Market Maker

Speakers debate whether AI is a productive inflection or a credit-fueled bubble amid supply constraints and mixed productivity data.

40 min Listen to the original Episode page
AI InfrastructureMarket BubblesProductivityCloud ComputingSemiconductorsLegal TechSpaceXMicrosoft

Summary

Speaker 2 argues the AI sector is driven by accelerating demand, sold-out supply from manufacturers like SK Hynix, and strong profitability across the stack, including cloud providers and Anthropic. Speaker 3 counters with a bear case citing a lack of broad productivity gains, circular financing risks, and a potential 'Minsky moment' as credit growth shifts to bad projects. The discussion highlights conflicting evidence on productivity, with Speaker 2 pointing to Microsoft's revenue growth and agentic AI adoption, while Speaker 3 notes stagnant app downloads and rising legal billable hours. Both speakers acknowledge government-controlled infrastructure constraints may prevent the overbuild typical of historical bubbles.

Key takeaways

  1. 1
    is planning to by the end of next year, is planning to have hit a run rate of building um one million Optimus road. robots per year, right? He'll be at that run rate by the end of next year. So obviously all of these robotic we're we're right at the start, I would say, of the robotic boom. And each of those robots are going to need AI licenses, let's say, to operate. So I think we're we're on the cusp of that demand. I In my view, the demand's only going gonna to go up, and therefore. the whole stack is gonna become even more profitable, so people are gonna want to carry on funding it. And so I think this thing can still go.

    Unidentified speaker (speaker_2) · 27:00

  2. 2
    Speaker 2 claims Microsoft's revenue increased by 50% over the last 24 months while their employee headcount remained flat.

    Unidentified speaker (speaker_2) · 32:58

Episode sections

  1. 0:00 – 9:24

    Introduction and Market Context

    Speakers discuss the absence of co-host Anthony Chung and engage in banter regarding listener comments on their ages. Speaker 2 notes a recent market sell-off following midterm elections, with the Nasdaq dropping over 5% on Thursday, while acknowledging the prior positive reaction to the election results.

  2. 9:24 – 14:45

    The Bull Case for AI

    Speaker 2 argues against the bubble narrative by citing accelerating demand, contracted backlogs for cloud providers, and sold-out supply from manufacturers like SK Hynix. He highlights that profitability exists across all layers of the AI stack, from chipmakers to model providers like Anthropic.

  3. 14:45 – 18:42

    The Bear Case for AI

    Speaker 3 presents a bear case centered on the lack of broad productivity gains and the risk of circular financing where money flows between AI companies rather than generating real economic value. He warns of a 'Minsky moment' as credit growth runs out of good projects, exacerbated by limited government fiscal capacity to stimulate the economy.

  4. 18:42 – 24:24

    Economic Risks and Fiscal Constraints

    Speaker 3 argues that governments in the US, UK, and France lack the fiscal space to intervene in a crisis due to high deficits and inflation. He contrasts this with Speaker 2's point that the AI boom is privately funded by hyperscalers, not government spending.

  5. 24:24 – 28:12

    Adoption and Future Demand Drivers

    Speaker 2 provides an anecdote about his brother's manufacturing business adopting AI tools, predicting widespread efficiency gains. He cites Elon Musk's plan to produce one million Optimus robots annually by the end of next year as a future driver of AI demand.

  6. 28:12 – 28:39

    Productivity Metrics and Digital Cocaine

    Speaker 3 counters with a 2025 study suggesting that while AI made coders 20% faster, the total number of code commits dropped by 20%, labeling the phenomenon 'digital cocaine' and questioning whether volume equates to value.

  7. 28:39 – 31:51

    AI Productivity and Legal Sector Trends

    Speaker 3 argues that increased volume in AI apps and tools does not equate to increased value, citing doubled app releases with only 2% download growth. They further claim that billable hours per legal case rose in the first half of 2026 despite AI adoption, suggesting a lack of productivity gains in deterministic industries.

  8. 31:51 – 32:26

    AI as a Driver of Legal Demand

    Speaker 3 suggests AI may be increasing legal billable hours by causing harm, such as cyberattacks and deepfakes, rather than providing productivity. Speaker 2 counters that AI might empower individuals to pursue legal claims they previously couldn't, potentially increasing litigation volume.

  9. 32:26 – 34:38

    Microsoft Revenue vs Headcount and Agentic AI

    Speaker 2 claims Microsoft increased revenue by 50% over the last 24 months while keeping headcount flat, attributing this to AI productivity. Speaker 3 notes this is specific to large tech companies. Speaker 2 then discusses the 'agentic phase' of AI, citing Meta's Muse app, and argues agents will consume exponentially more compute than individual users.

  10. 34:38 – 36:56

    Workforce Replacement and Bubble Theory

    Speaker 3 contrasts AI as a productivity booster versus a workforce replacer, referencing SpaceX's prospectus which they claim values AI products at $20 trillion. Speaker 3 introduces a theory by Byers and Huber distinguishing between 'mean reversion' (destructive) and 'inflection' (productive) bubbles, categorizing AI as the latter.

  11. 36:56 – 39:48

    Infrastructure Supply Constraints and Government Throttling

    Speaker 2 argues that unlike the dot-com bubble, the current AI infrastructure build is constrained by government-controlled supply of data centers and power. They claim these resources are in short supply due to regulatory and political factors, preventing the overbuild that typically causes bubble bursts.

Assets discussed

AssetKindStanceContext
NVIDIANVDAequitybullishAnd we can see this, like NVIDIA's revenue was more than doubled in the last year, right? 96 billion in one quarter.
Coppercommoditybullishyou look at other raw materials, copper, and actually in my research here, tin, I hadn't quite appreciated.
AlphabetGOOGLequitybullishCloud providers, you know, the big, you know, the big guns, AWS, Google, Microsoft, they're basically running operating operating

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

Claims worth verifying

NVIDIA's revenue was more than doubled in the last year, reaching $96 billion in one quarter.

Unidentified speaker (speaker_2) ·10:47

How to verify: Check NVIDIA's Q2 FY2025 earnings report on SEC EDGAR or NVIDIA Investor Relations for revenue figures. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Primary source

AWS and Google have over a trillion dollars of signed, contracted backlog.

Unidentified speaker (speaker_2) ·11:07

How to verify: Review AWS and Google Cloud financial disclosures in their respective 10-K filings on SEC EDGAR for backlog or deferred revenue metrics. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Primary source

SK Hynix is sold out through 2029.

Unidentified speaker (speaker_2) ·11:46

How to verify: Check SK Hynix press releases or investor presentations for capacity utilization and order book visibility. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

NVIDIA has 75% gross margins.

Unidentified speaker (UNKNOWN) ·12:46

How to verify: Verify in NVIDIA's most recent quarterly earnings report on SEC EDGAR. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Primary source

AWS, Google, and Microsoft run operating margins of 36 to 39%.

Unidentified speaker (speaker_2) ·13:06

How to verify: Check the respective 10-K filings for AWS (Amazon), Google (Alphabet), and Microsoft on SEC EDGAR for segment operating income margins. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Primary source

Anthropic made an operating profit in its S1 filing.

Unidentified speaker (UNKNOWN) ·13:16

How to verify: Review Anthropic's S-1 registration statement filed on SEC EDGAR for net income/loss and operating profit figures. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Primary source

Morgan Stanley estimates $570 billion of AI-related borrowing.

