Benedict Evans analyzes Anthropic's economics, arguing AI infrastructure may become low-margin commodity while value accrues to application-layer software.
Summary
Benedict Evans analyzes Anthropic's IPO prospectus, noting that while revenue has grown 10x, the $80 billion loss is largely accounting fiction due to inverted operating leverage where R&D costs rise with revenue. He argues that AI infrastructure is currently a once-in-a-generation shortage allowing pricing at customer ROI, but warns this will collapse as supply catches up, turning models into low-margin commodities. Evans contends that value will be captured by specialized SaaS companies building on top of these models, similar to how enterprise software captured value after the web browser era. He highlights that hyperscalers are spending over 50% of revenue on CapEx, forcing debt raises, and compares the current cycle to European mobile operators who failed to capture value from 3G infrastructure. The discussion also covers the competitive advantages of data-rich incumbents like Google and Meta in consumer agents versus model providers.
Key takeaways
- 1 use cases and those prices and those models, the more they become interchangeable. Which is part of the reason why everything they built before April is only twenty percent of revenue.
Unidentified speaker (speaker_1) · 30:41
Episode sections
- 0:02 – 3:43
Intro, Sponsors, and Guest Intro
Steve Plapham introduces the podcast, sponsor AlphaSense, and guest Benedict Evans to discuss the Anthropic IPO prospectus and AI economics.
- 3:43 – 6:22
Anthropic Prospectus Analysis and Unit Economics
Evans discusses the prospectus as a Rorschach test, noting the $80 billion loss is accounting fiction and operating leverage is inverted.
- 6:22 – 8:42
Supply, Demand, and Pricing Dynamics in AI
Evans describes a once-in-a-generation compute shortage allowing pricing at ROI, with inference costs dropping 10x annually.
- 8:42 – 11:58
Hyperscaler CapEx and Historical Parallels
Evans compares hyperscaler spending to European 3G operators, warning that infrastructure owners may not capture value.
- 11:58 – 15:52
Value Capture: Infrastructure vs. Application Layer
Evans argues value will accrue to specialized SaaS companies, comparing AI models to web browsers.
- 15:52 – 25:09
Chip Lifecycles and Data Center Constraints
Discussion on server lifecycle extensions, power constraints, and hyperscalers spending over 50% of revenue on CapEx.
- 25:09 – 29:45
Commoditization of Models and Price-Performance Curves
Evans argues models are becoming commodities with sharp price-performance curves, citing 80% of Anthropic's revenue from recent models.
- 29:45 – 31:32
AI Use Cases and Model Economics
Speakers discuss the shift to newer, more expensive models, with pre-April revenue accounting for only 20% of total revenue.
- 31:32 – 35:35
Automation vs. Abstraction in AI
Evans describes AI value through automation and abstraction, with abstraction driving higher token consumption.
- 35:35 – 42:52
The Application Layer and Domain Expertise
Debate on value accrual, citing Harvey and the need for domain expertise in vertical-specific software.
- 42:52 – 46:08
Valuation Challenges and Aswath Damodaran
Evans critiques Damodaran's Uber valuation, highlighting the difficulty of forecasting new AI revenue streams.
- 46:08 – 52:32
Infrastructure vs. Application Value Capture
Evans compares AI to the dot-com era, suggesting OpenAI/Anthropic may be like Netscape, with value captured downstream.
- 52:32 – 56:27
Market Volatility and Consumer Adoption Signals
Discussion on market volatility, Meta's Muse AI adoption, and the requirement for 10x cost reduction before scaling.
- 56:27 – 1:13:20
Consumer Agents and Data Visibility
Speaker_1 argues OpenAI/Anthropic lack private user data, putting them at a disadvantage to Google, Meta, and Amazon.
Assets discussed
| Asset | Kind | Stance | Context |
|---|---|---|---|
| Meta PlatformsMETA | equity | neutral | this year, like Meta, Google, Amazon, and Microsoft combined. |
| AlphabetGOOGL | equity | neutral | this year, like Meta, Google, Amazon, and Microsoft combined. |
| AmazonAMZN | equity | neutral | this year, like Meta, Google, Amazon, and Microsoft combined. |
| MicrosoftMSFT | equity | neutral | this year, like Meta, Google, Amazon, and Microsoft combined. |
| Anthropic | private company | neutral | headline benchmarks, anthropic, OpenAI, and now Google and Meta are all pretty close |
| OpenAI | private company | neutral | to think about the the future for anthropic and and indeed OpenAI. |
| SpaceX | private company | neutral | Which sounded kind of stupid, but what they're actually saying is, well, all of that is money that some people are going to do some stuff with with AI and so pay something for. |
“Stance” is how the speakers talked about the asset, not a recommendation.
Claims worth verifying
The Anthropic IPO prospectus shows an $80 billion loss which the speaker describes as 'pure accounting fiction.'
Unidentified speaker (UNKNOWN) ·4:34
How to verify: Review the S-1 filing for Anthropic on SEC EDGAR to check the reported net loss and the breakdown of revenue vs. expenses. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic's revenue has grown 10x in a year.
Unidentified speaker (speaker_1) ·4:33
How to verify: Review the S-1 filing for Anthropic on SEC EDGAR to check year-over-year revenue growth figures. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Primary source80% of Anthropic's revenue comes from models built in the last six months.
Unidentified speaker (speaker_1) ·28:24
How to verify: Review the S-1 filing for Anthropic on SEC EDGAR for details on revenue concentration by model version or product line. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Primary sourceMeta, Google, and Microsoft plan to spend over 50% of revenue on CapEx in the current year.
Unidentified speaker (speaker_1) ·17:15
How to verify: Review the latest 10-K or 10-Q filings for Meta Platforms, Alphabet, and Microsoft on SEC EDGAR to compare CapEx to revenue. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Primary sourceHyperscalers plan to spend $750 billion on CapEx this year, up from $100 billion five years ago.
Unidentified speaker (speaker_1) ·17:15
How to verify: Review earnings releases or 10-K filings for Meta, Google, Amazon, and Microsoft to sum their announced CapEx guidance. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Primary sourceEuropean mobile operators spent €110-120 billion on 3G licenses in 2000.
Unidentified speaker (speaker_1) ·10:13
How to verify: Search historical records of 3G spectrum auctions in Europe (e.g., UK, Germany) from 2026-2026 for total license fees. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Ramp data shows the top 1% of Anthropic's customers spend $10-15k per employee per month, while the average is $15.
Unidentified speaker (speaker_1) ·7:27
How to verify: Search for public reports or blog posts from Ramp citing their data on Anthropic customer spending. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Inference costs for AI models are dropping 10x a year.
Unidentified speaker (UNKNOWN) ·7:07
How to verify: Monitor public benchmarks from AI providers (e.g., OpenAI, Anthropic) or third-party trackers for price per token over time. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Anthropic is more or less breaking even on a cash flow basis.
Unidentified speaker (speaker_1) ·5:36
How to verify: Review the S-1 filing for Anthropic on SEC EDGAR to check Operating Cash Flow or Free Cash Flow metrics. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Primary sourceThe speaker's previous podcast episode on server lifecycle extension was published three years ago.
Unidentified speaker (speaker_0) ·15:52
How to verify: Search the Behind the Balance Sheet podcast archive for episodes discussing server lifecycle or chip longevity. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Pre-April revenue represents only 20% of Anthropic's total revenue.
Unidentified speaker (speaker_1) ·30:41
How to verify: Anthropic S-1 filing (SEC EDGAR) or company press releases regarding revenue composition. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Primary sourceAswath Damodaran valued Anthropic at $1.5 trillion based on a drug discovery hypothesis.
Unidentified speaker (speaker_2) ·42:52
How to verify: Aswath Damodaran's valuation blog or published articles from 2026-08-13 to 2026-11-13. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Damodaran's 2014 Uber valuation estimated a 2024 TAM of $183 billion with 10% share ($18 billion).
Unidentified speaker (speaker_1) ·43:55
How to verify: Damodaran's 2026 blog post on Uber valuation and Uber's 2024 earnings reports. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Uber's actual 2024 gross mobility was $83 billion, with $75 billion in delivery and $5 billion in ads.
Unidentified speaker (speaker_1) ·44:07
How to verify: Uber Q4 2024 Earnings Report or 2024 Annual Report (SEC EDGAR). For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Primary sourceMeta's Muse AI has 10 million users, with 25% using it on any given day.
Unidentified speaker (speaker_1) ·54:13
How to verify: Meta Q3/Q4 2026 Earnings Call transcript or press release. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Meta requires a 10x cost reduction before scaling Muse.
Unidentified speaker (speaker_1) ·54:33
How to verify: Meta Q3/Q4 2026 Earnings Call transcript. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Rocks Law states that the cost of a cutting-edge semiconductor fabrication plant doubles every four years.
Unidentified speaker (speaker_1) ·49:21
How to verify: Industry reports on semiconductor economics or Robert Rocks' publications. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Anthropic has 1,900 customers spending more than $1 million each.
Unidentified speaker (UNKNOWN) ·31:03
How to verify: Anthropic S-1 filing (SEC EDGAR) or company disclosures. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Primary sourceDamodaran's 2014 Uber valuation was viral and criticized for underestimating TAM.
Unidentified speaker (speaker_1) ·43:25
How to verify: Archived web versions of Damodaran's 2014 blog post and subsequent critiques. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Meta's Muse is not daily active users (DAU) but a different metric.
Unidentified speaker (speaker_1) ·54:13
How to verify: Meta Earnings Call transcript clarification on user metrics. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Primary sourceOpenAI and Anthropic do not have access to user email, messaging, purchasing, or social media data, seeing only what happens in ChatGPT.
Unidentified speaker (UNKNOWN) ·56:53
How to verify: Review Anthropic and OpenAI privacy policies and technical documentation regarding data retention and access; compare with Meta/Google data collection practices. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Google search growth has held up well, contradicting predictions that ChatGPT would replace search.
Unidentified speaker (UNKNOWN) ·57:37
How to verify: Check Google's quarterly earnings reports and search volume trends for the period 2026-2026. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Google and Meta have seen double-digit percentage point gains in ad revenue from AI optimization.
Unidentified speaker (speaker_2) ·1:00:14
How to verify: Review Meta and Alphabet (Google) quarterly earnings reports for ad revenue growth and AI-related commentary. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Amazon generated $75 billion of ad revenue in the last 12 months.
Unidentified speaker (UNKNOWN) ·1:02:36
How to verify: Check Amazon's most recent quarterly earnings report for advertising revenue figures. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Goldman Sachs employees now include software engineers making up easily a third of the workforce.
Unidentified speaker (speaker_1) ·1:08:22
How to verify: Review Goldman Sachs annual reports or public statements regarding workforce composition and engineering headcount. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.
Do your own research
- Review Anthropic's S-1 IPO prospectus for detailed unit economics, customer concentration, and accounting treatment of R&D versus COGS.
- Check hyperscaler (Meta, Google, Amazon, Microsoft) quarterly earnings calls for CapEx guidance, debt issuance, and return on invested capital metrics.
- Analyze Ramp or similar B2B spend data to verify customer spend distribution (top 1% vs average) for major AI model providers.
- Compare historical infrastructure cycles (3G mobile, cloud computing) for value capture patterns between infrastructure owners and application developers.
- Verify Google's search growth trends and user engagement metrics to assess the 'ChatGPT replacing search' narrative.
- Assess Meta's 'Muse' AI agent adoption metrics, specifically daily active users versus total users and cost-per-query trends.
- Review Aswath Damodaran's valuation methodologies for private AI companies and his historical accuracy in tech valuations.
- Check semiconductor fabrication cost trends (Rocks Law) and chip failure rates to evaluate infrastructure lifecycle assumptions.
