlog evaporates into non cash impairments, leaving the corporate balance sheet to absorb the fixed costs of customized, rapidly depreciating capital assets. 4. The loan book nobody calls a loan book. Look past the compute scarcity narrative and see these agreements for
57:36
Unidentified speaker (speaker_3): what they really are.
57:37
Unidentified speaker (UNKNOWN): A
57:38
Unidentified speaker (speaker_3): frontier lab. signs a contract promising to pay a counterparty tens of billions of dollars over several years for compute it has not yet consumed. The counterparty
57:51
Unidentified speaker (speaker_3): books that promises backlog and borrows against it, raising debt to pour concrete and rack GPUs. Reduce the arrangement to its skeleton, and it is a loan.
58:01
Unidentified speaker (UNKNOWN): The
58:02
Unidentified speaker (speaker_3): counterparty advances capital in kind, a building full of chips, against the borrower's commitment to pay it back, with interest and principal amortization baked directly into the take or pay rate. The data center is the collateral. The lab's contracted payments are the debt service. And the structure performs only so long as the
58:21
Unidentified speaker (speaker_3): borrower can keep funding those payments, which, for a company with no profits, means only
58:26
Unidentified speaker (UNKNOWN): the government.
58:27
Unidentified speaker (speaker_3): so long as it can keep raising money. This is precisely why the analogy is 2008, not 2000. AI capital expenditure is not a capital budget. It is a loan book. Leases in form, debt in substance.
58:42
Unidentified speaker (UNKNOWN): The
58:42
Unidentified speaker (speaker_3): CapEx is the funded principal. The contracted backlog, the remaining performance obligation in the filings, is the lease receivable.
58:51
Unidentified speaker (UNKNOWN): When
58:51
Unidentified speaker (speaker_3): a hyperscaler or a NeoCloud reports a record capex figure, the financial press reads it as confidence, as proof of demand. Read it instead as origination volume. Each gigawatt of committed build is a loan extended to whichever tenant has signed the take or pay beneath it, and the credit quality of that loan is precisely the
59:11
Unidentified speaker (speaker_3): credit quality of the tenant. The market is celebrating loan. growth and calling it revenue growth. 5. The borrower with no income.
59:21
Unidentified speaker (speaker_3): Every subprime cycle has a borrower who could only refinance, never repay. In this one, that borrower is OpenAI. OpenAI has committed to pay for compute on a scale without precedent in corporate history. Multi-year take or pay capacity contracts. whose aggregate obligations run to the
59:41
Unidentified speaker (speaker_3): hundreds of billions of dollars. Against them sits an operating business that does not yet earn a profit. Revenue real, large, and growing quickly, but short of covering the company's own cash burn and nowhere near covering the contracted payments. Those payments are therefore not serviced out of earnings, they are
1:00:01
Unidentified speaker (speaker_3): serviced out of financing. And financing for a borrower in this position is available on a single condition: that each new round price above the last. For OpenAI, the up round is not a measure of progress. It is the funding event itself, the mechanism by which the prior period's commitments
1:00:21
Unidentified speaker (speaker_3): are paid and the next periods are made signable. The markup is the cash flow.
1:00:26
Unidentified speaker (UNKNOWN): A
1:00:27
Unidentified speaker (speaker_3): company funded by its own appreciation is solvent not in proportion to how high the mark stands, but in proportion to how fast it is still rising. Because a burn rate is an accelerating schedule, indifferent to the size of the last round. Each new phase of compute expansion demands an exponentially larger cash injection.
1:00:47
Unidentified speaker (UNKNOWN): When
1:00:47
Unidentified speaker (speaker_3): the step up compresses from 1.91 times down to 1.2 1.23
1:00:50
Unidentified speaker (UNKNOWN): times.
1:00:51
Unidentified speaker (speaker_3): times, the math breaks. The valuation can print a record at the precise moment the company's capacity to fund its structural deficit is contracting. This is not a paradox. It is the arithmetic of a borrower whose liquidity is bound to the first derivative of its own price. Measured as a level, OpenAI's valuation
1:01:11
Unidentified speaker (speaker_3): is the most remarkable appreciation in the history of private markets.
1:01:15
Unidentified speaker (speaker_3): Roughly $86 billion in early 2024, then about $157 billion, $300 billion, $500 billion, and approximately $852 billion by the spring of 2026. Measured as a rate of change, the same series inverts.