Unidentified speaker (UNKNOWN) ·17:16

How to verify: Search for Morgan Stanley research reports or articles citing this figure, likely from late 2026 or early 2026. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Elon Musk plans to have a run rate of building one million Optimus robots per year by the end of next year.

Unidentified speaker (speaker_2) ·26:40

How to verify: Check Tesla's earnings calls or official announcements regarding Optimus production targets for 2026. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

A 2025 study by METR found coders felt 20% faster with AI but code commits dropped 20%.

Unidentified speaker (speaker_2) ·28:12

How to verify: Locate the specific METR (Machine Intelligence Research) study from 2026 discussing AI impact on coding productivity and commit volumes. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

The Nasdaq was down over five percent at one point on Thursday.

Unidentified speaker (speaker_2) ·6:23

How to verify: Check historical Nasdaq Composite index data for the Thursday prior to the recording date (Friday, November 6, 2026), which is November 5, 2026. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

The number of apps released on the App Store doubled in the first half of 2026, while downloads increased by 2%.

Unidentified speaker (speaker_3) ·28:39

How to verify: Check Apple's App Store statistics or third-party app analytics firms (e.g., Sensor Tower, App Annie) for H1 2026 release and download volume data. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Billable hours per legal case kept rising in the first half of 2026, despite AI adoption in law firms.

Unidentified speaker (speaker_3) ·29:19

How to verify: Review legal industry reports from sources like the American Bar Association, Thomson Reuters Legal Trends Report, or BigLaw metrics for H1 2026 billable hour data. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Microsoft's revenue increased by 50% over the last 24 months while headcount remained flat.

Unidentified speaker (speaker_2) ·32:58

How to verify: Check Microsoft's SEC 10-K filings for revenue figures for the trailing 24 months and employee count data from their investor relations page or proxy statements. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Primary source

Meta's Muse was the most downloaded app on the App Store.

Unidentified speaker (UNKNOWN) ·33:55

How to verify: Check App Store ranking archives or third-party analytics (e.g., Sensor Tower) for the most downloaded apps in the relevant period. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

SpaceX's prospectus states the total addressable market for their AI products is $20 trillion.

Unidentified speaker (UNKNOWN) ·34:56

How to verify: Locate the SpaceX S-1 or relevant prospectus filing on SEC EDGAR and search for 'total addressable market' or 'AI' TAM figures. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Primary source

AI is causing more harm than good, leading to increased legal cases due to cyberattacks, deepfakes, and hacking.

Unidentified speaker (speaker_3) ·31:51

How to verify: Review cybersecurity incident reports from firms like CrowdStrike, Mandiant, or government reports (e.g., CISA) for trends in AI-related crimes and resulting litigation in 2026-2026. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Data center build-outs and power supply are constrained by government forces, preventing infrastructure overbuild.

Unidentified speaker (speaker_2) ·38:03

How to verify: Review regulatory filings, zoning permits, and power grid capacity reports from energy regulators and local governments for data center projects in 2026. For recent events, verify 2026-08-08 through 2026-11-08. Use another period only when the speaker explicitly names it.