Full transcriptexpand
- 0:02
- Unidentified speaker (speaker_0): Hi, I'm Steve Plapham and welcome to the Behind the Balance Sheet podcast where we meet leading investors and commentators and educate ourselves about the world of investing and the world. Our mission is to remove some of the mystique around investing and improve our understanding of successful investors' strategies and
- 0:21
- Unidentified speaker (speaker_0): tactics. Behind the balance sheet is a research and training consultancy, the world's top investor. investors, from US firms like Fidelity in Wellington to Billy Gifford in Schroeder, UK, rely on us to train their staff. Check out our forensic analysis and valuation courses for institutional investment teams on our website. And
- 0:41
- Unidentified speaker (speaker_0): we're not just for teams, thousands of individuals have taken our Analyst Academy and other online investment training courses. You can find them on the website where you can also sign up for our free weekly newsletter. Check out our homepage behind the balance sheet.com. In the investment business, everyone has access to the
- 1:01
- Unidentified speaker (speaker_0): same data. Gaining an information edge comes from who can get through it faster and who spots what change before the rest of the market does. Alpha Sense just launched Super Analyst, an always always-on on agent. agent that works like a power user inside the platform on your behalf. It draws in over 500
- 1:20
- Unidentified speaker (speaker_0): million premium sources from company filings and earnings calls to broker research and expert transcripts, and it keeps working after you've closed your laptop. Think of it as an extension of your team. It remembers your projects, follows your process and your house templates, and every output links back to its source. So when a PM or
- 1:40
- Unidentified speaker (speaker_0): an investment committee asks where a conclusion came from, you can show them. So hand it the work that usually eats your week, like getting up to speed in the news sector, preparing for earnings across your book, or building the first draft of a research report or IC memo. Start your free trial at alphasense.com forward slash
- 2:00
- Unidentified speaker (speaker_0): BTBS. That's AlphaHyphenSense. alpha-sense.com forward slash BTBS for behind the balance sheet. Today I'm joined by my friend Benedict Evans, one of the most insightful tech analysts around, longtime publisher of a popular weekly newsletter, and former strategy director of Andresen Horowitz. I've
- 2:20
- Unidentified speaker (speaker_0): learned a lot from Benedict over the years. I think his particular skill is the ability to both understand the tech side of the industry and the financial ramifications of new developments. We discussed the publication of the Prospectus for the Anthropic IPO, which is our first real insight into the extraordinary economics of
- 2:40
- Unidentified speaker (speaker_0): the large language models driving artificial intelligence. There are some really fascinating questions here. Can anthropic and OpenAI develop sustainable competitive advantages? What happens when the cost of producing intelligence continues to fall very rapidly? And will the hyperscalers earn acceptable returns on
- 3:00
- Unidentified speaker (speaker_0): the hundreds of billions of dollars they're investing in AI infrastructure right now? We also discuss opportunities for specialist software providers, the implications for the big tech companies, and why the winners of the AI revolution may not be the companies investors currently expect. Benedict draws
- 3:20
- Unidentified speaker (speaker_0): some fascinating parallels with previous tech revolutions, particularly the internet and mobile telecoms. This is a conversation about the difference between technological success and investment performance, and I think it provides a really useful framework for thinking about investing in AI. I hope you'll really enjoy it. my conversation with
- 3:39
- Unidentified speaker (speaker_0): Benedict.
- 3:43
- Unidentified speaker (UNKNOWN): So,
- 3:43
- Unidentified speaker (speaker_0): Benedict, listen, it's been ages since we've had sat down and had a proper discussion. We sat down in New York about six months ago, I suppose.
- 3:52
- Unidentified speaker (UNKNOWN): Um,
- 3:53
- Unidentified speaker (speaker_0): welcome to the podcast. I I've been dying to interview you, and we're gonna going to talk about AI ananthropic and
- 4:01
- Unidentified speaker (UNKNOWN): how long
- 4:01
- Unidentified speaker (speaker_0): did it take you to read the IPO or you you used your favorite AI to read it for you?
- 4:07
- Unidentified speaker (speaker_1): I well I know this is a joke and there's like, you know, eleven million pages about how this is going to kill us all as well.
- 4:13
- Unidentified speaker (UNKNOWN): Um
- 4:13
- Unidentified speaker (speaker_1): I it kind of struck me that this is sort of a Rorschach blot in that if what you're uh interested in is debt and or you're interested in circular revenue or you're interested in operating detail. details. Um, whatever you want to know, you can sort of find stuff in there. So, you know, if you're an accelerationist, you can say,
- 4:33
- Unidentified speaker (speaker_1): well, you
- 4:34
- Unidentified speaker (UNKNOWN): know,
- 4:34
- Unidentified speaker (speaker_1): the revenue has grown 10x in a year and um or more, and a growing portion of the research is being done by Claude. If you are one of the diminishing band of people that thinks this is all kind of nonsense, well, there's like eighty $80 billion dollar loss if you don't bother to work out what that loss actually was, which is basically pure accounting fiction. Um
- 4:54
- Unidentified speaker (speaker_1): I think the challenge
- 4:57
- Unidentified speaker (speaker_1): I have is
- 5:00
- Unidentified speaker (speaker_1): what it doesn't tell us.
- 5:03
- Unidentified speaker (speaker_1): And
- 5:04
- Unidentified speaker (speaker_1): firstly, it doesn't tell us depth of usage and spread of usage. It tells us revenue and customers, but we there doesn't really say how well this is working outside of a narrow portion of people,
- 5:16
- Unidentified speaker (UNKNOWN): which
- 5:16
- Unidentified speaker (speaker_1): is partly and swapic can't see that because it's API usage. But it's also how much predictive value does this have. So we've had this complete transformation in this industry in the last 12 months, in that software using AI for software development suddenly got explosive product market fit, which is driving this
- 5:36
- Unidentified speaker (speaker_1): enormous driven this enormous growth in revenue. And means that the company is now more or less breaking even on a kind of cash flow basis.
- 5:45
- Unidentified speaker (UNKNOWN): But
- 5:45
- Unidentified speaker (speaker_1): we don't know what that's going to look like in the future. And the sort of, if you go and look at the PL,
- 5:53
- Unidentified speaker (UNKNOWN): it's
- 5:53
- Unidentified speaker (speaker_1): funny, I was I was talking about this with a friend yesterday. The fixed cost looks like marginal cost, and the marginal cost looks fixed, in that the the cost of inference, which is at the cost of
- 6:03
- Unidentified speaker (speaker_1): cost of revenue, has basically stayed flat as a revenue as increased NX because there's so much efficiency gains going on in the operations of this stuff. On the other hand, the training cost, which is your R D, which should be a fixed cost, has gone up in line with revenue. So if you didn't label the chart, you would think they were the other way around.
- 6:22
- Unidentified speaker (UNKNOWN): Yeah. Um And
- 6:24
- Unidentified speaker (speaker_1): so so there's two things, and we don't know, but the problem is we don't know what's going to
- 6:26
- Unidentified speaker (UNKNOWN): happen.
- 6:26
- Unidentified speaker (speaker_1): gonna happen either of those things. Um the easier thing to talk about is the pricing of the inference. So you're basically selling tokens, um, which is rather like selling bits in like 1998. Um right now there is a once-in-a-generation generation shortage because the software industry wants to buy far more of these things than you can
- 6:46
- Unidentified speaker (speaker_1): get. And so you can if you weren't making a profit now, your CFO isn't doing their job. because you can price at the customer's ROI.
- 6:55
- Unidentified speaker (speaker_1): But on the one side of the table, the combination of more capex, more model efficiency, more inference efficiency, more chips means that the cost is dropping sort of 10x a
- 7:07
- Unidentified speaker (UNKNOWN): year
- 7:07
- Unidentified speaker (speaker_1): or more. And on the other side, um, agentic means one, two, or whatever. orders of magnitude, maybe more tokens demand. And the usage is still broad but really very shallow in that like the the really kind of intense usage is very narrowly concentrated. So I think Ramps data and Ramps data
- 7:27
- Unidentified speaker (speaker_1): skews to startups and early adopting technology companies said that like the top one percent of their customer base is spending fifteen, 15, ten, 10, fifteen 15 grand a month per employee, but the average is fifteen. $15.
- 7:39
- Unidentified speaker (speaker_0): Yeah.
- 7:41
- Unidentified speaker (speaker_1): So
- 7:41
- Unidentified speaker (UNKNOWN): you've got this very
- 7:42
- Unidentified speaker (speaker_1): small I what I was going to say is like on the one hand, like you've got this very, very small portion of people who are using this a lot, but you don't have a billion people using this a lot, like in Instagram. So you know the demand is going to do this, you know the supply is going to do this, you don't know what the balance will be, and so you don't know what the pricing's going gonna to look like. You also don't know how the research is how this how the training is going
- 8:02
- Unidentified speaker (speaker_1): to evolve. You don't know. if the scaling will continue. You don't know if they're going to continue have to having to spend more and more money to get a better and better model. And you don't know whether they're going to be able to differentiate and have some kind of competitive diff competitive advantage over model companies. So you can say today this is what they're spending,
- 8:21
- Unidentified speaker (UNKNOWN): but
- 8:21
- Unidentified speaker (speaker_1): you don't know what any of those are going to settle out at in two, three, five years' time.
- 8:25
- Unidentified speaker (speaker_2): No, sure. No, I mean there's obviously we're and spoiler alert, we're not gonna solve the world's problems and the valuation of anthropic in the next half hour, 45 minutes, because there's a lot of unanswered questions. But let's try and figure out what we're looking for in terms of a roadmap because I think that'll
- 8:42
- Unidentified speaker (speaker_0): be
- 8:42
- Unidentified speaker (speaker_2): the the really helpful thing to listeners. And the thing
- 8:45
- Unidentified speaker (UNKNOWN): that, you
- 8:45
- Unidentified speaker (speaker_2): know, people always start
- 8:47
- Unidentified speaker (UNKNOWN): with
- 8:47
- Unidentified speaker (speaker_0): is the hyperscalers are spending all this money. And what does that mean in terms
- 8:54
- Unidentified speaker (UNKNOWN): of
- 8:55
- Unidentified speaker (speaker_0): how how much does anthropic and OpenAI need to sell in order to justify that spending? And I I I don't know if you have a perspective on this. I've seen figures of three trillion dollars by twenty thirty or two trillion dollars by twenty thirty, but whatever they've spent, they presumably want to earn a return. and that means
- 9:15
- Unidentified speaker (speaker_0): quite a lot of revenue. So half a billion dollars, three quarters of a billion dollars of revenue. Does that make sense? And is there any hope that by 2030 Anthropic and OpenAI will be generating that sort of level of turnover?
- 9:34
- Unidentified speaker (speaker_1): So I think there there's may maybe be two things within this. The first is this is a slightly iterative. calculation in that you don't really know what the run rate capex is. There's a sort of presumption that we've got this huge hump of capex now and then it'll kind of recede somewhat and the revenue will catch up. But like is the run rate capex half a trillion or a trillion?
- 9:53
- Unidentified speaker (UNKNOWN): I mean
- 9:53
- Unidentified speaker (speaker_1): it kind of depends like how much demon there is.
- 9:57
- Unidentified speaker (UNKNOWN): So
- 9:57
- Unidentified speaker (speaker_1): that's one answer. Secondly, um you
- 10:01
- Unidentified speaker (UNKNOWN): know
- 10:01
- Unidentified speaker (speaker_1): I started my career in ninety 99, nine. and then I became a telecoms analyst in 2000. And at that point, the mobile operators in Europe spent about 110, 120 billion euros on 3G licenses.
- 10:13
- Unidentified speaker (speaker_1): And
- 10:15
- Unidentified speaker (speaker_1): many years later, it turned out that all the stuff they thought people would do did in fact happen. We do all of this. But they didn't make any money from
- 10:23
- Unidentified speaker (UNKNOWN): it. Um,
- 10:24
- Unidentified speaker (speaker_1): you know, I haven't run the calculation, but I'm pretty sure they most of those operators did not make a return on that investment, even though turned up mobile internet was kind of a big deal. So it's entirely possible that you get an enormous amount of adoption for this stuff at quite low prices, that you get a pricing collapse. And so you don't produce
- 10:44
- Unidentified speaker (speaker_1): a return on this investment. It's also possible you get an enormous amount of new revenue, but the people but the infrastructure itself is a low margin uh is low margin commodity infrastructure with no pricing power. And so the value goes to other people, which is kind of what happened in mobile, like Uber and Amazon and Instagram and everybody else make money
- 11:04
- Unidentified speaker (speaker_1): on mobile, and that money doesn't go to the mobile operators. And that
- 11:08
- Unidentified speaker (speaker_0): would be possible. It's possible that both the model providers and the infrastructure providers, the the the data center providers, neither of them make any money because they're they're
- 11:18
- Unidentified speaker (speaker_2): both
- 11:18
- Unidentified speaker (speaker_1): in
- 11:19
- Unidentified speaker (speaker_0): commodity
- 11:19
- Unidentified speaker (speaker_2): businesses.