1:01:34
Unidentified speaker (UNKNOWN): The
1:01:34
Unidentified speaker (speaker_3): round-over-round round over round step up ran 1.83 times, $1.91 times.
1:01:39
Unidentified speaker (speaker_3): 1.67 times, 1.70x, and falls to roughly 1.23 times implied by the reported public offering target. Private marks are inherently lumpy, negotiated, episodic, set by a handful of insiders, so no single step is decisive. But the trend is
1:01:58
Unidentified speaker (speaker_3): unmistakable. It bends down, and it bends hardest at the one mark. set by the deepest, most unforgiving pool of capital, the public market. The implied IPO step up is both the lowest in the sequence and the hardest to negotiate, and it is the one the structure must actually clear. This arithmetic is also the most
1:02:18
Unidentified speaker (speaker_3): probable explanation for OpenAI's recent IPO PO delay. Here, the bulls raise their strongest objection. Yes, they say. the Frontier Labs burn cash now. But so did Amazon. So did every great compounding business in its infrastructure building phase. Burn is investment. The labs will grow into
1:02:38
Unidentified speaker (speaker_3): profitability. The borrower of today is the cash machine of tomorrow. Half right. There is no single Frontier Lab borrower. There are at least two, and they are not the same credit. Anthropic is a speculative grade but highly insulated credit. Its revenue is approximately 80% enterprise,
1:02:57
Unidentified speaker (speaker_3): sticky, recurring, contracted seats, and its unit economics are firmly above water, generating $1.70 of revenue for every dollar of compute. Bern converges to a manageable approximately 9% of revenue by 2027 as gross margins normalize.
1:03:15
Unidentified speaker (speaker_3): More importantly, Anthropics liabilities are protected by a classic 2008 monoline wrap maneuver. In the approximately $35 billion Apollo/slash Blackstone TPU facility, Anthropics paper borrows the rating of its investment grade backers. Google guarantees lease shortfalls, and Broadcom
1:03:35
Unidentified speaker (speaker_3): guarantees the residual value of the silicon on the approximately $31 billion senior tranche.
1:03:41
Unidentified speaker (UNKNOWN): The
1:03:41
Unidentified speaker (speaker_3): cosigners are standing behind the bills and protecting Anthropics counterparties. OpenAI is the naked borrower, making it the weakest and most volatile credit in the ecosystem. Its revenue mix is fragile, approximately 60% consumer. With Bern burn hovering at a crushing approximately 57% of
1:04:01
Unidentified speaker (speaker_3): revenue through 2027. and cumulative cash destruction marching toward $115 billion by 2029, OpenAI has no path to positive cash flow this decade. Worse, OpenAI has no real cosigner. While the market long priced in an implied Microsoft backstop,
1:04:21
Unidentified speaker (speaker_3): Microsoft stripped away every structural strut in April 2026, ending the revenue share. dropping exclusivity, and surrendering its right of first refusal to supply compute. Microsoft kept its 27% equity upside, but walked away from OpenAI's bills. Softbank, the other great OpenAI backer, is itself
1:04:41
Unidentified speaker (speaker_3): now trying to raise a $10 billion margin loan against its OpenAI Ice stake, offering a personal guarantee after lenders balked at the collateral.
1:04:49
Unidentified speaker (UNKNOWN): Even
1:04:49
Unidentified speaker (speaker_3): the cosigner. has no cosigner. Microsoft's own behavior is the signal. The most informed counterparty in the complex, the one that saw OpenAI's books from the inside for years, has recognized the credit risk. Rather than building its own compute on 15-year leases that outlast the chips, it foresaw the
1:05:09
Unidentified speaker (speaker_3): commoditization of frontier models and committed over $60 billion to NeoCloud providers. through shorter five-year capacity agreements, renting at the peak to avoid owning through the trough. That is not a bet on OpenAI's durability. It is a lender shortening duration on a borrower it has decided not to underwrite.