Primary source

Do your own research

  • Verify SK Hynix's supply backlog and capacity utilization reports for 2024-2029.
  • Check AWS and Google Cloud's latest earnings calls for contracted backlog figures and guidance.
  • Review Anthropic's S1 filing for detailed operating profit and revenue breakdowns.
  • Analyze Microsoft's latest quarterly earnings reports for revenue growth rates and employee headcount changes.
  • Examine App Store developer reports or third-party analytics firms for app release volumes versus download growth trends.
  • Read the SpaceX S-1 filing to confirm the stated Total Addressable Market (TAM) for AI products.
  • Investigate Morgan Stanley's research notes on AI-related borrowing and equity capital flows.
  • Assess government regulatory filings regarding data center power and zoning constraints in key markets.
Full transcriptexpand
0:00
Unidentified speaker (speaker_0): Okay. When I sell my business, I want the best tax and investment advice. I want to help my kids, and I want to give back to the community. Ooh, then it's the vacation of a lifetime.
0:12
Unidentified speaker (UNKNOWN): I
0:12
Unidentified speaker (speaker_0): wonder if my out of office has a forever setter.
0:15
Unidentified speaker (UNKNOWN): An
0:15
Unidentified speaker (speaker_1): IG private wealth advisor creates the clarity you need with plans that harmonize your business, your family, and your dreams. Get financial advice that puts you at the center. Find your advice. advisor at IgPrivatewealth.com.
0:31
Unidentified speaker (speaker_2): Hello, hello. Welcome back. Um, we've got a a little bit of a surprise uh episode here. Surprise in that your main man, uh Anthony Chung. Well, I he's actually decided to um sod off to Paris, can you believe?
0:49
Unidentified speaker (UNKNOWN): Um
0:50
Unidentified speaker (speaker_2): ditching. his responsibilities and his his duties to his audience. Um and so it just leaves me and you, Stephen, to come together.
0:59
Unidentified speaker (speaker_3): Yeah, this is this is amazing. It it's very, very rare that I speak to you and you know, we're
1:06
Unidentified speaker (UNKNOWN): both
1:07
Unidentified speaker (speaker_3): on the same podcast and uh we're just at different sides and it's it's like a it's like a solar eclipse, right?
1:15
Unidentified speaker (speaker_3): Very, very special. But if you don't wear those funny glasses, your eyes are gonna get burnt.
1:20
Unidentified speaker (speaker_2): This is it. I think we did do one episode. It actually might have been without though. So I think this is this may be the f I I'm not sure how many episodes we've done. It's probably like not sure what episode this is. It's gotta be over two hundred and fifty. Anyway, I think this might be the very first time we've uh we've we've gone solo without without without the boss man. So
1:37
Unidentified speaker (UNKNOWN): it's gonna be a car
1:38
Unidentified speaker (speaker_2): crash.
1:38
Unidentified speaker (speaker_3): It's gonna be an absolute car crash. I do what I do think that Ant holds this whole thing together, and he's very good at bringing me and you back to topic and back to a sense of time and structure. So
1:53
Unidentified speaker (UNKNOWN): whether
1:53
Unidentified speaker (speaker_3): this becomes a total garbled uh rant, we'll find out. But I tell you what,
1:59
Unidentified speaker (UNKNOWN): please.
2:00
Unidentified speaker (speaker_3): I thought I thought I'd do a little bit of prep. before this episode. So I'm quite frankly,
2:07
Unidentified speaker (speaker_3): I don't always listen to your episodes. Um I I tend to get my financial news elsewhere.
2:13
Unidentified speaker (UNKNOWN): Oh,
2:14
Unidentified speaker (speaker_2): thanks. Thanks.
2:15
Unidentified speaker (UNKNOWN): Sorry about that.
2:15
Unidentified speaker (speaker_3): But no, I I heard it was your birthday week. And uh I also heard that there was a bit of Of a sort bit of chat on the comments uh about what age, what birthday it was.
2:26
Unidentified speaker (UNKNOWN): Did you see?
2:27
Unidentified speaker (speaker_3): these comments?
2:28
Unidentified speaker (speaker_2): I did see the comments. I was yeah, there was a
2:30
Unidentified speaker (UNKNOWN): Did you do the
2:30
Unidentified speaker (speaker_3): analysis?
2:31
Unidentified speaker (UNKNOWN): Oh
2:31
Unidentified speaker (speaker_2): well, I didn't actually crunch the numbers to see
2:33
Unidentified speaker (UNKNOWN): what the
2:34
Unidentified speaker (speaker_2): exact average age was. I do feel it was lower than my actual age though.
2:39
Unidentified speaker (speaker_3): Uh well
2:39
Unidentified speaker (UNKNOWN): I mean I
2:40
Unidentified speaker (speaker_3): tell you what
2:40
Unidentified speaker (UNKNOWN): there
2:40
Unidentified speaker (speaker_2): were some pretty wide outliers.
2:42
Unidentified speaker (UNKNOWN): There were some outliers,
2:43
Unidentified speaker (speaker_3): yeah, exactly. So I did the I I ran the numbers. Oh,
2:46
Unidentified speaker (speaker_2): you did? And
2:47
Unidentified speaker (speaker_3): across all the comments so far, this is we're recording Friday the sixth. Across all the comments, your age,
2:54
Unidentified speaker (speaker_3): according to the listeners, forty four point eight years old. You take
3:00
Unidentified speaker (UNKNOWN): that.
3:00
Unidentified speaker (speaker_2): That is a win. That's a big
3:02
Unidentified speaker (UNKNOWN): now there was a really
3:03
Unidentified speaker (speaker_3): good clustering around that forty three, forty four mark. Now there was definitely one person that wanted to troll you and said you were sixty 65, five, but you look good for a sixty 65 five year old.
3:15
Unidentified speaker (speaker_3): And then and then I think Singh Mehu, one of the one of the uh speculators, I think is looking for a job. Uh
3:23
Unidentified speaker (UNKNOWN): she
3:23
Unidentified speaker (speaker_3): said you only look 25. So enough with the flattery, Singh.
3:27
Unidentified speaker (UNKNOWN): Yeah,
3:28
Unidentified speaker (speaker_2): absolutely. I mean 44, I'll take that. I'll take that because you know, obviously that's that's exactly right. Uh
3:33
Unidentified speaker (speaker_3): that's your age. Yeah.
3:34
Unidentified speaker (speaker_2): I I think actually my favorite one, and I'm I'm gonna misquote it, but it was something like uh um
3:40
Unidentified speaker (speaker_2): long uh long forty seven, take profit of forty-nine with a stop loss at forty six.
3:47
Unidentified speaker (speaker_3): I
3:47
Unidentified speaker (UNKNOWN): think was the
3:48
Unidentified speaker (speaker_2): comment which I I quite I quite enjoyed.
3:50
Unidentified speaker (UNKNOWN): Oh
3:51
Unidentified speaker (speaker_3): well
3:51
Unidentified speaker (UNKNOWN): I'm
3:51
Unidentified speaker (speaker_3): I'm not far away. I um I had my fortieth birthday last week no two weeks ago. Um and j a few weeks
3:58
Unidentified speaker (UNKNOWN): before my 40th fortieth
3:59
Unidentified speaker (speaker_3): birthday, I got ID'd buying glue in B and Q. So that's there you go.
4:03
Unidentified speaker (UNKNOWN): Well
4:03
Unidentified speaker (speaker_2): that's that's because your new your new haircut. It's kind of taken
4:06
Unidentified speaker (UNKNOWN): a
4:07
Unidentified speaker (speaker_2): decade off.
4:07
Unidentified speaker (speaker_3): Yeah, but I was with
4:08
Unidentified speaker (UNKNOWN): my
4:09
Unidentified speaker (speaker_3): my daughter at the time. So I uh I was a very precocious 17-year-old buying glue with my daughter on a Saturday morning. Um
4:17
Unidentified speaker (speaker_3): anyway.
4:17
Unidentified speaker (UNKNOWN): Um what
4:18
Unidentified speaker (speaker_3): are we gonna do?
4:19
Unidentified speaker (speaker_2): Let's well, this is it. We were we were on the on our WhatsApp group um over the weekend. It's like, what are we gonna do? Ant's not here. Um and and so I mean, I think it was your idea, and it's like, well, let's just have the conversation. Let's just have the conversation. I think there's pretty much only one in town. Um, and that is
4:39
Unidentified speaker (speaker_2): obviously AI. We've had the midterms, of course, um, which you know we'll talk about in a second. But we were on the WhatsApp group going, all right, let's let's do an AI, and your your suggestion was let's do bull. versus bear.
4:53
Unidentified speaker (UNKNOWN): So
4:53
Unidentified speaker (speaker_2): here we are. And I and then Stephen was like, Look, which side do you want to take? I don't mind. And I immediately and and quite confidently typed in the words I will definitely take the bull case.
5:05
Unidentified speaker (speaker_3): Yeah, it was almost that like that was a kind of pre written thing that you just you send whenever anyone asks you that question.
5:12
Unidentified speaker (speaker_2): This is it. Um, unfortunately for me, I sent that message. on Monday
5:16
Unidentified speaker (UNKNOWN): in
5:16
Unidentified speaker (speaker_2): the WhatsApp group. Uh then we had Tuesday um midterm elections, um, the blue wave, or some some are calling it um the blue the blue NAMI, I've heard.
5:30
Unidentified speaker (UNKNOWN): Blue
5:30
Unidentified speaker (speaker_3): Nami.
5:31
Unidentified speaker (speaker_3): I
5:31
Unidentified speaker (UNKNOWN): understand what
5:32
Unidentified speaker (speaker_3): they what's what they're doing there.
5:34
Unidentified speaker (speaker_2): Um but and then we dropped an episode, me and Anne. We did it on Wednesday and and
5:40
Unidentified speaker (speaker_2): we were, I mean, we I was getting quite excited. I mean, the market reaction was very positive initially. In certain in most corners of the market, there were some definitely some exceptions to that. Some corners of the market got heavily pummeled, like the semiconductors specifically. Um, some of the oil majors, for
6:00
Unidentified speaker (speaker_2): example, right? But broadly it was very risk-on, and we were