- 11:20
- Unidentified speaker (speaker_1): Exactly. And you know, the the the wrinkle at the moment is that there's such an enormous imbalance of supply and demand that you have a more or less undifferentiated commodity that can be sold at high prices, which is obviously what happens if you're an oil company. Um Oil is oil, well, obviously oil isn't exactly oil, but you
- 11:37
- Unidentified speaker (UNKNOWN): know,
- 11:37
- Unidentified speaker (speaker_1): one barrel of Western West Texas is the same as another barrel of West Texas. But if there's more demand than supply, then you can price at a very high margin. That doesn't last.
- 11:47
- Unidentified speaker (speaker_1): So, you know, it's possible that supply remains behind demand for quite a long time. And so you could have quite high margins on with zero differentiation. So that's kind of a you know another moving part in the question. I think
- 11:58
- Unidentified speaker (UNKNOWN): the
- 11:58
- Unidentified speaker (speaker_1): there's a sort of an a devo um there's an entirely different conversation here, which is what is the AI product? Do you give everybody We're sort of at the stage now where you gave everyone a model and for some people that was life-changing, which is kind of like giving everybody a spreadsheet in the mid 80s or giving everybody a web browser in the mid-90s. Like
- 12:19
- Unidentified speaker (speaker_1): spreadsheets were kind of life changing for five or 10% percent of people in the mid 80s, early 80s.
- 12:23
- Unidentified speaker (UNKNOWN): I
- 12:23
- Unidentified speaker (speaker_2): remember
- 12:25
- Unidentified speaker (speaker_1): but
- 12:25
- Unidentified speaker (UNKNOWN): that
- 12:25
- Unidentified speaker (speaker_1): that wasn't how you re rebuilt your business around computers and changed your financial operations. That was SAP and Oracle for the sake of argument. It wasn't like that you completely changed how you manage your internal financial operations by just giving all of the people who were doing paper and running mainframes spreadsheets. That wasn't what happened. Same thing with the internet. You gave everybody in the company web browser.
- 12:45
- Unidentified speaker (speaker_1): That wasn't how you built an online business. That was something else.
- 12:48
- Unidentified speaker (UNKNOWN): Um,
- 12:48
- Unidentified speaker (speaker_1): The same thing now you gave everybody Copilot or Claude or ChatGPT. That isn't how you work out how AI transforms your business. Um, does that happen as
- 12:58
- Unidentified speaker (UNKNOWN): in
- 12:58
- Unidentified speaker (speaker_1): some with things that are fairly close to Claude and ChatGPT and Gemini and everything else? Or is there a whole new world of enterprise software companies and consumer software companies that use models as low-margin low margin commodity infrastructure somewhere inside, kind of like cloud is? I mean, cloud is a good business,
- 13:16
- Unidentified speaker (UNKNOWN): you
- 13:16
- Unidentified speaker (speaker_1): know, but for some people, but you know, it does not set the agenda or control the entire internet. Um and so there's an outcome in which a high portion of what you do is something that's more or less clawed or more or less chat GPT, and they capture the value from that. And there's another outcome
- 13:36
- Unidentified speaker (speaker_1): in which no
- 13:38
- Unidentified speaker (UNKNOWN): it
- 13:38
- Unidentified speaker (speaker_1): it's Harvey, but lawyers buy Harvey or something else. And that the typical big company today in America has four or five hundred SaaS apps, and in ten years' time they will have a thousand apps, half of which are running on AI models, but they're not paying Flawed and they're not paying Anthropic, they're paying a
- 13:58
- Unidentified speaker (speaker_1): thousand new software companies.
- 14:00
- Unidentified speaker (UNKNOWN): Um,
- 14:00
- Unidentified speaker (speaker_1): And so the value goes further up the stack. stack. That's how it's always worked in the past.
- 14:04
- Unidentified speaker (speaker_2): Yeah.
- 14:05
- Unidentified speaker (speaker_1): Um and
- 14:07
- Unidentified speaker (speaker_1): we don't kind of know how this is going to evolve. This is part of the challenge. If you're kind of super accelerationist, you would say, look, the models are just going to get so much better that they'll be able to do the whole thing themselves. Or at any rate, the CIO will be able to get them to do the whole thing himself or herself. And so most of the value of all of that reconfiguration of how the enterprise works.
- 14:27
- Unidentified speaker (speaker_1): will be captured by the model companies themselves. There's another view, as I said, which is kind of no, I don't think that's how any of this ever works. Um, and that this will be infrastructure that is where most of the value gets captured by the people further up. Part of the challenge here
- 14:43
- Unidentified speaker (UNKNOWN): is
- 14:44
- Unidentified speaker (speaker_1): unlike every other platform The shift, we don't have a good understanding of how the technology works. Like we don't have a good first principles understanding of why this works so.
- 14:51
- Unidentified speaker (UNKNOWN): well.
- 14:52
- Unidentified speaker (speaker_1): So we knew with the internet, well, we didn't know that Google was going to get created, but we knew that PCs were expensive and telcos weren't going to give everyone fiber broadband next week. And like we knew the basic parameters of what was going on. Whereas with this, we don't know what the science is going to produce next. And so we don't know how good the models will get, on what dimension, or what cost.
- 15:12
- Unidentified speaker (speaker_1): And so we don't kind of know what's going to happen. back to the PL or the Anthropic S1. We don't really know what that R and D line is going to look like. Because there's like a worst case scenario, which is the whole thing kind of flattens out next summer. And the R and D cost drops off the PL, but so does the revenue growth.
- 15:31
- Unidentified speaker (UNKNOWN): Well,
- 15:32
- Unidentified speaker (speaker_0): I'm going to come back to the point points you were raising about who captures the value, because I think that's very important. But just before we get there, I just want to just finish off on the hyperscalers because I wrote, I don't know, three years ago about the fact that the these companies were extending the
- 15:52
- Unidentified speaker (speaker_0): the lives of their servers. And so they were they started off to pre setting their servers over three years and they ended up after several changes to pre pre-sating setting them over five or six years, depending on the on the company. And I pointed out that I didn't know how long a chip lasted, but I knew which direction they were going and what they were doing was they're boosting their earnings growth.
- 16:11
- Unidentified speaker (speaker_0): And so I go, well, you know, this is I I don't know what the right answer is, but what I do know is that they're accelerating their earnings growth and making their companies look more voluble. And that that's quite dangerous from the perspective of an investor. But it's now become like a series. serious issue. I mean, when I was talking about this, nobody's worried about service
- 16:31
- Unidentified speaker (speaker_0): spending that much. Um now Michael Burry's been been you know shed the spotlight on this. Just give us a sense uh your sense of if the chips last five years,
- 16:44
- Unidentified speaker (UNKNOWN): what
- 16:44
- Unidentified speaker (speaker_0): happens in year six? They just use them for inference or reduced demand, you know, not as high.
- 16:51
- Unidentified speaker (speaker_0): pressure kind of applications. They're still useful. Or do they need to just go back and and and put a whole new set of chips in?
- 17:01
- Unidentified speaker (speaker_1): So I think there's there's probably two or four bits to put things to threads to pull on here. First of all, I mean, as you point out, like five years ago these companies probably spent a hundred $100 billion dollars on CapEx this year, like Meta, Google, Amazon, and Microsoft combined.
- 17:15
- Unidentified speaker (speaker_1): plan to spend about $750 billion this year. Five years ago was $100 billion. So there's this enormous surge in CapEx. And that means that three of those four, like the other than the one Microsoft, so Meta, Amazon and Google, sorry, Meta, Microsoft and Google are all going to spend over 50% of revenue on CapEx this year. And so one of the big structural changes is that the CapEx has
- 17:35
- Unidentified speaker (speaker_1): outgrowing outgrown the free cash for these companies. And so they've had to go to the capital markets, which is why suddenly this becomes people start talking to credit analysts. Like, well, well, how much credit is there in the world? Like, like no one, like you didn't have to pay attention to
- 17:46
- Unidentified speaker (UNKNOWN): this
- 17:46
- Unidentified speaker (speaker_1): if you're a tech analyst before. Um, like what what what is what does Ig and H Y stand for, please? Like those kind of questions, suddenly, like you have to ask. Um, I think secondly, um
- 17:59
- Unidentified speaker (speaker_1): you've got this continuous acceleration in efficiency of running the models.
- 18:03
- Unidentified speaker (speaker_1): And so between more CAPEX, but also better chips and better models, you've got like you've got the capacity, raw capacity people are measuring in gigawatts or megawatts, like how many gigawatts of data center have you got? But then there's how many tokens can you get um per gigawatt?
- 18:23
- Unidentified speaker (speaker_1): Because the chips get better, and the models get better. And then there's how good a result can you get from a token? And so there's sort of a lot of moving parts in the return on all of that investment.
- 18:36
- Unidentified speaker (speaker_1): Then you have, and of course, then what can you charge ultimately for the results? Are you selling tokens? Are you selling results? What are you selling? What can you charge for that? Do you have differentiation? Um, so that's one piece. The second is um
- 18:50
- Unidentified speaker (speaker_1): at what point do you decide it's a better overall return to throw those chips out and put in new ones that get better results per watt? And obviously that's a spreadsheet calculation and a capacity and a timing infrastructure, you know, that starts for infrastructure people to kind of work out like does it make sense for us to rip that stuff out or to leave the stuff that's in there and put new stuff in a new data
- 19:10
- Unidentified speaker (speaker_1): center and a new shell? And obviously part of the character the the calculation calculation data. there is your ability to get access to power, because if you know if you could have as much power as you were willing to pay for, then it would be easy, you would be more likely to leave the old ones in place if they were producing an economic return on their own right in their own right. Whereas if you're power constrained, then
- 19:30
- Unidentified speaker (speaker_1): it makes sense to put in new chips that get a higher um return per watt. Um And so that's even without asking, does the thing actually work? work?
- 19:40
- Unidentified speaker (speaker_1): Um
- 19:41
- Unidentified speaker (UNKNOWN): then
- 19:42
- Unidentified speaker (speaker_1): how what's the working life of chip of a chip?
- 19:44
- Unidentified speaker (UNKNOWN): You
- 19:44
- Unidentified speaker (speaker_1): know, does it actually start kind of packing in? Does the failure rate start rising? I mean I think the answer is you have a failure rate and then there's a curve. And as you you
- 19:51
- Unidentified speaker (UNKNOWN): know
- 19:52
- Unidentified speaker (speaker_1): over time the ch the failure rate starts getting worse. I don't know what that looks like.
- 19:57
- Unidentified speaker (UNKNOWN): I'm
- 19:57
- Unidentified speaker (speaker_1): not a data center thread analyst. I don't think anyone has a really, really good sense of what that looks like until you've just run these things maxed out for X amount of time. And then there are all sorts of things you can do to mitigate that and what does the cooling look like and so on. So you very quick get quickly get into a lot of quite nitty gritty technical questions on you know what is the
- 20:17
- Unidentified speaker (speaker_1): um what's the actual um mechanical life of these things as opposed to the useful economic you know the the life at which you're still making an economic econom economic return on these things. I think and that also of course depends on whether you're using them for inference or for training and what's the loading on your system and how fungible are they between inference and training. All
- 20:37
- Unidentified speaker (speaker_1): of which is a long way of saying I don't know go ask your data go ask a data center
- 20:41
- Unidentified speaker (UNKNOWN): analyst.