1:05:29
Unidentified speaker (speaker_3): The same subprime credit this section describes, priced by the party that knows it best. The critical divergence here is counterparty risk. When an investor or lessor underwrites Anthropic, they are ultimately looking through the structure to underwrite the pristine balance sheets of Google and Broadcom. The credit risk is
1:05:49
Unidentified speaker (speaker_3): synthetically lifted to investment grade. When underwriting OpenAI, there is no look through. Counterparties are exposed to a naked, standalone startup sitting on an underwater unit economic model. Without a parental balance sheet, OpenAI is entirely dependent on a continuous refinancing treadmill, servicing each old
1:06:09
Unidentified speaker (speaker_3): obligation with the proceeds of the next, larger equity raise. That is the counterparty risk hiding inside roughly half of the $2.1 trillion backlog the market has priced as bankable. 7. The reflexive flip.
1:06:24
Unidentified speaker (UNKNOWN): The
1:06:24
Unidentified speaker (speaker_3): take or pay contracts are the structural. foundation, the collateral that makes the borrowing possible. But they are not a passive constraint. The Nash equilibrium is what pushes the hyperscalers to sign those contracts in the first place, and to sign them at ever larger scales. The race does not bypass the loan book. It writes the
1:06:44
Unidentified speaker (speaker_3): loan book. To justify the next gigawatt of spend, a hyperscaler needs the next gigawatt of backlog. so it pushes its tenants to commit further forward. The $2.1 trillion backlog is not a pre existing limit on the ARMS race. It is the ARMS race's own paper trail. The contracts are the collateral. The
1:07:04
Unidentified speaker (speaker_3): race is the demand for more collateral. Once that collateral is signed, it must be converted into infrastructure before the cash arrives. That conversion, the act of turning a signed contract into a live data center, is what drives the CapEx machine. And that machine is now consuming cash faster than the backlog can
1:07:24
Unidentified speaker (speaker_3): validate it. Aggregate CapEx as a share of operating cash flow ran near 30% in 2022, roughly 42% in 2023, about 50% in 2024,
1:07:37
Unidentified speaker (UNKNOWN): and
1:07:37
Unidentified speaker (speaker_3): approximately 60%. in 2025.
1:07:40
Unidentified speaker (UNKNOWN): On
1:07:41
Unidentified speaker (speaker_3): consensus spending, it reaches 100% in 2026. Above that line, by definition, every marginal dollar of capacity is funded not from internal cash, but from the balance sheet, debt or equity. The Fortress balance sheet self-funded story is true only below 100%,
1:08:00
Unidentified speaker (UNKNOWN): and
1:08:00
Unidentified speaker (speaker_3): the consensus path crosses 100%.
1:08:03
Unidentified speaker (speaker_3): this year. The fortress is not being defended, it is being spent. Past 100% of cash flow, accelerating CapEx means borrowing more, faster, every quarter, against a rating that only has so many notches left. Hyperscaler debt issuance has to climb steeply over the coming year.
1:08:23
Unidentified speaker (speaker_3): The bond market becomes the marginal funder of the entire bill. build. Why are hyperscalers betting over 100% of operating cash flow on an uncertain return? Because until now, they have been paid to. Capital expenditure has gone vertical. Roughly $150 billion
1:08:43
Unidentified speaker (speaker_3): in 2023, $226 two hundred twenty six billion dollars in 2024, twenty twenty four, $410 four hundred ten billion in 2023. dollars twenty twenty five, an estimated seven hundred twenty five billion dollars in twenty twenty six, and approaching one point one trillion dollars in twenty twenty seven. As a level, it is the largest private capital
1:09:03
Unidentified speaker (speaker_3): formation event in history. As a first derivative, the growth rates read plus fifty one 61%, percent, plus plus eighty 81%, one per cent, plus plus seventy 77%, seven percent, plus plus fifty 52%. two per cent.
1:09:15
Unidentified speaker (UNKNOWN): And
1:09:15
Unidentified speaker (speaker_3): as a second derivative, the acceleration peaked at roughly plus 30 percentage points into 2025 and has turned negative, about minus 4 points, then about minus 25. The level is at records. The velocity is still high. The acceleration has already rolled over.
1:09:34
Unidentified speaker (UNKNOWN): There
1:09:34
Unidentified speaker (speaker_3): is a recursion here that the headline numbers obscure. Hyperscaler Hyperscalar CapEx. in this cycle is not primarily a response to AI demand. To a substantial degree, it is the demand. The lab's revenue is, in large part, hyperscalar spending recycled. Cloud credits, compute commitments,
1:09:53
Unidentified speaker (speaker_3): equity-funded consumption. Nvidia's revenue is hyperscalar Scalar CapEx. The NeoCloud's revenue is hyperscalar capex, levered. Strip out the spending and the demand it manufactures, and the organic capex independent demand is a fraction of the headline figure. Which means the single most important growth rate in
1:10:13
Unidentified speaker (speaker_3): the system is the second derivative of hyperscalar capex, and it has already gone negative while every level chart still points to the sky.