6:03
Unidentified speaker (speaker_2): spinning the argument as to why. Unfortunately, then the following day, Thursday, so that's yesterday, as we're recording this, we had actually one of the biggest market blowouts for many, many, I don't know how long it is, to be honest, but you know, the Nasdaq was down four, well, over five percent at one point. You know, you had a proper
6:23
Unidentified speaker (speaker_2): old old-fashioned fashioned um sell everything. that's not nailed down uh kind of session, and it felt pretty ugly.
6:32
Unidentified speaker (speaker_3): Is this, you know, I'm not I'm not a markets person, and you definitely are. Is this is this the equival was yesterday, and again we're speaking on Friday the sixth, was yesterday the equivalent of, you know, when you really, really want to break up with someone and you've been waiting for the right moment and maybe the night before, you know,
6:52
Unidentified speaker (speaker_3): they dance with your best friend or something like that, and you use that as the trigger to break up. Was that was that what happened with the midterms or not not quite that severe?
7:02
Unidentified speaker (speaker_2): Uh
7:03
Unidentified speaker (UNKNOWN): I
7:03
Unidentified speaker (speaker_2): d I d yeah, possibly. I mean, I I think that there's two parts to this, I believe. Um, and we're gonna get into it, right? Because obviously one of your arguments, you taking the bear case. I'm assuming, is gonna be peddling the whole bubble um
7:19
Unidentified speaker (speaker_3): argument.
7:19
Unidentified speaker (speaker_2): Um
7:21
Unidentified speaker (UNKNOWN): and
7:21
Unidentified speaker (speaker_2): maybe I'm getting ahead of the conversation a bit here, but
7:24
Unidentified speaker (UNKNOWN): you
7:25
Unidentified speaker (speaker_2): know, I think what happened was a direct reaction to uh obviously the the scale of the win from the Democrat side, meaning that,
7:34
Unidentified speaker (UNKNOWN): and
7:35
Unidentified speaker (speaker_2): particularly that data center
7:37
Unidentified speaker (UNKNOWN): and
7:38
Unidentified speaker (speaker_2): I'm feeding feeding your argument here, but that that kind of that that that the voters have said their word and data centers are not for us, kind of um interpretation of the result, right? And so then you've had that panic, I would say, on Thursday, anybody who's who's in the business of building data
7:58
Unidentified speaker (speaker_2): centers and and and that's the service and the or the or products that they're supplying for that process, you know, I think they're the ones that got immediately tanked. Yeah. But is it a bubble bursting? I mean, you gotta I would say the semiconductor sector very specifically, that narrow sector,
8:18
Unidentified speaker (UNKNOWN): you know,
8:18
Unidentified speaker (speaker_2): that's been selling off for for a few months.
8:21
Unidentified speaker (UNKNOWN): June, right?
8:21
Unidentified speaker (speaker_3): Yeah.
8:22
Unidentified speaker (speaker_2): I'd say it's been a slow deflation. Um and you know, we're fifty percent down on some of these stocks, right? But um anyway, I look I'm getting ahead of myself
8:31
Unidentified speaker (UNKNOWN): here.
8:31
Unidentified speaker (speaker_3): You are. And look, and I think it's really important to know we're gonna do the bulls versus the bears, and we're gonna do a bit of a debate style where you're gonna introduce your kind of bull case and I'll and I'll
8:39
Unidentified speaker (UNKNOWN): do the
8:39
Unidentified speaker (speaker_3): same with the bear case, and then we'll have a chat. But I think it's I think it's also worth
8:45
Unidentified speaker (speaker_3): using uh the volatility in the market at the moment as a learning point to say, you know. there's a lot of noise out there, right? I mean, we're looking at markets today, and, you know, there's a lot of there's a lot of recovery. There's a decent amount
8:58
Unidentified speaker (UNKNOWN): of
8:59
Unidentified speaker (speaker_3): bounce back. And by the time this episode goes out on uh Sunday through Monday, something else might have happened, right? So we're gonna
9:07
Unidentified speaker (speaker_0): try and
9:07
Unidentified speaker (speaker_3): present arguments maybe from a not a 30,000 thirty thousand foot view, but at least a 10,000 ten thousand foot view and not not get overly distracted by the data deck. So Peter,
9:17
Unidentified speaker (UNKNOWN): exactly
9:17
Unidentified speaker (speaker_3): I want you to give it a go. Give me your bullcase because I need uh I need some reassurance.
9:24
Unidentified speaker (speaker_2): Well, um this is not investment advice.
9:27
Unidentified speaker (UNKNOWN): Um and
9:28
Unidentified speaker (speaker_2): uh spoiler, I probably do own some of these stocks that I might be discussing uh in the next ten, fifteen minutes. That said, this is my bullcase. bull case. Are you ready? Basically, look, the bear case, as I've said, partially, and we're going to hear from you, is this bubble argument. So firstly, I want to address that.
9:48
Unidentified speaker (speaker_2): Bubbles historically have a signature, okay? And you can go back the railway mania of the eighteen forties, housing bubble bursting in two 2008, thousand and eight, shale oil. Look, there've there have been many, many bubbles that have then crashed. but there's been one commonality that preceded the crashes. And you get a combination
10:08
Unidentified speaker (speaker_2): of three factors at the same time: falling prices, rising inventory or supply, and then idle or slowing capacity. Okay. So it tends to be supply A overtaking demand is the typical function that then leads to a collapse. Now, I'm going
10:27
Unidentified speaker (speaker_2): to argue very strongly that in this AI specific case, we've actually got the exact opposite. So that's my so, like on the demand side, I'm going to put forward look, demand is accelerating. Also, a lot of the demand side from some of the big players in this data center build-out race and
10:47
Unidentified speaker (speaker_2): so on have legally binding controls. contracted, you know, orders going way out into the future. And we can see this, like NVIDIA's revenue was more than doubled in the last year, right? 96 billion in one quarter. AWS and Google, you know, some of the big cloud providers, of course, they're sitting on over a trillion
11:07
Unidentified speaker (speaker_2): dollars of signed, contracted backlog.
11:13
Unidentified speaker (UNKNOWN): And
11:13
Unidentified speaker (speaker_2): look, this is only at the very start. I would say, you know, we're we're we're going from, if you think about it from a consumer demand or a business demand point of view, I think we're in the less than one AI agent per person.
11:27
Unidentified speaker (speaker_2): A lot less than one. And I think we're going to move into multiple AI agents per person, you know, over over the next few years, right? So that's my demand argument. I'll dive into that in more details. On the supply. side, they were just sold out. Um, the world's biggest AI memory maker, S.K. Heindicks, they're
11:46
Unidentified speaker (speaker_2): sold out through to 2029,
11:52
Unidentified speaker (speaker_2): which is just staggering. Uh, you look at other raw materials, copper, and actually in my research here, tin, I hadn't quite appreciated. These are two really key metals that are required. Tin is a really big component in the whole data center build-out point of view. They're both right up at record prices. Yes, they came off a little
12:12
Unidentified speaker (speaker_2): bit yesterday. Very true. But they're but buying it, you know, to all intents and purposes, still up at around their all-time highs. So um so really, you know, you've got AI labs like Anthropic, they're actually rationing usage at the moment, right? So I would say from a from a demand and a supply point of view, the dynamics are very
12:32
Unidentified speaker (speaker_2): strong. And just finally, before I hand over to you for your opening argument, um, everyone in this AI stat,
12:41
Unidentified speaker (speaker_2): every player on every layer is making money.
12:46
Unidentified speaker (UNKNOWN): And
12:46
Unidentified speaker (speaker_2): actually, there's some pretty interesting profit margins on each layer. Like NVIDIA, 75% gross margins. Take data center landlords, they're actually raising their rents, so their margins are going up. Cloud providers, you know, the big, you know, the big guns, AWS, Google, Microsoft, they're basically running operating operating
13:06
Unidentified speaker (speaker_2): margins of 36 to 39%. Sent, And then finally the big kind of the big reveal in Anthropics S1. they're making an operating profit.
13:16
Unidentified speaker (UNKNOWN): So
13:17
Unidentified speaker (speaker_2): you've kind of got four layers there, right? You've got the chips, you've got the data centers, you've got the cloud providers, and then finally you've got the model providers, and they're all making profit. And so I would say my argument is that this is very much a bullish secular shift. And yes, we're going gonna to get wobbles
13:37
Unidentified speaker (speaker_2): and volatility,
13:38
Unidentified speaker (UNKNOWN): like
13:38
Unidentified speaker (speaker_2): we've definitely
13:39
Unidentified speaker (speaker_2): just had, but I'm staying strong.
13:43
Unidentified speaker (speaker_0): Okay. When I sell my business, I want the best tax and investment advice. I want to help my kids, and I want to give back to the community. Ooh, then it's the vacation of a lifetime. I wonder if my out of office has a forever setting.
13:59
Unidentified speaker (speaker_1): An IG private wealth advisor creates the clarity you need with. plans that harmonize your business, your family, and your dreams. Get financial advice that puts you at the center. Find your advisor at IGPrivatewealth.com.
14:13