- 20:41
- Unidentified speaker (speaker_1): But
- 20:42
- Unidentified speaker (UNKNOWN): the big
- 20:43
- Unidentified speaker (UNKNOWN): things
- 20:44
- Unidentified speaker (speaker_1): there's also I suppose there's a there's a are we focusing on EPS or are we focusing on cash flow? Because obviously Amazon has always said, you know, ignore the EPS look at free cash flow. And if your objective is to manage capex to sales and to go out and raise debt to pay for that, um is that an EPS question or a capex or a cash flow
- 21:04
- Unidentified speaker (speaker_1): question? Um I mean one of the things that's occurred to me with these companies raising It's external capital is obviously the question is do you want to
- 21:14
- Unidentified speaker (UNKNOWN): if you
- 21:14
- Unidentified speaker (speaker_1): believe that you're going to grow a huge amount, then you would by default you would go with debt rather than equity. Because you're going to grow into the equ debt without giving out the equity, or you do a convert, maybe. As a credit investor, it kind of doesn't matter if Google
- 21:34
- Unidentified speaker (speaker_1): gets zero return from all of these data centers as long as the ad revenue is still there because they've got no debt and you know massive cash flow. So if the data centers all empt uh and end up as space eaters, as a creditor investor, you kind of don't care because the ad revenue will still be there. Um as an equity investor, obviously, well that was all of your money that's just gone into a hole in the ground.
- 21:54
- Unidentified speaker (speaker_1): Um but
- 21:54
- Unidentified speaker (UNKNOWN): there's a sort of
- 21:55
- Unidentified speaker (speaker_0): But if you look at the big picture,
- 21:58
- Unidentified speaker (UNKNOWN): the
- 21:58
- Unidentified speaker (speaker_0): big picture is that we're having trouble building enough data centers. So in five years' time, having a data center that's got power and all the rest of it, water, whatever,
- 22:12
- Unidentified speaker (speaker_0): that's going to be the constraint, right? So it's quite it makes it more likely that you're going to bin the old chips and replace them with new ones because you can't just go and build another data center.
- 22:24
- Unidentified speaker (UNKNOWN): Which would
- 22:25
- Unidentified speaker (speaker_0): normally be what you would do, right?
- 22:27
- Unidentified speaker (speaker_1): Yeah, exactly. Then of course the question becomes well, as I was saying, well, what is the supply, what does the demand for tokens end up looking
- 22:35
- Unidentified speaker (UNKNOWN): like?
- 22:36
- Unidentified speaker (speaker_1): And you know, this is a comparison people like to do is with fiber 25 years ago, which isn't quite right because fiber was built out massively um ahead of demand, whereas this is being built out behind demand. But what also happened with fiber is that it got massively disaggregated and fragmented, and then everything at every layer in the stack was being sold
- 22:56
- Unidentified speaker (speaker_1): at marginal cost. And that is certainly what could end up happening here.
- 23:02
- Unidentified speaker (speaker_0): Expert calls have always been one of the most powerful ways to build conviction. But today investors are asked to cover more companies, move faster, and do it with leaner teams. With Office Office's Sense's AI AI-led led expert calls, their TKIS call service team. sources experts based on your research criteria and lets the AI interviewer get to work. The magic is in
- 23:22
- Unidentified speaker (speaker_0): that AI interviewer, purpose-built and knowledge base informed to conduct high-quality, context-rich conversations on your behalf, acting as a trusted extension of your team. Then they take it one step further. Your call transcripts flow natively into your Office OfficeSense Sense. experience and become queryable, searchable,
- 23:42
- Unidentified speaker (speaker_0): and comparable. So your primary insights plug directly into earnings prep, diligence work streams, and pitchbooks with zero tool switching. And with OfficeSense's expert call services, the AI-led expert calls are just one option because they know the importance of a hybrid expert research approach. AI for coverage and
- 24:02
- Unidentified speaker (speaker_0): efficiency. humans for complexity and conviction. It's the institutional edge that scales research without scaling headcount. For hedge funds, that means validating thesis assumptions across dozens of experts before earnings instead of just a handful. For private equity, it means faster
- 24:22
- Unidentified speaker (speaker_0): pre-IOI scans and deeper commercial diligence. For investment banks, and asset managers, it means pulling real operator perspective straight into models and sector positioning without disconnected tools or manual handoffs. Although it lives inside the AlphaSense platform, trusted by 75% of the world's top
- 24:42
- Unidentified speaker (speaker_0): hedge funds, alongside filings, broker research, news, and more than 240,000 two hundred and forty thousand expert call transcripts, turning realm. raw conversations into comparable auditable insight. Take advantage of AlphaSense's AI led expert calls now. The first to see wins, the rest
- 25:01
- Unidentified speaker (speaker_0): follow. Learn more at alpha-sense dot com forward slash BTBS.
- 25:09
- Unidentified speaker (speaker_2): So let me uh just turn to the this idea of the commodities Um
- 25:16
- Unidentified speaker (UNKNOWN): are
- 25:16
- Unidentified speaker (speaker_2): the models commodities? Because what I don't really understand is
- 25:22
- Unidentified speaker (speaker_2): why would you I mean uh I know from my personal circumstances that I like to use Chat GPT when I'm wanting help with editing my newsletter and finding a title, and I don't find Claude so good at it, and I find Claude better at producing. spreadsheets than Chat
- 25:42
- Unidentified speaker (speaker_2): GPT. Presumably there are some some differences, but at the end of the day, is there a really distinct business with a mort as we recognize
- 25:53
- Unidentified speaker (UNKNOWN): it
- 25:53
- Unidentified speaker (speaker_2): in a model, in an LLM?
- 25:57
- Unidentified speaker (speaker_1): So the simple answer is today not really.
- 26:02
- Unidentified speaker (speaker_1): Um if you and it depended and it but then it depends what you're doing. So the aggregate benchmarks,
- 26:10
- Unidentified speaker (UNKNOWN): the
- 26:11
- Unidentified speaker (speaker_1): headline benchmarks, anthropic, OpenAI, and now Google and Meta are all pretty close and Chinese models are not very far behind and SpaceX is not very far behind, and for the sake of argument, we'll get back onto the top soon. When you get in into specific use cases, then you can start seeing more variance because
- 26:31
- Unidentified speaker (speaker_1): you're disaggregating those benchmarks. When you are pushing right at the frontier
- 26:36
- Unidentified speaker (UNKNOWN): and
- 26:36
- Unidentified speaker (speaker_1): you're doing something super hard that the model is only just able to do something out, then you can see differences. But then the question is: well, what are your use cases? So, you know, sort of pick kind of, you know, because you know, the interesting benchmark. curve is actually to look at price versus performance, where there's a very, very, very sharp curve. So you
- 26:56
- Unidentified speaker (speaker_1): can have absolutely the latest full fat model that's very expensive. You can have the model from six months ago that's sort of 90% on the benchmarks and 10% of the price. I'm making these numbers up, but that's kind of the story. So the curve does that. And at the extreme, like
- 27:14
- Unidentified speaker (UNKNOWN): voice.
- 27:14
- Unidentified speaker (speaker_1): recognition is done and it runs on your phone.
- 27:18
- Unidentified speaker (speaker_1): So as a company, whether you're a startup building something for consumers or you're whether you're building something for enterprise, what is it that you're trying to do? Is your objective to do um summarization of the call what's being said in the call center? That's not going to need the frontier model that won't appear for 10 ten years. Maybe next year.
- 27:38
- Unidentified speaker (speaker_1): year's frontier model will do that a bit better. But clearly, as you sort of actually separate out, if you try and say, well, we've got 500 SAS apps, what would an AI be what what model would be able to do that? Um, it's obviously it's never as simple as that. But like there's clearly there's going to be some use cases that work fine on the open source model from two years ago, or that work fine on
- 27:58
- Unidentified speaker (speaker_1): today's model, but only have an ROI when the prices come down by ninety 90% percent in six months.
- 28:05
- Unidentified speaker (UNKNOWN): And
- 28:05
- Unidentified speaker (speaker_1): then there's other things at the other extreme where things really, really will benefit from absolutely the latest model. I mean, this is the um, you know, this is the thing we should have mentioned earlier talking about the anthropic PL. 80% of revenue comes from models built in the last six months. And that's partly because people only want the model built in the last six months, and it's also because the model from a year
- 28:24
- Unidentified speaker (speaker_1): ago is such a commodity. and it's super, super cheap, and maybe it runs on prem and so you can't make any money from it. And there's a lot of use cases in that category. There's a lot of things companies are doing with those models. So the other extreme would be, you know, um go look at all of our customer emails and every recording with a
- 28:44
- Unidentified speaker (speaker_1): customer call and all of our usage telemetry and do a competitive review and tell me why our churn is changing.
- 28:52
- Unidentified speaker (speaker_1): And that might or might not work now, but you can imagine like that will get better answers with a better model for quite some time. But this is the kind of the question: is do you need the newest, fastest, fullest fan model? And for that, there's pricing power and maybe some differentiation. I like this the result I get from this one more than the result I get from that one for this
- 29:12
- Unidentified speaker (speaker_1): use case or that use case. But as soon as you start doing going down that curve, then they all blow into one. And if you are building
- 29:22
- Unidentified speaker (speaker_1): a tool to do that summarization of what are the trends in customer calls, then you're going to be looking at your price per token, you know, which model running on which cloud. And you're going to be swapping them in and out. And the enterprise. is going to be asking itself, just as it asked for
- 29:42
- Unidentified speaker (speaker_1): cloud, do we build this ourselves?
- 29:45
- Unidentified speaker (UNKNOWN): Or
- 29:45
- Unidentified speaker (speaker_1): for SaaS, do we build this ourselves or do we buy
- 29:49
- Unidentified speaker (UNKNOWN): it?
- 29:49
- Unidentified speaker (speaker_1): And the answer as it was with SAS was kind of depends. We're not going to build our own email server. We're not going to build our own cyber intrusion detection systems. We may well build our own thing to manage. that thing inside our payment proflows.
- 30:06
- Unidentified speaker (speaker_2): Well the answer to your qu to my question is that um there's a relatively small number of use cases, probably, that require the most expensive, most up to date, most profitable model.
- 30:20
- Unidentified speaker (UNKNOWN): Well
- 30:21
- Unidentified speaker (speaker_1): I'm not sure whether if it's small, but but I think the point is that though it's a distribution of many use cases. many models, many prices. And there are some use cases where you absolutely want the newest, full fat, most expensive model, and where you also may be able to see differences between them. And as soon as you start moving back down the curve of those
- 30:41
- Unidentified speaker (speaker_1): use cases and those prices and those models, the more they become interchangeable. Which is part of the reason why everything they built before April is only twenty percent of revenue.
- 30:53
- Unidentified speaker (speaker_2): So I mean just talk us through I mean what would be an example of a customer that needed to have that latest type of
- 31:03
- Unidentified speaker (UNKNOWN): model?
- 31:04
- Unidentified speaker (speaker_2): So they've got nineteen hundred customers that are spending more than a million dollars. What do you know, do you have a a feeling for what sort of companies those are? I mean, presumably some of them are
- 31:13
- Unidentified speaker (speaker_1): I think I it's
- 31:13
- Unidentified speaker (UNKNOWN): concentrated in I
- 31:14
- Unidentified speaker (speaker_1): think we all understand that this is very, very highly concentrated in software development. Um but I think that's you know, that's transitory, and this is one of the questions is how fast does that expand outside of software, out of software companies? The way that I've been sort of trying to think about this is that
- 31:32
- Unidentified speaker (speaker_1): you can think of AI as unlocking two things, maybe. One of them is automation and the other is abstraction. And so automation is faster cheaper, doing the old thing faster cheaper. Doing things that you needed people for, maybe that people couldn't do. Like listen to every call coming into the call center. Tell me what people are really
- 31:52
- Unidentified speaker (speaker_1): saying. Tell me if people are more angry. Are people more upset than they used to be? Are they happier with the product? Questions. And these are things that you kind of couldn't ask a database and that you didn't really type into Salesforce.
- 32:05
- Unidentified speaker (UNKNOWN): And
- 32:05
- Unidentified speaker (speaker_1): sometimes that's transformative. And this is obviously what you see in cyber now. If you throw 10,000 agents at your software at your code base and say try every possible vulnerability in every possible combination, then they'll find stuff that people could never have found before.
- 32:18
- Unidentified speaker (UNKNOWN): So
- 32:19
- Unidentified speaker (speaker_1): that's kind of if you do stuff faster and cheaper, that might just be price elasticity, but it might actually change the way the business works and it might unlock new stuff. The or indeed new kinds of competition. But the other side that I think is and so that some of that is okay, well then, you know, we will throw 10,000 agents at that and say we'll use way more tokens. The other side that I think is
- 32:39
- Unidentified speaker (speaker_1): more intellectually interesting is to say what are the new questions you can ask. And so there's this old saying in computer science that a computer should never ask you a question if it should be able to work out the answer.