1:10:21
Unidentified speaker (UNKNOWN): The
1:10:22
Unidentified speaker (speaker_3): CapEx Arms race is a Nash equilibrium, but a conditional one. It holds only while the market rewards the next dollar of spending as a call option on growth. In that regime, the boom regime, the dominant strategy for every hyperscaler is to spend, because the alternative is to be the one player who blinked and
1:10:41
Unidentified speaker (speaker_3): seeded the future. Mutual escalation is stable precisely because the market applauds it. Each CFO spends because every other CFO is spending. and the multiple rewards the spender. Morgan Stanley caught the psychology exactly when it described 2027 CapEx estimates leaping 30%
1:11:01
Unidentified speaker (speaker_3): in a single quarter toward $1.1 trillion as the dynamics of an auction. An auction is the right frame because in an auction, the price is set by the most optimistic bidder and the act. of bidding is itself the signal, the applause, the proof of seriousness. Keynes's
1:11:21
Unidentified speaker (speaker_3): Beauty Contest with chips. You are not spending on what you think the compute is worth. You are spending on what you think the market will reward you for being seen to spend. That equilibrium is not anchored to anything physical, it is anchored to a belief, the market's reading of what
1:11:38
Unidentified speaker (UNKNOWN): you're doing.
1:11:39
Unidentified speaker (speaker_3): the next dollar of CapEx means, and beliefs reprice.
1:11:43
Unidentified speaker (UNKNOWN): The
1:11:43
Unidentified speaker (speaker_3): flip comes the first time a hyperscaler announces a capex cut, and its multiple expands on the news rather than contracting. The instant discipline is rewarded instead of punished. Every payoff on the board rewrites. Spending, formerly the dominant strategy, becomes the move that gets you punished alone. Holding,
1:12:03
Unidentified speaker (speaker_3): formally surrender, becomes the move that gets you re-rated.
1:12:07
Unidentified speaker (UNKNOWN): The
1:12:07
Unidentified speaker (speaker_3): Nash equilibrium inverts from everyone spends to everyone cuts. And because it is a coordination equilibrium, the inversion is not gradual. The first mover rewarded for cutting gives every other CFO both the cover and the incentive to follow, and discipline cascades as fast as the spending it replaces. The
1:12:27
Unidentified speaker (speaker_3): day the market cheers a cut is the day the arms race ends. Goldman's head of Delta One Trading put it as plainly as it can be put. The first hyperscaler to signal that it can slow the pace of spending will likely see its share price rewarded and will crush semiconductor stocks. If that happens, others
1:12:47
Unidentified speaker (speaker_3): will take notice. That is the reflexivity that ultimately stops. stalls the CapEx cycle. Not a lack of demand, but investors deciding that incremental returns on the next dollar of spend are no longer attractive.
1:13:00
Unidentified speaker (UNKNOWN): The
1:13:00
Unidentified speaker (speaker_3): cruelty of the flip is what it does to the contracts.
1:13:03
Unidentified speaker (UNKNOWN): In
1:13:03
Unidentified speaker (speaker_3): the boom regime, a signed take-or-pay commitment is an asset to everyone who touches it. Forward demand for the hyperscaler, bankable backlog for the neo NeoCloud, cloud, collateral for the lender.
1:13:15
Unidentified speaker (UNKNOWN): In
1:13:15
Unidentified speaker (speaker_3): the repriced regime, the identical contract is a liability for all of them simultaneously.
1:13:22
Unidentified speaker (UNKNOWN): The
1:13:22
Unidentified speaker (speaker_3): lab cannot fund the payments it locked in. The hyperscaler holds a receivable from a visibly distressed counterparty. The NeoCloud is left servicing debt against data centers it financed on a contract now worth less than the debt. This is negative convexity wired directly. into the demand side. The same instrument is an
1:13:42
Unidentified speaker (speaker_3): asset on the way up and a liability on the way down. And the transition between the two states is a repricing of belief, not a change in the underlying hardware. Nothing physical has to break. The market only has to change its mind. It lands hardest on the Frontier Labs, who can carry these contracts only by raising more
1:14:02
Unidentified speaker (speaker_3): capital.