Unidentified speaker (speaker_3): All right, Piers, that was that was brilliant and also extremely compelling. Uh I love I love your point about well, I mean it's just the n the size of these numbers is is is just staggering. Trillion dollars a backlog, uh profitable at all levels. I take it. Here's my
14:33
Unidentified speaker (speaker_3): riposte. And by the way, sorry, I've I've kind of gone into uh undergraduate essay mode here, so apologies. I got
14:42
Unidentified speaker (UNKNOWN): carried
14:43
Unidentified speaker (speaker_3): away. Here we go.
14:45
Unidentified speaker (UNKNOWN): All
14:45
Unidentified speaker (speaker_3): right, Piers. I remember when during my first year at Cambridge, we were lucky enough to witness one of the few remaining Phillips machines, which was a hydraulic model of the UK economy that
15:02
Unidentified speaker (UNKNOWN): used
15:02
Unidentified speaker (speaker_3): colored water in place of money, if you remember.
15:05
Unidentified speaker (speaker_2): Yes.
15:06
Unidentified speaker (speaker_3): The Phillips machine, or to be precise, the Monetary National Income Analogue Computer, was built by this. uh Kiwi engineer turned economist Bill Phillips in 1949 and served to show thousands of economics undergraduates how the circular flow of the
15:25
Unidentified speaker (speaker_3): economy really worked.
15:28
Unidentified speaker (speaker_3): Phillips was building this thing, which is quite a cool piece of kit, and we'll put a uh photo of it on the um
15:34
Unidentified speaker (UNKNOWN): There's one in the There's one in
15:35
Unidentified speaker (speaker_2): the Bank of England Museum.
15:37
Unidentified speaker (UNKNOWN): Is
15:38
Unidentified speaker (speaker_3): it working?
15:39
Unidentified speaker (speaker_2): I'm not sure if it's working if anybody's ever in town by Bank Tube Station. Bank of
15:44
Unidentified speaker (UNKNOWN): England.
15:44
Unidentified speaker (speaker_2): That's why it's called Bank. Um they've got a museum, go check it
15:48
Unidentified speaker (UNKNOWN): out.
15:49
Unidentified speaker (speaker_3): Yeah, absolutely. And it's a it's a fascinating old school piece of kit. But anyway, so Phillips was building this thing at the time of peak Keynesianism, Keynesianism, post post-war, war, post post-beverage beverage reports, reports. And a nation of new garden towns and the birth of the NHS and the welfare state and things like
16:08
Unidentified speaker (speaker_3): that. And the Phillips machine showed how by opening opening the spigots of government spending, money would trickle down into the pockets of individuals who would play their economic role as consumers and start buying more. More consumption resulted in more money flowing to the PLs of companies. and ultimately back to the government in the form of
16:28
Unidentified speaker (speaker_3): new taxes.
16:29
Unidentified speaker (UNKNOWN): You
16:30
Unidentified speaker (speaker_3): know, Keynesian economics, right? Yeah. Um according to economic logic, the economy will also grow when the taps of investments and consequently, one would hope, increased productivity are also switched on. This is where AI comes in. So the entire outsized
16:49
Unidentified speaker (speaker_3): AI bet. is premised on whole economy productivity gains resulting from significant capital investments. If this turns out not to be true, then much of this investment could and probably will be wiped out. That's
17:08
Unidentified speaker (speaker_3): $570 billion of AI-related borrowing, according to. Morgan Stanley.
17:16
Unidentified speaker (UNKNOWN): And
17:16
Unidentified speaker (speaker_3): that number doesn't take into account, obviously, the equity capital that's bound up in this story, which is probably orders of magnitude higher. And to be totally honest, so far the productivity gains are nowhere to be seen outside of the Mag 7 and this increasingly circular network of
17:36
Unidentified speaker (speaker_3): AI companies, right?
17:38
Unidentified speaker (UNKNOWN): From the
17:38
Unidentified speaker (speaker_3): pri the picks and the shovels to the labs and the favs.
17:42
Unidentified speaker (UNKNOWN): So
17:42
Unidentified speaker (speaker_3): what are we left
17:43
Unidentified speaker (UNKNOWN): with? When
17:44
Unidentified speaker (speaker_3): it comes to the AI build-out, forget the Philips machine and the circular flow and think instead of the circular washing machine. You like that piece. In the incestuous world of artificial intelligence, money circles round and around. I'm trying to get a job as a journalist here. Uh Through equity investments and
18:04
Unidentified speaker (speaker_3): loan guarantees, through compute agreements and rental deals for spare capacity. This investment and the resulting expenditure produced extremely narrow, very short-term GDP growth, growth which doesn't trickle across to the rest of the real economy.
18:21
Unidentified speaker (UNKNOWN): As
18:22
Unidentified speaker (speaker_3): Katie Martet put it, Katie Martin put it in the FT, in a crisis. circular financing becomes a circular firing squad. And I think we're starting to see a little bit of that at the moment. But wait, I want to pull back and get into your territory, Piers, and talk about the wider economy.
18:42
Unidentified speaker (speaker_3): We started by talking about government spending as a key lever in the Phillips machine, stepping in to stimulate the economy in times of capital. retrenchment. Here is where I'm probably most pessimistic. Governments, as you well know, from Europe to the US, no longer have the capability for fiscal intervention in their
19:02
Unidentified speaker (speaker_3): economies. Whatever Andy Burnham says about not being in hoc to the bond markets, the fact of the matter is: just as I am in hoc to my mortgage provider, we are, right? So what does this mean? It means we can't spend our way out of an economic crisis or a prolonged recession. Due
19:22
Unidentified speaker (speaker_3): to a catastrophic White House pot uh set of policies, we're left with persistent inflation, a result of oil prices and trade wars. This keeps interest rates unnaturally high for this stage in the business cycle, this stage of the economic cycle, which leaves government with limited. wiggle room to spend their way out of
19:41
Unidentified speaker (speaker_3): trouble. We saw this with last week's budget here in the UK. We're seeing this in France at the moment. We see this in the US's two trillion dollar 2026 deficit. How bad could this all get?
19:56
Unidentified speaker (UNKNOWN): Well,
19:57
Unidentified speaker (speaker_3): in the short term, it could be very bad. According to investor Paul Kedroski, AI has reached its Minsky moment when credit growth runs out of good projects and starts funding bad ones. As with any bubble, we are simply desperate for the party to
20:16
Unidentified speaker (speaker_3): keep going,
20:17
Unidentified speaker (UNKNOWN): the
20:18
Unidentified speaker (speaker_3): equivalent of shouting one more song at a beleaguered provincial DJ. But when the bubble does burst, or maybe more, optimistically deflate, we will be left with a few very challenging years.
20:33
Unidentified speaker (UNKNOWN): Alright.
20:34
Unidentified speaker (speaker_3): Oof. There we go. Get me a job at the FT. Alpha
20:39
Unidentified speaker (UNKNOWN): Ville.
20:39
Unidentified speaker (speaker_3): I'm coming for you. I
20:41
Unidentified speaker (speaker_2): see what you're
20:41
Unidentified speaker (UNKNOWN): doing.
20:42
Unidentified speaker (speaker_2): You're kind of going macro here. So
20:45
Unidentified speaker (speaker_2): and
20:46
Unidentified speaker (speaker_2): excuse me, I think that yeah, I've kind of had this feeling as well. over the last the last months. It's been that sort of um
20:57
Unidentified speaker (speaker_2): that almost like a two-horse race where you've got on the one side you've got the economy, the broader economy, and then on the other side you've got this AI secular revolution. And it's like, well, can the AI can the productivity gains
21:13
Unidentified speaker (UNKNOWN): of the
21:14
Unidentified speaker (speaker_2): AI revolution?
21:15
Unidentified speaker (speaker_2): come fast enough to actually solve all of these fiscal and broader kind of macro issues? And I
21:24
Unidentified speaker (UNKNOWN): think,
21:25
Unidentified speaker (speaker_2): yeah, I mean, I think you've got definitely got some strong arguments there. It's the timing of it all. I want to say cancer with two things.
21:34
Unidentified speaker (speaker_2): Currently, at least as I understand it.
21:39
Unidentified speaker (speaker_2): the AI boom has not
21:42
Unidentified speaker (speaker_2): relied on government funding. Uh well, I mean you might argue differently in China, but let me just talk like let's say US, right? At the moment, it hasn't relied on any government funding. It has been private or privately backed, pri the private sector backing um and obviously
22:02
Unidentified speaker (speaker_2): the hyperscalers are stumping up the cash. Um and so I think that whilst you are 100% right, fiscally, these governments are not in any position to, you know, start uh ramp up spending and stimulate. Um I do think these private businesses have a war chest, and I know cash
22:22
Unidentified speaker (speaker_2): flow issues are starting to raise some alarm bells and stuff. so on. Um, but I think the privately funded AI boom is still there. I would I would then say, like, is it the case that when you run out of stuff to, you know, when credit runs out of good places to go,
22:42
Unidentified speaker (speaker_2): does it then just end up going to bad, you know, projects?
22:48
Unidentified speaker (UNKNOWN): And
22:48
Unidentified speaker (speaker_2): I go back to that point that everyone in the stack is now. making money. So is a bad project one that you get a return on investment from? And I I think that therefore
23:01