- 32:51
- Unidentified speaker (UNKNOWN): And
- 32:52
- Unidentified speaker (speaker_1): that used to mean things like what printer do you
- 32:54
- Unidentified speaker (UNKNOWN): have?
- 32:54
- Unidentified speaker (speaker_1): But now it might mean you open Salesforce and it says, What question do you want to ask?
- 33:03
- Unidentified speaker (speaker_1): And you construct your carefully devised 35-step database query that produces a list of who you're going to call this week. Customers who've said this and done this and we called them at this time and they're using the product this much and this is how much they paid. You know, you construct your complicated database query.
- 33:22
- Unidentified speaker (UNKNOWN): But
- 33:23
- Unidentified speaker (speaker_1): what you actually want to know is who should I call this week.
- 33:26
- Unidentified speaker (UNKNOWN): And
- 33:26
- Unidentified speaker (speaker_1): so you
- 33:26
- Unidentified speaker (UNKNOWN): should
- 33:27
- Unidentified speaker (speaker_1): be able to ask the computer that you should be able to call say which customers are going to churn instead of doing a 30 30-step step query. You should be able to say what invoices are due this week and what invoices are overdue and why. Pick your business question as opposed to
- 33:44
- Unidentified speaker (speaker_1): pulling lots and lots of levers and switches and knobs to get the computer to give you the question. And you should be able to tell the computer what you want to do rather than give it lots and lots and lots of complicated criteria. That's new software. That's a new software company. And that's a lot of tokens.
- 34:05
- Unidentified speaker (speaker_1): Especially when you start saying, look, instead
- 34:09
- Unidentified speaker (UNKNOWN): of
- 34:09
- Unidentified speaker (speaker_1): asking this question to Salesforce, I'm going to it's got all of our app telemetry and it's got all of our recordings of all of our customer calls. And it's got all of our payment data. So ask me who's going to churn. And it should be able to work that out by looking at all of those things. And when you start building things.
- 34:27
- Unidentified speaker (speaker_1): like that, then you need a bigger, better model. You could imagine a bigger, better model is going to produce better results for those. And it's probably ten models that are changed together and built in various complex ways. When you start doing that, that is also probably not a copy of Claude running on the CFA's laptop. Any
- 34:47
- Unidentified speaker (speaker_1): more than the CFO ran inventory for a ten thousand. 10,000-person company from a copy of Excel running on their laptop. Yes, they have Excel, but that's not how that works. Yeah,
- 34:57
- Unidentified speaker (speaker_2): true.
- 34:57
- Unidentified speaker (speaker_1): That's something that needs to be built carefully and thoughtfully by a bunch of people whose job is to think about how this stuff works.
- 35:05
- Unidentified speaker (UNKNOWN): And
- 35:05
- Unidentified speaker (speaker_1): you're probably going to pay them. And they may not work for you. They may be a whole other company. That might be Accenture, it might be Bain, it might be a new startup, it might be SAP. It's not anthropic. Anthropic isn't going to be able to create all of those by themselves.
- 35:19
- Unidentified speaker (UNKNOWN): And
- 35:19
- Unidentified speaker (speaker_1): so all of that sort of gets to a hypothesis that says, well, you know, explain to me why this doesn't work like cloud. Explain to me why this isn't infrastructure, that other people have to go out and invent, work out what that stuff is.
- 35:35
- Unidentified speaker (speaker_2): But that's where the real added value will be. And I mean I mean my impression is that the that the models are largely commodity businesses and that the real added value will be in the application layer, the vertical, the the different industry verticals. And where the domain
- 35:54
- Unidentified speaker (speaker_2): expertise will be significant. So I was asking, I met um
- 35:59
- Unidentified speaker (UNKNOWN): the
- 35:59
- Unidentified speaker (speaker_2): Harvey CEO, Winston Weinberg, uh, he was speaking at London Tech Conference, and he's saying, Well, actually, I think you know the young lawyers have got a great future because they'll be able to do more interesting work. And he says the thing that's really important is having the domain expertise. And that's something I I took away. And I mean
- 36:17
- Unidentified speaker (UNKNOWN): Well, in the
- 36:17
- Unidentified speaker (speaker_1): words of in the words of Andy Mandy Rice Davis, he would say that, wouldn't he?
- 36:21
- Unidentified speaker (speaker_2): Yeah. Well uh
- 36:24
- Unidentified speaker (speaker_2): he would say that, but it I means mean i there was a certain log. logic to that that I uh I could come up
- 36:30
- Unidentified speaker (UNKNOWN): um
- 36:31
- Unidentified speaker (speaker_2): There's
- 36:31
- Unidentified speaker (speaker_1): a sort of yeah, I mean there's
- 36:32
- Unidentified speaker (UNKNOWN): a sort of an
- 36:33
- Unidentified speaker (speaker_1): ontological accelerationist argument here, which is to say, look, if I
- 36:38
- Unidentified speaker (speaker_1): define
- 36:40
- Unidentified speaker (UNKNOWN): the
- 36:40
- Unidentified speaker (speaker_1): future of AI as something that's going to be able to grow to the point that it can do anything,
- 36:46
- Unidentified speaker (UNKNOWN): then
- 36:46
- Unidentified speaker (speaker_1): it's going to be able to do anything.
- 36:49
- Unidentified speaker (UNKNOWN): And
- 36:49
- Unidentified speaker (speaker_1): it gets kind of circular. Um I mean there's a thing I always like, obviously, you've got a fairly intellectual audience. So there's a there's a um there's there's an argument made by Anselm, he's a medieval theologian, who said, um, step one, definition of God is that God is the greatest thing imaginable. Because if you could imagine something greater, then that would be
- 37:09
- Unidentified speaker (UNKNOWN): God.
- 37:09
- Unidentified speaker (speaker_1): Okay, that feels reasonable. Like step two, something that exists is greater than something. that doesn't exist. Because if there was an X that existed, that would be a greater thing than an X that didn't exist. Also seems hard to argue with that. Step three, therefore God exists.
- 37:26
- Unidentified speaker (speaker_1): And every other theologian in the room says, but that obviously doesn't work. And Ansom says, okay, prove it. And it takes like eight or nine hundred years of philosophers arguing about this to explain, well, why is that? not true? Why does that not work? Um
- 37:42
- Unidentified speaker (speaker_1): Bert and Russell said it's actually much more interesting to work out why it doesn't work than the the actual proposition itself. It's like it's obviously not true, but try explaining why that doesn't work. No,
- 37:51
- Unidentified speaker (speaker_2): sure.
- 37:51
- Unidentified speaker (speaker_1): And you get this talking about AGI, which is like, well, clearly these models keep getting better. And so if they keep getting better indefinitely, then they're going to get to the point that they're massively cleverer than us. And if they do, then clearly they would have all the domain knowledge you're talking about and they would know what lawyers are and they'd be able to just do what the lawyer does.
- 38:08
- Unidentified speaker (UNKNOWN): It's like,
- 38:09
- Unidentified speaker (speaker_1): well, yes, but there's two giant ifs in that statement. Like there's you're just kind of presuming you're defining the problem away. You're defining the future of AI as something that can do everything, and then saying, well, therefore it can do everything. And maybe, which is the the point I said earlier, we don't have a good knowledge of what the physical limits. of this are. And so if you just presume there
- 38:29
- Unidentified speaker (speaker_1): are no physical limits and presume that it can do anything, then it's going to be able to do everything, fine. Otherwise, um let's kind of plant this in reality and say,
- 38:40
- Unidentified speaker (speaker_1): yeah, you're going to need to build an actual piece of software for this, not just give everybody flawed. And who's going to do that? And who's going to build 500 or 1,000 a thousand individual pieces of software with support and integrations? and compliance and auditing and
- 38:54
- Unidentified speaker (UNKNOWN): go to
- 38:54
- Unidentified speaker (speaker_1): market and everything else?
- 38:56
- Unidentified speaker (speaker_3): But this is kind of fundamental to how
- 38:59
- Unidentified speaker (speaker_2): to think about the the future for anthropic and and indeed OpenAI. And if you believe that Harvey and Ligora, they've got a big lead, so it'd be very difficult and pointless in a way for Anthropic or OpenAI to try. and devote resources to try and catch up with them. If you look at sort of my
- 39:19
- Unidentified speaker (speaker_2): area, finance, well, there's you know, models everywhere, software applications everywhere in tiny little corners. People, you know, you you wouldn't even have thought of. So it seems it seems to me it's almost impossible for the model to to the model company to catch. capture all
- 39:39
- Unidentified speaker (speaker_2): these verticals because it does require a certain amount of knowledge at the start. See and once
- 39:44
- Unidentified speaker (UNKNOWN): you've
- 39:45
- Unidentified speaker (speaker_2): once you've started, once you've built a product and you've got customers, and the customers are using your product, it's very difficult to make them shift. So this
- 39:52
- Unidentified speaker (UNKNOWN): seems to
- 39:53
- Unidentified speaker (speaker_2): me the the the case for
- 39:57
- Unidentified speaker (speaker_2): evaluation.
- 39:58
- Unidentified speaker (UNKNOWN): Yeah,
- 39:58
- Unidentified speaker (speaker_3): I would come at this from a slightly different angle, which is on the one hand, yes, in principle, you should be able to ask the model like work out 200 things that a financial analyst would need and then make them.
- 40:10
- Unidentified speaker (speaker_3): Maybe. What's more interesting as a CAP is
- 40:17
- Unidentified speaker (speaker_3): where does this actually have that information? Where is the stuff that it needs publicly accessible? Where is it available? Where is that implementation? implicit knowledge? Where is it that you want this done the way most people would probably do it? Because what these systems do is they give you the answer that most people will probably do. They do that the
- 40:37
- Unidentified speaker (speaker_3): way most people would do that. And often that's what you want. Sometimes that's not at all what you want. Um what are the kind of conceptual limits on what this can and can't do?
- 40:50
- Unidentified speaker (UNKNOWN): Um
- 40:52
- Unidentified speaker (speaker_3): Can you how much do you need to be in the room? How much is this not something that you can automate? How much is the whole point that you don't want it done the same way? Is can you automate it is is is is a is a kind of a tricky statement. Because can you automate it is actually asking two different questions. One of them is is it physically possible for a machine to do
- 41:12
- Unidentified speaker (speaker_3): this thing? And the other is do you want it done in the same way every time? Those are two very different kinds of questions.
- 41:19
- Unidentified speaker (speaker_0): Yeah.
- 41:19
- Unidentified speaker (UNKNOWN): Okay.
- 41:20
- Unidentified speaker (speaker_3): And so if you say, can I automate making the NDA, one answer is
- 41:25
- Unidentified speaker (UNKNOWN): d
- 41:25
- Unidentified speaker (speaker_3): is the ANDA any good? The other answer is do I want the NDA done the way most people would probably do the NDA? And the answer is sometimes yes. Like sometimes you just want it done the way most people would do it. Sometimes no. Sometimes that's not at all what you want. And then you kind of spiral out, the conversation becomes fragile. Like,
- 41:42
- Unidentified speaker (speaker_1): is this is the data available? Who has the data? Can And Twappy get the data? What does it take to plug that in?
- 41:48
- Unidentified speaker (speaker_3): Who is going
- 41:48
- Unidentified speaker (UNKNOWN): to
- 41:48
- Unidentified speaker (speaker_1): be able to deploy this? What does it actually mean? You know, go back to my you know my example of looking at across every system in the company.
- 41:56
- Unidentified speaker (speaker_3): Who does
- 41:56
- Unidentified speaker (UNKNOWN): that?
- 41:57
- Unidentified speaker (speaker_3): Who is it that actually makes that happen? Because it's not the CFO.