1:14:02
Unidentified speaker (UNKNOWN): and
1:14:03
Unidentified speaker (speaker_3): the flip closes that window. What follows is not a clean default, but a negotiation. Volumes cut, schedules stretched, contracts restructured. The contracts do not vanish, they reprice, beginning with the borrower who needs the next round most.
1:14:20
Unidentified speaker (speaker_3): 8. Who blinks first?
1:14:25
Unidentified speaker (speaker_3): Every reflexive cascade. needs a first mover. So which hyperscaler cuts first? Who blinks? The instinct is to say the weakest balance sheet, and the instinct is wrong. The first to cut will be the one with the best information, the credibility to reframe the cut as strength, and the balance
1:14:45
Unidentified speaker (speaker_3): sheet room to be rewarded rather than punished for it. Zuckerberg holds dual class control. He has run this exact playbook before and was rewarded with a tripling of the stock. And of all the hyperscalers, Meta has the weakest direct monetization of its AI CapEx, no public cloud to sell the capacity into, which makes
1:15:05
Unidentified speaker (speaker_3): its spend the hardest to defend and the easiest to cut. The only reason it has not cut yet is the Nash equilibrium. Zuckerberg is waiting for the market to tell him it is. safe to stop spending. The others array predictably. Google will not blink. It builds TPUs at a structural cost advantage and reports a
1:15:25
Unidentified speaker (speaker_3): cloud backlog north of $460 billion, so it benefits if rivals retrench. Oracle cannot blink. At roughly 86% of sales going to CapEx, with a balance sheet stretched around Stargate. its stress will surface as a credit event. Amazon may have its hand forced from the other direction,
1:15:44
Unidentified speaker (speaker_3): free cash flow already turning negative under the build. Negative free cash flow is the kind of thing capital markets eventually vote on, whether management calls the election or not.
1:15:55
Unidentified speaker (UNKNOWN): The
1:15:55
Unidentified speaker (speaker_3): numbers tell the same story. Morgan Stanley pegs hyperscaler investment grade leverage at roughly 1.8 turns of gross debt, double what it was a year ago, and now higher than the entire energy sector. That figure does not count the 100 billion plus parked off the balance sheet in the vehicles. What stands in its place is a
1:16:15
Unidentified speaker (speaker_3): leveraged, hard asset, refinance-dependent balance sheet. And the marginal gigawatt, the thing cut first, is the most discretionary line on it.
1:16:25
Unidentified speaker (speaker_3): 9. The blast radius.
1:16:29
Unidentified speaker (speaker_3): Let's say OpenAI is subprime. The regime shifts, belief reprices, the capital window slams shut, and a hyperscaler cuts that marginal gigawatt to protect its own leverage. Who is exposed? OpenAI is the single largest customer, by direct contract or one counterparty removed, of
1:16:49
Unidentified speaker (speaker_3): very nearly every national. name that sells into the AI build. Oracle's contracted backlog is more than half OpenAI. Coreweaves Book, once its Microsoft routed capacity is traced through to the underlying tenant, runs to roughly two-thirds OpenAI. Softbank's commitments, through Stargate, are almost entirely
1:17:08
Unidentified speaker (speaker_3): OpenAI. This is precisely the structure that made 2008's senior tranches lethal.
1:17:15
Unidentified speaker (speaker_3): Thousands of individual mortgages, geographically dispersed, statistically independent, until the one macro variable they all depended on, national home prices, turned, and the correlation the models had assumed away revealed itself to be one. Here, the single variable is not home prices. It is
1:17:35
Unidentified speaker (speaker_3): whether OpenAI can clear its next mark.
1:17:38
Unidentified speaker (UNKNOWN): That
1:17:38
Unidentified speaker (speaker_3): is why. chip stocks fell when OpenAI signaled it may delay its IPO from 2026 to 2027.