Unidentified speaker (speaker_2): my opinion is this this train is still in motion.
23:05
Unidentified speaker (speaker_3): Is it that everyone in the stack is making money because we are yet to accept that we are we hope that this wider economy productivity boost will happen in the short or at least in the
23:25
Unidentified speaker (speaker_3): medium term. And therefore the equity investors, the VC guys, the private equity guys, the private credit guys are willing to keep rolling the dice in the hope that this follow because that is that is the propulsion that is resulting in everyone quote unquote making money. That money is
23:45
Unidentified speaker (speaker_3): coming, if you think that money's not coming necessarily from spending in the real economy on AI yet. It's starting to a little bit, you know, clawed, impressive, but still really, really small relative to the magnitude of the build-out. So that profit is coming from. capital as opposed
24:05
Unidentified speaker (speaker_3): from demand.
24:07
Unidentified speaker (speaker_2): Well, I I look at, you know, I look at the use cases out there. You know, we are very much, I would say, still in the early, and I would probably even go as far as say the very early adoption phase. No. Now
24:24
Unidentified speaker (UNKNOWN): And
24:24
Unidentified speaker (speaker_2): my my brother's a good example. Now, how many episodes of this show have we done? I I was guessing at 250 plus no
24:32
Unidentified speaker (speaker_3): more.
24:32
Unidentified speaker (UNKNOWN): He was reminding me. Oh,
24:33
Unidentified speaker (speaker_2): is it that many? Wow. Okay.
24:35
Unidentified speaker (UNKNOWN): Yeah, I know,
24:35
Unidentified speaker (speaker_3): I know. Um
24:36
Unidentified speaker (speaker_2): yeah,
24:37
Unidentified speaker (speaker_3): for for a forty forty for a 44.8 year old, you've done a lot of that
24:41
Unidentified speaker (speaker_3): He
24:41
Unidentified speaker (UNKNOWN): was
24:41
Unidentified speaker (speaker_2): reminding me last weekend that I'd never given him a shout out um on the podcast. So Dean,
24:47
Unidentified speaker (speaker_2): my brother, here's your shout out, here's your moment. I'm gonna give g use you as a as a an example in my argument here. So he runs a manufacturing business. Um he d he builds like PCBs, right? So electronic manufacturing business. His business, he's got like 300 employees, roughly. I'm gonna get these numbers slightly wrong.
25:07
Unidentified speaker (speaker_2): And I would say he was telling me at the weekend that right now there's probably about three employees that use AI. tools. Three. And those three are like the absolute champions, and they are all over it in inverted commas and are starting to now really
25:27
Unidentified speaker (speaker_2): make some headway in terms of making processes way more efficient. And they're about to now, you know, really roll this out and um and drive a one. wider adoption. So they're going to go from three users to something like 30 users like next year. Okay. And this is going to drive
25:47
Unidentified speaker (speaker_2): huge efficiencies and therefore productivity gains for that business. I think that's just quite that's probably a very typical
25:56
Unidentified speaker (speaker_2): situation. And so I think, hang on, I've got a s I've got a backup point as well.
26:00
Unidentified speaker (speaker_3): Okay. All right.
26:01
Unidentified speaker (speaker_2): That's one point.
26:03
Unidentified speaker (UNKNOWN): So
26:03
Unidentified speaker (speaker_2): I think that adoption phase is gathering momentum.
26:07
Unidentified speaker (UNKNOWN): Now
26:07
Unidentified speaker (speaker_2): look, there'll be some casualties there. In this adoption phase, people are trying it out. Everyone in s in other companies, and if that's like in in Amplify, to be honest, everyone uses AI tools.
26:19
Unidentified speaker (UNKNOWN): And
26:19
Unidentified speaker (speaker_2): to a certain degree, some of that spend is probably
26:24
Unidentified speaker (UNKNOWN): I
26:24
Unidentified speaker (speaker_2): mean, wasted is one word you could use. in that we're just trying it out, seeing if it's going to work. And so that's that adoption phase can be quite messy, right? But I would say
26:34
Unidentified speaker (UNKNOWN): this
26:34
Unidentified speaker (speaker_2): is a different point entirely. Um Elon Musk
26:40
Unidentified speaker (speaker_2): is planning to by the end of next year, is planning to have hit a run rate of building um one million Optimus road. robots per year, right? He'll be at that run rate by the end of next year. So obviously all of these robotic we're we're right at the start, I would say, of the robotic
27:00
Unidentified speaker (speaker_2): boom. And each of those robots are going to need AI licenses, let's say, to operate. So I think we're we're on the cusp of that demand. I In my view, the demand's only going gonna to go up,
27:14
Unidentified speaker (UNKNOWN): and
27:14
Unidentified speaker (speaker_2): therefore.
27:15
Unidentified speaker (speaker_2): the whole stack is gonna become even more profitable, so people are gonna want to carry on funding
27:20
Unidentified speaker (UNKNOWN): it. And
27:21
Unidentified speaker (speaker_2): so I think this thing can still
27:23
Unidentified speaker (UNKNOWN): go.
27:25
Unidentified speaker (speaker_3): Yeah. So I think if we were if we were doing this debate live at the Oxford Union, let's just
27:31
Unidentified speaker (UNKNOWN): say
27:32
Unidentified speaker (speaker_3): I I would I would I would have been nodding along with you uh during the uh the Dean, Brother Dean anecdote and the the optimization and the adoption phase. And then I would have given myself a nice big smile when you said Elon Musk and Optimus Robots. Because I think I would have just won the argument
27:51
Unidentified speaker (speaker_3): by you introducing the fantastical nature of this speculative bubble that we are seeing, right? I want to see cold, hard numbers, and I'm going to give you some. some real world stories, anecdotes, and some numbers about how this how volume doesn't necessarily equate
28:11
Unidentified speaker (speaker_3): to value.
28:12
Unidentified speaker (speaker_2): Okay.
28:13
Unidentified speaker (speaker_3): So there was that classic, there was that classic meter study. So M E T R 2025 studies that said that coders Has felt 20% faster with the AI, that the number of commits ended up being down 20%. They
28:31
Unidentified speaker (UNKNOWN): called
28:31
Unidentified speaker (speaker_3): it digital cocaine, right? And obviously.
28:36
Unidentified speaker (UNKNOWN): This is
28:36
Unidentified speaker (speaker_3): late 2025. I know that
28:37
Unidentified speaker (UNKNOWN): since
28:38
Unidentified speaker (speaker_3): then we've
28:38
Unidentified speaker (UNKNOWN): got to be a little bit more.
28:39
Unidentified speaker (speaker_3): Yeah, we've had, you know, codecs and and things like that. But there definitely is this feeling that more volume doesn't equal more value. I'm Going to give you two more examples. So new apps. In the first half of 2026, the apps, the number of apps released on the App Store doubled. Downloads
28:59
Unidentified speaker (speaker_3): increased by 2%. Right? So volume doesn't necessarily equal value. Thirdly, you would have thought that AI for s it would be kind of ground zero for US law firms, right? We we've got all of these venture back companies delivering brilliant, well,
29:19
Unidentified speaker (speaker_3): purportedly brilliant AI tools for for legal services. Build hours in 20 in 2025, in the first half of 2026, build hours per case kept rising. Right? So
29:35
Unidentified speaker (speaker_3): even though we have all of this productivity unlocked. unlock in a what you might consider to be a relatively a somewhat deterministic and therefore AI appropriate industry,
29:49
Unidentified speaker (speaker_3): filled number of hours per case kept on rising. So something is not adding up here in the real world. Worlds. And just to speak personally, I definitely do more with AI. than I used to, but I also feel more like a spinning top, right?
30:09
Unidentified speaker (speaker_3): Doing lots and lots of things relatively frenetically and wondering whether I'm heading in the right direction or creating real significant value. And that might just be from a personal perspective. But I'm still yet to see,
30:23
Unidentified speaker (UNKNOWN): you
30:23
Unidentified speaker (speaker_3): know, that major, major productivity unlock.
30:27
Unidentified speaker (speaker_2): Interesting. Yeah, that's an interesting one with the lawyers. I think
30:30
Unidentified speaker (UNKNOWN): that, you
30:31
Unidentified speaker (speaker_2): know, the way one of the way use cases I would use Claude, for example, is on legal matters uh regarding running of of our business, right? And then those legal matters can be quite broad. Um, it could be we're uh dealing with uh a new contracts. contract that
30:51
Unidentified speaker (speaker_2): we're having with a new customer, for example, an SOW. It might be, um, I don't know, we we've hired someone new and so there's an employment contract that we're doing. Or it might be a HR, you know, a different HR situation. Lots of legal things come up on a weekly basis. I'm able to now use Claude
31:10
Unidentified speaker (speaker_2): to at least get a lot further down the road than I would have been able to. Um
31:16
Unidentified speaker (UNKNOWN): now
31:16
Unidentified speaker (speaker_2): I know that's maybe suggesting that, well, the billable hours should have gone down then, because if if people are now being empowered by Claude to be their own lawyers, essentially. But I I think actually you can only go so far down the track as a kind of armchair expert that now Claude enables you to be. And and I think, is it the case that
31:36
Unidentified speaker (speaker_2): people are getting uh you know, they're going to court more
31:40