- 42:02
- Unidentified speaker (UNKNOWN): Is
- 42:02
- Unidentified speaker (speaker_3): it the CIO hiring Accenture? Who is it that works out that that's the thing that you would want to do? Is it the is it the CIO? Is it Accenture? Is it Bain? Is it a software company? Who is it that invents all of those things and
- 42:15
- Unidentified speaker (speaker_1): imagines all of those
- 42:15
- Unidentified speaker (speaker_3): things
- 42:16
- Unidentified speaker (speaker_1): and then actually go and goes and gets them done? It's not Claude running on the CFO's laptop. That's one thing you can say with certainty. That's not how any of that happens. So
- 42:25
- Unidentified speaker (UNKNOWN): how is
- 42:25
- Unidentified speaker (speaker_3): it that that
- 42:25
- Unidentified speaker (speaker_1): happens and who built it? builds that?
- 42:27
- Unidentified speaker (UNKNOWN): Um,
- 42:28
- Unidentified speaker (speaker_1): who is it that comes to the company and says, Hey, did you realize that this thing could do this for you?
- 42:32
- Unidentified speaker (UNKNOWN): And
- 42:32
- Unidentified speaker (speaker_3): again,
- 42:33
- Unidentified speaker (speaker_1): you can propose, well, you know, you just have like the giant world brain and you say, Yes, I'm going to give you access to all of my stuff. And you hire Accenture to plug it into all of your stuff. And then you say, Great, now what can you do? And it comes back and like then you get into like a C now, the that Hitchhiker's Hart is Guide to the Galaxy or something. Um
- 42:49
- Unidentified speaker (speaker_3): But like in the meantime, this is a world of professional services.
- 42:52
- Unidentified speaker (speaker_2): So Aswath de Moderan, who I always enjoyed reading his pieces, um did the evaluation of anthropic which came up with one and a half trillion.
- 43:02
- Unidentified speaker (UNKNOWN): And
- 43:02
- Unidentified speaker (speaker_2): he was assuming that it would come up with a big drugs company.
- 43:05
- Unidentified speaker (UNKNOWN): And
- 43:06
- Unidentified speaker (speaker_2): I'm looking at that and thinking, well, I I absolutely accept that AI could be brilliant for drug discovery.
- 43:13
- Unidentified speaker (UNKNOWN): And
- 43:14
- Unidentified speaker (speaker_2): who's more likely to be successful? Is it Anthropic, who's really good at building a model, or is it a drugs company that's quite good at developing drugs? And I I don't know the answer to that. Do you know the answer to that?
- 43:25
- Unidentified speaker (speaker_1): Back in 2014, I know the name because back in 2014 he wrote this very um viral piece saying that Uber was massively overvalued at the time.
- 43:35
- Unidentified speaker (speaker_2): Yeah.
- 43:35
- Unidentified speaker (speaker_1): And his argument for why Iber Uber was massively overvalued. was that he said, well, the TAM is taxis. And what's the market size for taxis? And what share of that will they get? And first of all, the TAM wasn't the existing taxi market. It was completely transforming the taxi market. And secondly, then they did
- 43:55
- Unidentified speaker (speaker_1): delivery and ads. So his estimate was 2024 TAM of $183 billion. and Uber would get 10% of that. So $18 billion.
- 44:07
- Unidentified speaker (speaker_1): Actual gross mobility in 2024 was $83 billion, not 18.
- 44:12
- Unidentified speaker (speaker_3): Delivery was another 75, and then trade was five
- 44:15
- Unidentified speaker (speaker_1): on top of
- 44:16
- Unidentified speaker (UNKNOWN): that.
- 44:18
- Unidentified speaker (speaker_3): So I'm
- 44:20
- Unidentified speaker (speaker_2): not saying he's come up with the right answer.
- 44:22
- Unidentified speaker (UNKNOWN): And I'm not like
- 44:22
- Unidentified speaker (speaker_3): and I'm not doing this as like a gotcha of like, therefore, I'm not gonna believe going to what he's
- 44:27
- Unidentified speaker (speaker_1): saying now.
- 44:28
- Unidentified speaker (speaker_3): The reason that I'm because like, you know, predictions are hard, especially about the future, as Yogi Berra tells us. The reason
- 44:34
- Unidentified speaker (speaker_1): I point
- 44:35
- Unidentified speaker (speaker_3): to it is like
- 44:36
- Unidentified speaker (speaker_1): you can't value things like that.
- 44:40
- Unidentified speaker (speaker_1): You don't know what the new things are going to be.
- 44:44
- Unidentified speaker (speaker_2): Sure.
- 44:45
- Unidentified speaker (speaker_1): You can do a very, very high level exercise and you can say, well, what if productivity? growth, and GDP did this? Then what would be the aggregate value of that? That doesn't capture consumer surplus. It doesn't capture what happens if every law firm buys this and they have exactly the same number of associates just doing 10x more stuff
- 45:05
- Unidentified speaker (speaker_1): in the same hours.
- 45:08
- Unidentified speaker (speaker_1): It doesn't capture why is it the number of accountants went up every decade in the 20th century, even as we automated all accounting work.
- 45:15
- Unidentified speaker (speaker_1): Why are there more people working at PWC today than there were before spreadsheets? Where is that in your TAM?
- 45:21
- Unidentified speaker (speaker_3): Like what what what happened with that? Where is all the value that PW is all the value that PwC got out of spreadsheets is not in the PWC
- 45:30
- Unidentified speaker (speaker_1): PL, nor is it in the Microsoft PL. Nor is it in productivity
- 45:35
- Unidentified speaker (speaker_3): statistics.
- 45:36
- Unidentified speaker (speaker_1): So, you know,
- 45:37
- Unidentified speaker (UNKNOWN): sitting
- 45:38
- Unidentified speaker (speaker_1): in the and saying, well, you know, imagine if they can create a a drug and the drug would be worth X and that, like, I
- 45:45
- Unidentified speaker (speaker_1): Yeah, you can make that spreadsheet. But I don't know what that tells us. I mean, it's back to what I said at the beginning. I don't know what predictive value there is in the S1.
- 45:54
- Unidentified speaker (speaker_2): So
- 45:55
- Unidentified speaker (UNKNOWN): I
- 45:55
- Unidentified speaker (speaker_2): mean have you have you got sort of a range of of ideas as to where anthropic and open air AI will be, say, in five years' time? What what
- 46:05
- Unidentified speaker (speaker_3): will
- 46:05
- Unidentified speaker (UNKNOWN): they
- 46:07
- Unidentified speaker (UNKNOWN): what
- 46:07
- Unidentified speaker (speaker_2): will they be?
- 46:08
- Unidentified speaker (speaker_1): There's a view that says that OpenAI is is the new Netscape.
- 46:12
- Unidentified speaker (speaker_1): First, but
- 46:13
- Unidentified speaker (UNKNOWN): then
- 46:14
- Unidentified speaker (speaker_1): the value went got captured elsewhere. I mean the challenge is
- 46:19
- Unidentified speaker (UNKNOWN): you know,
- 46:20
- Unidentified speaker (speaker_1): maybe this is why I was a bad equity analyst. I hate forecasts. Um
- 46:25
- Unidentified speaker (speaker_1): that
- 46:26
- Unidentified speaker (UNKNOWN): you know
- 46:26
- Unidentified speaker (speaker_1): imagine we're back in 1997 and Larry and Serge are in grad school. Mark Zuckerberg is in junior high or middle school. Jeff Bezos has a terrible haircut and a baggy sweater and a goofy smile, and he sells books.
- 46:43
- Unidentified speaker (speaker_1): You knew the internet was going to be this huge thing. You didn't really know where real oil value was going to be captured, and you were excited about Yahoo. and AOL and ISPs. Like the biggest internet stocks in the dot-com bubble in Europe were ISPs.
- 46:56
- Unidentified speaker (speaker_2): Yep.
- 46:59
- Unidentified speaker (speaker_1): So you don't know what the things are going to be that people do with this and where that's going to be captured and how. And all of the stuff that we ended up doing sort of was on the Yahoo homepage in, I don't know, 1999, ninety nine, 2000, two thousand, 2001. two thousand one, But it ended up not being Yahoo.
- 47:15
- Unidentified speaker (speaker_1): And it was all on like the Y AOL homepage and the Yahoo homepage. And that but and that all happened, but not then.
- 47:22
- Unidentified speaker (speaker_1): So
- 47:24
- Unidentified speaker (speaker_1): is there some mechanic that says you got here you know
- 47:28
- Unidentified speaker (speaker_1): Well everything we've just been kind of talking about are all the reasons why it's going to be really tough for the stuff that gets built with this all to accrue to anthropic and OpenAI.
- 47:40
- Unidentified speaker (speaker_1): Now, you can propose, no, but actually the models themselves will be able to invent all of that stuff. And so and Claude itself will be able to make those things in a way that SAP was not.
- 47:54
- Unidentified speaker (speaker_1): I mean, those 500 apps, SAS apps, 4500 SAS apps are the typical big company in America has. They're all basically something you could have done in SAP or Salesforce, or something. you could have done in Excel at email. And they unbundled that they unbundled that workflow out of email or out of Excel or out of SAP and turned it into a company and went and sold it to people. They're all basically Thin SQL wrappers.
- 48:14
- Unidentified speaker (speaker_1): And theoretically, um those 500 new things that will get created with AI, know no, actually Claude can think of all of those things and build them. So that is different. So then you say yes theoretically, but actually just thinking of that idea and then making the software isn't the
- 48:34
- Unidentified speaker (speaker_1): hard part. It's everything else around having that software existing that's hard. And if you propose this as like this sort of Dut computer that can do anything, then fine, but otherwise someone's going to have to go out and build those things. And that means that you can. keep coming back to saying, isn't this infrastructure?
- 48:55
- Unidentified speaker (speaker_1): And
- 48:56
- Unidentified speaker (speaker_1): I have a comparison I've made is, you know, I made the comparison with mobile operators, which is all the stuff they thought was going to happen did happen, but they didn't get any of the value from it. You could also look at cloud, where cloud was infrastructure, Structure, that was the idea. They Amazon didn't expect to build all the SaaS SAS apps, and they make good money from that, but they don't control the whole thing. They don't capture the value of the whole
- 49:16
- Unidentified speaker (speaker_1): software industry. It could also look like chips where um
- 49:21
- Unidentified speaker (speaker_1): there's no winner takes all effect per se, or there's certainly no network effect in chips. But over the last couple of decades, it got with each generation it got more expensive and more difficult. And I think the Rocks law says that the cost of a cutting-edge semiconductor fabrication plant doubles every four years.
- 49:36
- Unidentified speaker (UNKNOWN): I
- 49:36
- Unidentified speaker (speaker_1): think that's a number. Crucially it's it's it's slower than Moore's Law, otherwise, we'd be in trouble. Um but um so what happened is you went from dozens of fabric semic of companies at the cutting edge to just one in T SMC. And so um it may be that that's what models look like. That it turns out that with each generation they get harder and more expensive,
- 49:56
- Unidentified speaker (speaker_1): and then you only have one or two companies doing it. Um and they make good markets, then they can charge whatever price they want. Um, or they can charge it up. at ROI. Um
- 50:05
- Unidentified speaker (UNKNOWN): may
- 50:05
- Unidentified speaker (speaker_1): look like mobile operators, which are commodity infrastructure, may look like something else. Um but as I've sort of said a couple of times, we can't answer any of those questions by reading the S1.
- 50:18
- Unidentified speaker (UNKNOWN): This
- 50:18
- Unidentified speaker (speaker_0): episode is brought to you by Oxford Data Plan. Every quarter, by the time a company reports, the numbers are old news to anyone who's paying attention. Oxford Data Pan tracked more than 800 KPIs across over 260 listed companies, built from a variety of exclusive and proprietary data sources like receipt data, card
- 50:38
- Unidentified speaker (speaker_0): transactions, and an advertising agency panel. Key revenue drivers like Ad Spend, Order Cans, AOV, and more. Updated every day, 12 hours after the close. It won't tell you what a company's worth, but it will tell you whether the quarter is landing ahead of your model or behind it, before the print, while there's still time to act. I've been using it myself
- 50:58
- Unidentified speaker (speaker_0): and I've been impressed with the depth and breadth of the data. It's been built by people who used to run money for people who still do. Find out more at BITspotly forward slash BTBS for behind the balance sheet, ODP for Oxford Data Plan, and the number one. That's bitly for. forward slash BTBS ODP1.
- 51:18
- Unidentified speaker (speaker_0): Baseline is an equity research tool and screening system which scores 3,600 companies on over 30 qualitative fields. Pricing power, barriers to switching, management quality, AI exposure and more. It allows you to screen on qualitative traits, low pricing power and high AI risk, founder led names with accounting red flags.