1:17:45
Unidentified speaker (UNKNOWN): A
1:17:45
Unidentified speaker (speaker_3): correlation of one is invisible until it is tested. Then it is a transmission line. When the borrower at the center cannot clear its next mark, the loss does not stay put. It runs the length of the chain into every counterparty that booked its commitment as demand. The naked borrower is not merely the weakest credit in the
1:18:05
Unidentified speaker (speaker_3): complex, it is the credit the complex is wired to. That correlated exposure is now being securitized. In May 2026, CoreWeave closed its DDTL 5.0 facility, $3.1 billion, issued through a bankruptcy remote financing subsidiary. CoreWeave CoreWave disclosed that
1:18:25
Unidentified speaker (speaker_3): the underlying capacity serves two large non-investment grade customers, OpenAI and Cohere. But the distinction that matters is structural. DDTL 5.0 was the first publicly syndicated GPU-backed facility built to trade in the secondary market. The paper has left the originator's balance sheet and
1:18:45
Unidentified speaker (speaker_3): entered the broad credit complex, the distribution step, the moment originate to distribute. stops being a metaphor. The D and DTL isn't serviced by OpenAI's earnings. It's serviced by OpenAI's ability to keep raising, which is underwritten by the AI CapEx narrative continuing to compound.
1:19:05
Unidentified speaker (UNKNOWN): X.
1:19:06
Unidentified speaker (speaker_3): The refinance of last resort.
1:19:10
Unidentified speaker (speaker_3): Trace the refinancing chain to its end, and you arrive at the public market.
1:19:15
Unidentified speaker (speaker_3): Private capital is deep but finite. Softbank, the sovereign funds, the hyperscalers, the mega funds each can absorb a round or two. But the lab's burn is measured in tens of billions a year and compounding. And at some point, the only pool of capital large enough to keep refinancing it is the one the index
1:19:35
Unidentified speaker (speaker_3): funds and the retail bids sit in.
1:19:37
Unidentified speaker (UNKNOWN): The
1:19:37
Unidentified speaker (speaker_3): IPO is not an exit in this structure. It is the refinancing of last resort, the final deepest teaser into which the whole edifice expects to roll once the private rounds can no longer carry the burn. Which is why news of OpenAI's delayed IPO matters far more than the market initially understood.
1:19:57
Unidentified speaker (speaker_3): The terminal refinance carries a trap the private rounds did not.
1:20:01
Unidentified speaker (UNKNOWN): To
1:20:01
Unidentified speaker (speaker_3): reach the public pool, the borrower must file an S1.
1:20:05
Unidentified speaker (UNKNOWN): And
1:20:06
Unidentified speaker (speaker_3): the S1 discloses exactly the fragility that made the refinance necessary: audited losses, customer concentration, the full $600 billion plus of take or pay obligations laid out for any reader.
1:20:20
Unidentified speaker (UNKNOWN): The
1:20:20
Unidentified speaker (speaker_3): document that unlocks the capital is the same document that prices the risk. OpenAI needs the market's money. and cannot fully afford the market's scrutiny, the bind of a company whose story is better than its statements.
1:20:34
Unidentified speaker (UNKNOWN): Now
1:20:34
Unidentified speaker (speaker_3): do the arithmetic the delay is hiding.
1:20:37
Unidentified speaker (UNKNOWN): The
1:20:37
Unidentified speaker (speaker_3): step up from roughly $852 billion to the reported greater than $1 trillion target is the next hurdle. Barely 1.23 times, the lowest step step-up up in the entire sequence. and far below the 1.7x to 1.9x multiples that funded the prior burns. It
1:20:57
Unidentified speaker (speaker_3): must do two incompatible things at once. Clear at a level the public market will actually pay, and raise enough to retire a cumulative burn approaching $115 billion. The implied step-up cannot do both. The price that clears the market does not retire the burn. and the price that retires the burn
1:21:17
Unidentified speaker (speaker_3): does not clear the market. The refinance of last resort is failing quietly, pricing below the mark the structure requires and waiting.
1:21:27
Unidentified speaker (speaker_3): 11. How it breaks.
1:21:31
Unidentified speaker (UNKNOWN): The
1:21:31
Unidentified speaker (speaker_3): trigger is narrow and specific. The next mark fails to clear at the required step step-up. up.