Unidentified speaker (UNKNOWN): because
31:41
Unidentified speaker (speaker_2): they're able to go down the track on their own far enough to see that actually I do have a case here, let's lawyer up. So there might be some of that going on. But um
31:51
Unidentified speaker (speaker_3): Yeah, and it might also be just by the by, it might also be that AI is causing much more harm than it's causing that it's providing good. You know,
31:58
Unidentified speaker (UNKNOWN): there
31:58
Unidentified speaker (speaker_3): might be a lot more there might be a lot more legal cases because fringes fridges
32:02
Unidentified speaker (UNKNOWN): are
32:02
Unidentified speaker (speaker_3): getting hacked in Sweden and France. Uh and there's gonna be a lot of uh legal cases there, cyber security issues the the world over, AI slot, deep fakes, all of this stuff. Maybe AI uh legal billable hours are going up because AI is infecting the world. And that's
32:19
Unidentified speaker (UNKNOWN): why
32:19
Unidentified speaker (speaker_3): so many Americans hate it.
32:20
Unidentified speaker (speaker_2): So this is like we're straying into that argument of, you know, how is it gonna disrupt the job market?
32:26
Unidentified speaker (UNKNOWN): You
32:26
Unidentified speaker (speaker_2): know, how is AI going to impact on human beings doing jobs? And I think when you look back at other revolutions, you you tend to get this phase where to start with, it doesn't eliminate human beings in jobs. It actually enables human beings in those jobs, they keep their
32:46
Unidentified speaker (speaker_2): jobs, but it enables them to be more productive and do more. I mean, actually, there's a really good stat. I'm gonna get the numbers wrong, so forgive me, I'm gonna roughly guess. But Microsoft,
32:58
Unidentified speaker (speaker_2): in the last 24 months, I believe their revenue's gone up by
33:05
Unidentified speaker (speaker_2): 50%,
33:07
Unidentified speaker (speaker_2): very broad, roughly, right? Their revenue has increased by 50%. And in that time, their headcount in employees, is flat.
33:17
Unidentified speaker (speaker_3): But you'll find, you'll find that that is a very very that is a Mag 7 or a tech,
33:25
Unidentified speaker (UNKNOWN): you know,
33:26
Unidentified speaker (speaker_3): VC backed tech company focused phenomenon. That doesn't mean that it won't trickle out. Um
33:32
Unidentified speaker (UNKNOWN): that we're But not. We're not
33:34
Unidentified speaker (speaker_3): seeing that in any in other econom in in other industries.
33:37
Unidentified speaker (speaker_2): That's why I brought up my brother.
33:39
Unidentified speaker (speaker_2): Because for me, that was like, oh, okay, that's interesting. And I will say there's one other strand to this, which is the agentic phase. And look, you know, Meta, I think the most downloaded app on the App Store was Meta's Muse.
33:55
Unidentified speaker (UNKNOWN): I
33:55
Unidentified speaker (speaker_2): think it was the number one app, which is this kind of agencing, agentic, you know, you you get this thing to do stuff for you in your personal. life, right? Well, this agentic phase, and look, is it gonna come through? Is it gonna be popular? Is there gonna be adoption or not? And how quick? Obviously, lots of question marks there. But typically, what
34:15
Unidentified speaker (speaker_2): happens if if you get an agent working for you, that agent will use more kind of compute and AI than you would as an individual. So if we do share. shift into this agentic phase, again, I think the demand curve is only not only sloping higher, I still think
34:34
Unidentified speaker (speaker_2): it's accelerating exponentially.
34:38
Unidentified speaker (speaker_3): The agentic phase is interesting. I I think I think a lot of this this AI thing comes down to whether you believe that AI is a modest productivity booster or it is a
34:53
Unidentified speaker (speaker_3): uh a workforce replacer, right? And if
34:56
Unidentified speaker (UNKNOWN): you
34:56
Unidentified speaker (speaker_3): read the SpaceX uh prospectus and they say the total addressable market for their AI products is 20 trillion, they are betting that this is a workforce replacer, an agentic workforce replacer. And it is almost one of those things that if you believe, you believe, and if you don't you don't. And
35:16
Unidentified speaker (speaker_3): you it's very hard to convince the other side. But I'm just gonna, I'm just gonna wrap wrap this up with
35:25
Unidentified speaker (speaker_3): maybe a comment and then a question. So I was reading about different types of bubbles, and I think there's two guys called Byr Bayer and Hobart and Tobias Huber. And he's got a book called Boom. And they talk about two different types of bubbles: the mean reversion bubble, which is a destructive bubble, which
35:46
Unidentified speaker (speaker_3): is about, you know, there is a financial miracle, promising returns, like the South Sea bubble, portfolio insurance, subprime mortgages, and you revert to the mean, and all of that value is destroyed, right? Because it's basically like a Ponzi scheme, right? But then you have the second type of bubble, which is an
36:06
Unidentified speaker (speaker_3): inflection bubble, which is a productive bubble, a genuine technological breakthrough, railroads, electricity, the internet. And I think we all agree that AI is one of those, which might result, like those other bubbles, in short short-term term pain, where we
36:26
Unidentified speaker (speaker_3): all get over. our skis from a credit perspective and from a valuation perspective. But then in five to 10 years' time, we see this kind of new crop of companies that have figured it out. You know, the green shoots coming after a bit of bit of a controlled burn. And that's
36:45
Unidentified speaker (speaker_3): where, you know, maybe we can. I want wanna to I just want us to end well. Uh Maybe uh we can. both be right, is what I'm gonna say. Yeah.
36:54
Unidentified speaker (UNKNOWN): It's
36:54
Unidentified speaker (speaker_3): gonna be bad in the short, good in the long.
36:56
Unidentified speaker (speaker_2): Yeah, and I think we're I think we're having that bad in the short now, but I don't think that necessarily the setup is right
37:05
Unidentified speaker (speaker_2): as in now, you know, start of November 2026. I don't think the setup is quite right for a big blowout episode like we had dot dot-com com bubble. bursting. It it might be that we come to one of those moments later. Um, but I get it's kind of back to my point that the supply and demand dynamics, they're kind of you well,
37:25
Unidentified speaker (speaker_2): do I say unique, maybe not unique, but right now they're still very powerfully pointing in one direction. And I think there's there's an interesting new s element to this that I don't think you can really look back on other bubbles and and kind of see a similar thing. And that that is the the
37:45
Unidentified speaker (speaker_2): the rollout, the supply, because that's what happened, you know, the internet, the dot com bubble crashed because or the railroad crash. We had this massive overbuild and the infrastructure build got ahead of the demand and then the whole thing crashed, right? There there's there's one throttling
38:03
Unidentified speaker (speaker_2): lever on all of this because in and it's out of the control of any of those people in the stack that I mentioned. And that is it's really at government level, and this comes back to the midterms. Data center build out, number one. Number two, just straight up power, like electrons to power any of this, right?
38:24
Unidentified speaker (speaker_2): These are these are very, very inverted. in very short supply. Have they just has the supply of those just got even shorter given the midterm election results? So I think that that there's a barrier to getting an infrastructure overbuild because of those third-party
38:43
Unidentified speaker (speaker_2): government forces that are kind of throttling it back. And I think that's an interesting dynamic we perhaps haven't seen in one of these other previous revolutions.
38:51
Unidentified speaker (speaker_3): According to Piers Curren, this time it's different. This
38:55
Unidentified speaker (UNKNOWN): is not
38:55
Unidentified speaker (speaker_3): going to be like the other bubbles we've seen before. Well, look, let's
38:58
Unidentified speaker (UNKNOWN): go.
38:59
Unidentified speaker (UNKNOWN): I
38:59
Unidentified speaker (speaker_3): think we did all right without Ant.
39:00
Unidentified speaker (speaker_2): Yeah, not bad.
39:01
Unidentified speaker (speaker_3): No, not bad. A little bit waffly.
39:03
Unidentified speaker (UNKNOWN): Well the comments
39:04
Unidentified speaker (speaker_2): the comments will will be the judge of that. And and actually, what do the listeners think? Who's on who's on my bull side? Who's on Steven's bear side? Get your comments in the in the chat. I'd be interested in
39:15
Unidentified speaker (UNKNOWN): that. Who's on
39:15
Unidentified speaker (speaker_3): Ant's side just having a lovely time in Paris? Yeah. Not worrying about all this market doom and gloom.
39:21
Unidentified speaker (speaker_2): Yeah, indeed.
39:22
Unidentified speaker (speaker_3): Good for him. Well, Pierce, thank you for your time. Until next time, let's not let's not leave it two hundred and fifty episodes.
39:28
Unidentified speaker (speaker_2): Yeah, agreed.
39:30
Unidentified speaker (speaker_3): All
39:30
Unidentified speaker (speaker_2): right. See you later.
39:31
Unidentified speaker (speaker_3): Take care.
39:32
Unidentified speaker (speaker_0): Should you Would grow a playoff beard? What about a playoff mustache? A playoff soul patch? Rated braided playoff moustache? mutton chops, or maybe just playoff stubble. Whatever you grow, it's up to you, just like it's up to you to ask your doctor about Wigovi.
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 Market Maker