- 51:39
- Unidentified speaker (speaker_0): You can't do this on conventional screens, and each score has a detailed explanation of the reasoning behind it. If you ask, does ASML have pricing power? Baseline says 4 out of 5 because of X, Y, and Z. This saves a lot of analyst time when reviewing a new company or an IPO. Private investors
- 51:59
- Unidentified speaker (speaker_0): pay a low monthly cost, funds get additional features, and the ability to customize the system to suit their specific needs, including additional coverage, fields, and other tools. To learn more about baseline, visit B I T spotly forward slash BTBS for behind the balance sheet, B-A-S-E, and the number
- 52:19
- Unidentified speaker (speaker_0): one. So that's BTBS base one.
- 52:22
- Unidentified speaker (UNKNOWN): So
- 52:23
- Unidentified speaker (speaker_2): what signal should we be watching for that will give us some clues as to which path we're going that?
- 52:31
- Unidentified speaker (UNKNOWN): So,
- 52:32
- Unidentified speaker (speaker_3): I mean, I think this is the
- 52:33
- Unidentified speaker (speaker_1): thing I wrote in my newsletter last night is um
- 52:37
- Unidentified speaker (UNKNOWN): we
- 52:38
- Unidentified speaker (speaker_3): have a lot of volatility
- 52:40
- Unidentified speaker (speaker_1): because all of the numbers are very elevated and all the ratios are very elevated, which makes people nervous. And so people jump on anything that looks like signal.
- 52:47
- Unidentified speaker (speaker_1): Um so um and
- 52:50
- Unidentified speaker (UNKNOWN): when
- 52:50
- Unidentified speaker (speaker_1): you really would know you're really in a a bubble or a correction when stuff gets priced in twice. Like, well, we didn't we had that story a week ago. Why are you pricing it again? Because we are, like, deal with it. Um But you know, obviously we saw that with wobbles over Google search revenue. Like maybe this is the moment that everything flips.
- 53:10
- Unidentified speaker (speaker_1): You see it in in e every one of those stories suddenly. gets leverage applied to it, or rather people think maybe this is the moment. Um and the problem is on the way up, you know, it goes tick, tick, tick, tick, tick, and on the way down it goes tick, tick, tick, tick, tick. There's like I don't know what's the equivalent of faller catching a falling knife on the way
- 53:30
- Unidentified speaker (UNKNOWN): up?
- 53:31
- Unidentified speaker (speaker_1): There should be a phrase that means the same
- 53:33
- Unidentified speaker (UNKNOWN): point on the way up.
- 53:34
- Unidentified speaker (speaker_1): Yeah, but you know it's the problem. Do you want to do study new at what point do you want to do you want to step off? I don't there's not like some clear answer to that question. Um
- 53:44
- Unidentified speaker (speaker_1): clearly we have explosive product market fit for software development right
- 53:49
- Unidentified speaker (UNKNOWN): now.
- 53:49
- Unidentified speaker (speaker_1): And then you have cycles, you know, every everyone's saying, wait a minute, but we're spending way too much on this stuff, so we need to moderate our spending a bit. But that doesn't change the underlying story. We don't have consumer market product market fit at scale for something with real intensity. We have lots of people using this a little bit. We have a billion, billion and a half people who can't who are using this but can't think of anything to do with it today.
- 54:09
- Unidentified speaker (speaker_1): Maybe that's agents. Um
- 54:13
- Unidentified speaker (speaker_1): Mark Zuckerberg on his quarterly call said that Muse has product market fit. Well, it's got 10 million users. Um he says 25% twenty-five percent of them are using them on any given day, which note is not daily active users. It's like slightly different. And so when people don't give you a number or give you a number that looks like a number but isn't, that's always always signal. Um
- 54:33
- Unidentified speaker (speaker_1): he also said you can't scale it until costs have gone down by 10x. Which will only take about a year at current rates. Um so that's not quite product market fit if you can't actually scale it. Um
- 54:45
- Unidentified speaker (UNKNOWN): if
- 54:46
- Unidentified speaker (speaker_1): we had a thing that looked like Instagram, we had a thing that where consumers were like, No, I really, really want this all the time.
- 54:51
- Unidentified speaker (speaker_1): then that would be another shift. Maybe that will happen next later this year. Maybe that will happen in the next 12 months. Right.
- 54:59
- Unidentified speaker (speaker_1): Well, it is, but it is and it isn't, because it is actually yes, great, but it's actually 10 million people.
- 55:06
- Unidentified speaker (UNKNOWN): If
- 55:06
- Unidentified speaker (speaker_1): 10 million after a month, okay. Give me some charts. Show me this is really, really working and not it's just working for everyone who gets
- 55:13
- Unidentified speaker (UNKNOWN): to
- 55:13
- Unidentified speaker (speaker_1): our Walkeshi's case's podcast.
- 55:15
- Unidentified speaker (speaker_1): Um tell me it's working outside the valley. Um tell me it's working at scale and sustainably and really working. Um maybe. Like at the moment, could be. We'll
- 55:26
- Unidentified speaker (UNKNOWN): see.
- 55:27
- Unidentified speaker (speaker_1): Um certainly it's working in a way that open claw didn't work. Maybe. Um
- 55:34
- Unidentified speaker (UNKNOWN): I
- 55:35
- Unidentified speaker (speaker_1): mean all of this is to say like, you know, like the problem is it's like the internet. Yes, clearly this is working. Clearly, this is going to be a huge deal. Clearly, this is going to change everything.
- 55:43
- Unidentified speaker (UNKNOWN): But
- 55:43
- Unidentified speaker (speaker_1): where's the value capture going to
- 55:44
- Unidentified speaker (UNKNOWN): be?
- 55:45
- Unidentified speaker (speaker_1): Who is it going to
- 55:46
- Unidentified speaker (UNKNOWN): be?
- 55:47
- Unidentified speaker (speaker_1): And
- 55:49
- Unidentified speaker (speaker_1): we don't know what the usage patterns are going to look like, what the apps are, what people are going to do with it. Just you know, on Muse. Is the right part of the point of Muse is that part of the consumer problem with AI? is a blank screen problem. You don't really know what you can ask or when to ask it or how and how to form a pattern around that. Maybe
- 56:09
- Unidentified speaker (speaker_1): maybe agents solve that because they expand what you can ask, because you can ask it to do a thing for you rather than just answering a question. That's good, but it's still a blank screen problem. Is the solution that you make it passive so it watches everything, and then it can pop up and say, hey, I've noticed your bank account.
- 56:27
- Unidentified speaker (speaker_1): you should move on to this account, you should be sweeping your money and you know, you should be doing X and Y. I've noticed you're on the wrong credit card. Is it that it can watch you and learn and see? And so you can pull the use cases through like that.
- 56:41
- Unidentified speaker (UNKNOWN): If
- 56:41
- Unidentified speaker (speaker_1): that's your thesis, that's Meta can has some of that, Google has some of that, Amazon has a bit of that, Apple has a bit of that, but doesn't kind of doesn't want to use it. OpenAI and Anthropic don't have any of that.
- 56:53
- Unidentified speaker (UNKNOWN): So if
- 56:53
- Unidentified speaker (speaker_1): your thesis is consumer agents that are watching and seeing all of your behavior will drive a new wave of adoption. Well, OpenAI doesn't see your email, it doesn't see your messaging or your purchasing or your social media, it doesn't see anything except what you do in ChatGPT. So, which is one reason maybe why they want a device.
- 57:13
- Unidentified speaker (UNKNOWN): So
- 57:13
- Unidentified speaker (speaker_1): if that's your thesis, that's a thesis. for Google and Meta. It's not a thesis for open air and anthropic, or it's a thesis where they come at a disadvantage. Um again, we're back in nineteen ninety seven. Like, is it the ISPs or is it going to be Yahoo or is it Netscape?
- 57:27
- Unidentified speaker (speaker_2): These are unanswerable questions. Have you been surprised that um Google search has held up as well as it has? And I I mean obviously last week there was a bit of
- 57:37
- Unidentified speaker (UNKNOWN): a
- 57:37
- Unidentified speaker (speaker_2): wobble, but I
- 57:38
- Unidentified speaker (UNKNOWN): mean
- 57:38
- Unidentified speaker (speaker_2): the growth the rate. it might not met Anna's expectations, still a pretty good growth rate if you'd thought back to the chat two years ago that said nobody was going to be searching. It was all going to be done via Chat GPT.
- 57:50
- Unidentified speaker (speaker_1): I think
- 57:52
- Unidentified speaker (speaker_1): so. Some of this is habit. Some of this is bubble thinking.
- 57:57
- Unidentified speaker (UNKNOWN): Like,
- 57:57
- Unidentified speaker (speaker_1): oh I'm not even doing this anymore. Everyone must be doing this. Some of it is um
- 58:03
- Unidentified speaker (speaker_1): what are you trying to achieve? If if you're only doing, you know, using you don't change the world by doing the old thing with the new technology.
- 58:13
- Unidentified speaker (speaker_1): Using Chat GPT as a search engine is not very interesting. And it's not really any better. And Google can do that too. And they can probably do it a better.
- 58:22
- Unidentified speaker (speaker_2): Yeah.
- 58:22
- Unidentified speaker (speaker_1): What you have to do is get to not
- 58:28
- Unidentified speaker (speaker_1): um what's a good credit card, but look at my spending and my behavior and work out what credit card I should
- 58:37
- Unidentified speaker (UNKNOWN): have.
- 58:38
- Unidentified speaker (speaker_1): It's not here's a coat, where can I buy it? It's look at my Instagram and suggest a new winter coat for me that would change my look, but not too much. It's actually doing something. different. It's not doing the old thing, but slightly more or slightly better with the new thing. And
- 58:58
- Unidentified speaker (speaker_1): um, maybe that's agents, maybe that's dots, maybe that's Muse, maybe it's something else. Is that properly done horizontally or vertically? Is it properly done within a finance app? Or is it properly done with an application that looks everything and has my bank, my bank account and my
- 59:17
- Unidentified speaker (speaker_1): purchasing, and my travel and my camera roll and everything else? Who is it that's and who is it that's best placed to work out those use cases, persuade the user to do those use cases, surface and suggest that? Who has the visibility on you? Um
- 59:36
- Unidentified speaker (UNKNOWN): You know,
- 59:36
- Unidentified speaker (speaker_1): I used to describe this as like the blind man, the problem is. the blind man feeling the elephant. Yeah. Google and Meta and Amazon each feel different bits of the elephant and think it's a different thing. Apple theoretically sees a whole other bunch of it, doesn't do anything.
- 59:50
- Unidentified speaker (speaker_1): OpenAI and Antropic aren't either of those.
- 59:54
- Unidentified speaker (speaker_2): And um do you have any sort of strong feelings about the the sort of relative winners between the existing big team? tech companies. I mean, do you feel that you know the advantages that Facebook is getting or Meta is getting in in advertising from using the AI? Is that kind of like
- 1:00:14
- Unidentified speaker (speaker_2): a sustainable competitive advantage?
- 1:00:17
- Unidentified speaker (UNKNOWN): Well,
- 1:00:17
- Unidentified speaker (speaker_1): so clearly they you know they've all got like double digit percentage points gains in their ad revenue simply from optimizing. Partly optimizing is more than optimizing. It is going from X, therefore, Y correlation. You bought X, other people bought who bought X, bought Y, therefore, you might like Y, to knowing what those products are
- 1:00:36
- Unidentified speaker (UNKNOWN): and
- 1:00:36
- Unidentified speaker (speaker_1): the reasons why the people make those purchases.
- 1:00:45
- Unidentified speaker (speaker_1): For Amazon on that.
- 1:00:48
- Unidentified speaker (speaker_1): Then there is a question of what do they have and what are they? Meta is a company that has always served user behavior.
- 1:00:58
- Unidentified speaker (speaker_1): What it does is it looks and sees and follows and builds and tracks and pivots. It's protein. It's continually changing based on what people are and what they want. And so that's the bet for Meta for AI that it's a bet on a company that always works it out and always follows people. and works it out. For Google, like you know, if this is
- 1:01:18
- Unidentified speaker (speaker_1): search, this is search. Google does search. Um Apple is sort of interesting here in the sense that, like
- 1:01:29
- Unidentified speaker (speaker_1): whatever happens here, I'm going to want a device that I have with me that does this. Who is good at making beautiful devices with a strong privacy story, great battery life, a great camera, and really strong. custom silicon?