1:21:37
Unidentified speaker (UNKNOWN): Not
1:21:38
Unidentified speaker (speaker_3): a collapse. merely a deceleration below the threshold the structure needs. This is the 2006 dynamic replayed. The velocity rolled over while the level was still climbing. From there, the sequence runs in order. 1. The terminal refinance prices below the required mark. The
1:21:58
Unidentified speaker (speaker_3): step from about $852 billion to more than $1 trillion depends. does not clear, or clears at a level that cannot retire the burn. The delay is the signal. 2. The borrower pulls back on compute commitments to conserve cash. And a pullback on a take or pay obligation is a covenant breach against
1:22:17
Unidentified speaker (speaker_3): the provider whose debt is collateralized by that commitment. 3. The breach lands first and hardest on the neo NeoClouds clouds. Coreweave, Corweave, Lambda, Crusoe, whose entire business is the spread between borrowed money and resold compute.
1:22:32
Unidentified speaker (UNKNOWN): A
1:22:32
Unidentified speaker (speaker_3): NeoCloud is not a business so much as a spread trade with no balance sheet to warehouse the risk. When the spread inverts, it is insolvent by definition, not by choice. Oracle, corporate funded but dangerously concentrated, takes the next blow. It's impairment deeper than the hyperscalers. but it does not seize. It
1:22:52
Unidentified speaker (speaker_3): bleeds.
1:22:53
Unidentified speaker (UNKNOWN): A
1:22:53
Unidentified speaker (speaker_3): hyperscaler can fund a missed payment out of search or Windows or retail. The NeoCloud has no second cash flow. 4. Credit freezes across the complex. RPO reprices from forward demand to counterparty risk. GPU-backed CPU-backed notes cannot roll. The originate-to-distribute originate to distribute machine
1:23:13
Unidentified speaker (speaker_3): seizes. 5. Equity decimates. Negative convexity in reverse. CapEx repriced from option to cost. Multiples compressing across every name in the chain. 6. The strong survive. The best capitalized actors with the least exposure buy stranded data centers for pennies and
1:23:32
Unidentified speaker (speaker_3): backstops the leases that must endure. A necessary concession. I do not know when. The trigger could be quarters away or further. The borrowed time window between the second derivative rolling over and the first derivative crossing zero can stretch further than any short seller's patience. There are three stretches that can extend
1:23:52
Unidentified speaker (speaker_3): it: a larger than expected private round, a sovereign or strategic backstop that postpones the terminal refinance, like an Intel-style federal equity stake, and the
1:24:02
Unidentified speaker (UNKNOWN): hyper
1:24:02
Unidentified speaker (speaker_3): hyperscalers' continued ability to lever up, borrowing against the very backlog this article has described. The last of these is the most powerful near-term stabilizer, because the hyperscalers have real balance sheets, real cash flows, and real access to debt markets. But it is not infinite. Investment
1:24:22
Unidentified speaker (speaker_3): grade leverage across the group has already doubled in a year, and the rating agents. agencies have only so many notches left. The sequence above is not a calendar, it is a mechanism conditioned on a single variable, where the growth decelerates below the rate the refinance requires.
1:24:39
Unidentified speaker (UNKNOWN): But
1:24:40
Unidentified speaker (speaker_3): with OpenAI's IPO already delayed, the clock is ticking.
1:24:45
Unidentified speaker (speaker_3): 12. The strongest case against this.
1:24:50
Unidentified speaker (speaker_3): Grant the bulls their strongest case. Demand is real. Backlogs are exploding. Inference is in its infancy, and the risk of underbuilding a generational platform is acute. Supply is locked years out, and even skeptics see paths to $1.4 trillion in annual CapEx. I take this case
1:25:10
Unidentified speaker (speaker_3): seriously, but it does not save the structure. Every bull claim is about the level or the first derivative. Backlogs, inference ramping, supply growth. Not one speaks to the second derivative. I do not need demand to fail. I need the rate of capex growth to flatten. And a structure this levered and
1:25:30
Unidentified speaker (speaker_3): dependent on perpetual acceleration breaks on the flattening alone. Grant every level argument. Housing demand was real in 2006. and the financing detonated on deceleration, not the level. There are three bull cases to address. First, the fortress balance sheet.
1:25:49
Unidentified speaker (speaker_3): Hyperscalers generate enormous cash flow. A tenant impairment is absorbable. This misses the wound. The impairment is accounting. Margin collapse is structural. AI capacity carries a massive fixed. cost base, depreciation, power, interest, that does not flex when a tenant defaults.