Market Maker

Is Venture Capital in an AI Bubble? The $750 Billion Boom

Episode discussion: VC investment hits $750B in 2026 as AI valuations shift to trillion-dollar scale

Speakers’ claims: Speaker 0 reports that venture capital investment reached $750 billion in 2026, surpassing the 2021 record, driven by AI exuberance and interest rate environments. The speaker notes that valuation caps have shifted from $100 billion to $1-1.5 trillion for companies like OpenAI and Anthropic, causing upstream valuations to inflate disproportionately. Geographic concentration is accelerating, with US startups capturing over 75% of global VC funding, leading firms like Sequoia to close international offices. The speaker contrasts this with Ramtim Naimi's view that there is insufficient capital for the AI build-out, expressing skepticism about visible productivity gains in the real economy. Despite high internal rates of return, the speaker highlights a liquidity problem for limited partners due to few exits outside of top-tier deals. Top quartile VC firms continue to outperform due to signaling effects and entrenched networks that secure superior deal flow.

Venture CapitalArtificial IntelligenceStartup Valuations
29 min3 assets10 claims to verify
Market Maker

Nvidia's Jensen Huang Scandal & Anthropic's IPO Deflates to $1.4 Trillion

Episode discussion: Nvidia decouples amid CEO scandal while Anthropic targets $1.4T IPO valuation

Speakers’ claims: Speakers analyze Nvidia's rare 4.3% stock drop driven by an espionage scandal involving CEO Jensen Huang, noting the company's unique governance structure and role as an AI infrastructure financier. Simultaneously, they discuss Anthropic's IPO, which aims to raise $64 billion at a ~$1.4 trillion valuation, down from earlier whispers of $2 trillion. The episode covers the book-building process led by Goldman Sachs, historical funding rounds including Sam Bankman-Fried's involvement, and competitive dynamics with Google's TPUs.

NvidiaAnthropicIPO
43 min3 assets20 claims to verify
All insights