1:26:09
Unidentified speaker (speaker_3): Utilization drops, but OPEX does not. Revenue falls, yet costs remain anchored to the peak build. The same operating leverage that supercharged profits now destroys margins on the way down. Second, the cross-subsidization defense. If AI margins crater, search and Windows cash flows
1:26:29
Unidentified speaker (speaker_3): carry the division. The rebuttal is the conglomerate discount. Investors buy hyperscalers for growth, not to subsidize perpetual losses. If AI consumes tens of billions without profitability, consolidated ROIC declines. A high ROIC growth compounder that becomes a low ROIC
1:26:48
Unidentified speaker (speaker_3): capital-intensive operator loses its growth. premium and trades down to a utility multiple. Worse, legacy cash cows are not infinite engines. Search faces structural erosion. Retail operates on thin margins. Windows is mature. Using shrinking profits from declining units to fill vacancies
1:27:08
Unidentified speaker (speaker_3): is not patient capital. It is value destruction. The conglomerate trades as a utility with a venture capital problem. commanding a lower multiple. Third, the physical rebuttal. If OpenAI defaults, the provider releases the capacity. This is the housing never loses value argument of
1:27:27
Unidentified speaker (speaker_3): 2006.
1:27:28
Unidentified speaker (UNKNOWN): An
1:27:29
Unidentified speaker (speaker_3): OpenAI default will not occur in isolation. It will coincide with a broader deceleration, meaning hyperscalers bring gigawatts online into a softening environment. You do not re-lease into a GLUT. You compete on price, and the clearing price falls below the debt service coverage ratio. The replacement tenant, facing the
1:27:49
Unidentified speaker (speaker_3): same decelerating demand, will demand a 30 to 50% discount and a shorter commitment, turning a long-duration, high-yield asset into a distressed instrument. Releasing merely transforms a clean default into a prolonged. vacancy crisis, the same mechanism that turned 2007's subprime reperformance
1:28:08
Unidentified speaker (speaker_3): hopes into a five-year grind.
1:28:11
Unidentified speaker (UNKNOWN): The
1:28:11
Unidentified speaker (speaker_3): bulls and I do not disagree about AI. We disagree about which derivative the structure is written on. They are watching the level. I am watching its acceleration. That is not a difference about technology, it is a difference about arithmetic. And arithmetic, eventually, does not take opinions.
1:28:32
Unidentified speaker (speaker_3): 13. The number nobody watches.
1:28:37
Unidentified speaker (speaker_3): Three errors. Stacked, recreate 2008. The market is pricing AI as a technology cycle when its financing is the machinery of a credit and real estate cycle.
1:28:49
Unidentified speaker (UNKNOWN): It
1:28:49
Unidentified speaker (speaker_3): is watching the level and the velocity. while the structure breaks on the acceleration.
1:28:54
Unidentified speaker (UNKNOWN): And
1:28:54
Unidentified speaker (speaker_3): it is treating a concentrated single-borrower loan book as though it were diversified forward demand. Each error alone might be survivable. Together, they reconstruct, feature for feature, the conditions of the last great credit event. The same negatively convex structures, the same originate to distribute plumbing, the
1:29:14
Unidentified speaker (UNKNOWN): same
1:29:14
Unidentified speaker (speaker_3): correlation. of one hiding inside the appearance of diversification, the same blindness to the one derivative that matters. The law from the opening sections holds, unchanged. Any structure whose serviceability depends on refinancing into growth does not need a decline. It needs only a deceleration.
1:29:34
Unidentified speaker (speaker_3): That deceleration is already happening.
1:29:37
Unidentified speaker (UNKNOWN): The
1:29:37
Unidentified speaker (speaker_3): market remembers two. 2008 backwards. The defaults didn't come when prices fell. They came when prices stopped rising faster. And this build out is engineered with exquisite precision to break on the one number nobody watches the second derivative.
1:29:58
Unidentified speaker (speaker_2): Thanks for joining us for today's episode. We'd really appreciate it if you could like us. and subscribe the podcast if you got value from it, as this helps us grow our listenership. If you're a sophisticated investor under Australian law, we'd welcome you to sign up to our investor syndicate, where we provide curated investment opportunities in great B2B focused companies that we help scale into the
1:30:18
Unidentified speaker (speaker_2): US and the UK. If you're a founder of a B2B To be company that has US or the UK U County on your roadmap, we'd also love to have a chat to see how we can help. Go to tribe global.vc for more links and information and to get in contact with us. See you at the next episode.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from Transparent Venture Capital by Tribe Global Ventures

All insights