- 1:01:41
- Unidentified speaker (speaker_1): Sounds like Apple to
- 1:01:42
- Unidentified speaker (UNKNOWN): me.
- 1:01:45
- Unidentified speaker (speaker_1): So they're sort of well positioned by not playing or playing, choosing very carefully which bits they play in.
- 1:01:54
- Unidentified speaker (speaker_1): Amazon, you know, the core of Amazon is a network effect around logistics and delivery.
- 1:02:01
- Unidentified speaker (speaker_1): The other side of this, of course, is the other side. of the agent story, which is why Amazon decided to block Muse. There's two parts of that. One of them is a purely tactical operational question, which is why all the banks are blocking it, which is like, where's the liability and the security? Like we spend 20 years blocking bots because they're not the user. We don't know who that is. Okay, well, now we're going to let like any random connection from Muse connect. Well, we
- 1:02:21
- Unidentified speaker (speaker_1): don't know if that's the a user.
- 1:02:22
- Unidentified speaker (UNKNOWN): We
- 1:02:22
- Unidentified speaker (speaker_1): don't know if that's you. Maybe there's someone who stole your someone stole your password. They make a transaction where's the liability? Where's the chargeback? So there's a bunch of like operational questions around it. Then there's a strategic question, which is nobody wants to be somebody else's dumb API call.
- 1:02:36
- Unidentified speaker (UNKNOWN): True.
- 1:02:36
- Unidentified speaker (speaker_1): No, Amazon does not want a situation where I say to me, Amazon did $75 billion of ad revenue in the last 12 months. They don't want a situation in which you ask somebody else's agent to make you a purchase and it goes on Amazon and makes that purchase for you.
- 1:02:52
- Unidentified speaker (speaker_1): They want to show you that. They want the opportunity, the customer ownership, the upsell, everything else. And of course, this trade-off is as old as organized retail. Like, do I want my own shop or do I want to be in the department store? Well, it depends.
- 1:03:03
- Unidentified speaker (UNKNOWN): You know,
- 1:03:03
- Unidentified speaker (speaker_1): do I want the customer ownership? And am I willing to spend to build the beautiful shop on Bond Street or, you know, Ferber, Furbourg Saint Santo John and Array? Um,
- 1:03:12
- Unidentified speaker (UNKNOWN): Or
- 1:03:12
- Unidentified speaker (speaker_1): Bulevard Sandwich, you
- 1:03:13
- Unidentified speaker (UNKNOWN): know,
- 1:03:13
- Unidentified speaker (speaker_1): or Bulevard Bulevalt Saint. Santon Array, or do I want to be in the supermarket? And it depends, like what brand are you, what's the product, what's a customer. Um, Amazon is Walmart. Amazon wants to own that customer relationship. They don't want to be somebody else's dumb API call. And all of those old questions get repeated over and over again. I have a slide about this that says that all AI questions have one of two answers.
- 1:03:33
- Unidentified speaker (speaker_1): The answer is either no one knows, and we've had a lot of no one knows, or
- 1:03:37
- Unidentified speaker (UNKNOWN): how did
- 1:03:38
- Unidentified speaker (speaker_1): it work last time? Right. Are we going to give this startup complete access to all of our customers and our business without anything in return?
- 1:03:47
- Unidentified speaker (speaker_1): Hmm. How did that work the last time? How did we answer that question last time? Should we build our own software or should we pay Accenture to build it for us or should we buy it from a vendor? How do you do cloud? Depends.
- 1:04:02
- Unidentified speaker (UNKNOWN): So
- 1:04:02
- Unidentified speaker (speaker_2): Luke, I'm gonna finish now. I I mean
- 1:04:06
- Unidentified speaker (speaker_2): my sort of thinking about this is you you know, in the one hand, you've got the sort of bull case, you've got a magical product with sort of almost unlimited demand. Efficiency is improving incredibly fast, and demand's gonna grow to the skies. And the computer just gets cheaper, and so we all use more of it, and and there's a
- 1:04:26
- Unidentified speaker (speaker_2): a huge profit pool. Where it exactly gets divided up, I don't know. And the bare case is that you've got so much money being invested in just the sheer scale of investment requires an impossible level of revenue. And when the market starts to worry about that, this the plug gets pulled, the cost of money is going
- 1:04:46
- Unidentified speaker (speaker_2): up anyway, and because the whole the whole edifice is interconnected there's a a a vulnerability
- 1:04:53
- Unidentified speaker (UNKNOWN): to
- 1:04:53
- Unidentified speaker (speaker_2): it. And I think that this financing issue is really
- 1:04:57
- Unidentified speaker (UNKNOWN): a
- 1:04:57
- Unidentified speaker (speaker_2): critical issue because
- 1:04:59
- Unidentified speaker (UNKNOWN): the
- 1:04:59
- Unidentified speaker (speaker_3): sheer scale
- 1:05:00
- Unidentified speaker (UNKNOWN): of the
- 1:05:01
- Unidentified speaker (speaker_3): money
- 1:05:01
- Unidentified speaker (speaker_2): involved
- 1:05:02
- Unidentified speaker (UNKNOWN): means
- 1:05:02
- Unidentified speaker (speaker_3): that financing,
- 1:05:03
- Unidentified speaker (UNKNOWN): the cost of
- 1:05:04
- Unidentified speaker (speaker_3): the money is a a critical component.
- 1:05:07
- Unidentified speaker (speaker_1): Yeah. So there's I mean the the the interesting thing there is like the spread of who can get capital. And some of that is timing and some of that is who you are. I mean, I I talked about, you know, obviously Google has the cash flow. If you're neo cloud, you don't. So who has the cash flow, who has the access to get to that capital is something that's changed this year. I mean, it's funny,
- 1:05:27
- Unidentified speaker (speaker_1): you've you've brought up implicitly, Tam, a couple of times, and I've and I've sort of kind of not answered it because we've spent most of our time talking about where who would get the money rather than how much money. Um and
- 1:05:37
- Unidentified speaker (UNKNOWN): obviously
- 1:05:37
- Unidentified speaker (speaker_1): how much money is obviously kind of important here.
- 1:05:42
- Unidentified speaker (speaker_1): The puzzle I have here is,
- 1:05:45
- Unidentified speaker (UNKNOWN): and
- 1:05:45
- Unidentified speaker (speaker_1): I've sort of alluded to it in a couple of different ways.
- 1:05:49
- Unidentified speaker (speaker_1): So
- 1:05:51
- Unidentified speaker (speaker_1): supposing I go and ask,
- 1:05:54
- Unidentified speaker (UNKNOWN): well,
- 1:05:54
- Unidentified speaker (speaker_1): so let's start by talking about behavioral economics. How much money is left on the table as a result of behavioral economics? The fact that you don't bother to do that, that it's too hard, it's too much friction, it's too painful, it's too difficult, it's too complex. You don't bother to move that money into the savings account because, hey, it's only £15 a month in
- 1:06:14
- Unidentified speaker (speaker_1): interest income. You are on the credit card that was the highest scored on the comparison site rather than the one that exactly matches your behavior. Yeah. You don't cash in that, you know, $10 a month discount on Spotify. And if you cash in all of those, that would be like a grand a year. in all the bits and pieces that are floating around. How much does all of that add up to? And how would you put that?
- 1:06:34
- Unidentified speaker (speaker_1): If you have agents and AIs optimizing everything, maybe all that money just does this. Maybe it does that, or maybe it goes up and you just rechange the price. Which industries will give all of that up and which industries will rebalance their pricing to hang on to half of it or all of it? Where's that in your TAM calculation for Anthropic? Um,
- 1:06:53
- Unidentified speaker (speaker_1): every law firm buys AI. Okay, set aside the question of whether they're buying it directly from Anthropic or buying it from Harvey. Um, how much do they pay?
- 1:07:04
- Unidentified speaker (speaker_1): What's the pricing power of that? What's the pricing value of that? What's the ROI? Does it replace 20% of your associates or not? Um, you know, if you've been sitting and saying what's the TAM
- 1:07:14
- Unidentified speaker (UNKNOWN): for
- 1:07:14
- Unidentified speaker (speaker_1): spreadsheets in the 1980, what number would you have come up with? On the one hand, you didn't really know what the economic value of it was going to be. And the answer is in the end, you can't charge more than 50 or 100 quid for it. And on the other side, you didn't know how many people were going to do this because you'd have built your you'd have done your calculation and you said, well, how many CPAs are there in the world? And you would not have said And one and a half billion
- 1:07:34
- Unidentified speaker (speaker_1): people will be using Excel.
- 1:07:36
- Unidentified speaker (UNKNOWN): Because
- 1:07:36
- Unidentified speaker (speaker_1): of course, what actually happens is there's a small people of people. number of people who do a lot in Excel, and everyone kind of has it and uses a little bit.
- 1:07:46
- Unidentified speaker (speaker_1): Where's Uber in that? Where's Instacart in that? Um, where are all the things that we actually do with this? How much of that value disappears as consumer surplus? How much how much of it doesn't even show up in productivity statistics? Um how much of it just gets competed away and becomes part of the cost of doing
- 1:08:06
- Unidentified speaker (speaker_1): business? Um
- 1:08:09
- Unidentified speaker (UNKNOWN): I
- 1:08:09
- Unidentified speaker (speaker_1): don't know, I asked this question, I was at an event last week with with Lloyd Blankfind. You know, I said, you know, in the the eighties you deployed PCs and spreadsheets and yet the number of associates didn't go down by ninety percent.
- 1:08:22
- Unidentified speaker (speaker_1): And like uh yeah and easily a third of Goldman's employees now are software engineers.
- 1:08:28
- Unidentified speaker (speaker_1): Where's that in your TAM for PCs that you calculated in 1980? Um and so I think the all half of the answer to this is there'll be a bunch of value that doesn't accrue to anybody. It's just efficiency and maybe lower prices or consumer surplus. There'll be also be a bunch of value that
- 1:08:47
- Unidentified speaker (speaker_1): is new spend or spend directed from different places, but you don't know who's going to capture that.
- 1:08:54
- Unidentified speaker (UNKNOWN): And
- 1:08:55
- Unidentified speaker (speaker_1): I've we've argued at some length that it's probably not the model companies, but hey, you know, we could be completely wrong and it could be all go to the model companies.
- 1:09:01
- Unidentified speaker (UNKNOWN): Sure. Yeah, no,
- 1:09:02
- Unidentified speaker (speaker_0): sure.
- 1:09:03
- Unidentified speaker (speaker_1): And like I can make a slide that says, well, this is the value of the global professional services industry.
- 1:09:10
- Unidentified speaker (speaker_1): This is the value of the global money transfer industry. This is the value of the global software industry is a trillion dollars. Okay. The global professional services industry is again make up a number. The global marketing industry is a trillion dollars. Global retail is $30 trillion, $25 trillion. Which percentage of each of those do you think it's going to
- 1:09:30
- Unidentified speaker (speaker_1): be? And this is what gets you this kind of absurd thing. What it kind of looked absurd where SpaceX said our TAM is like global GDP or whatever it was or half of global GDP. Which sounded kind of stupid, but what they're actually saying is, well, all of that is money that some people are going to do some stuff with with AI and so pay something for.
- 1:09:47
- Unidentified speaker (UNKNOWN): Like what
- 1:09:47
- Unidentified speaker (speaker_1): was the term for the what was the term for the internet? Global GDP.
- 1:09:51
- Unidentified speaker (UNKNOWN): Yeah. But it
- 1:09:52
- Unidentified speaker (speaker_3): didn't
- 1:09:52
- Unidentified speaker (UNKNOWN): mean
- 1:09:53
- Unidentified speaker (speaker_1): that like all of that went to the ISPs or Yahoo.
- 1:09:56
- Unidentified speaker (speaker_2): And of course the difference this time is that there's a physical dimension. So the physical world is is being disim dis intermediated. But we can
- 1:10:03
- Unidentified speaker (UNKNOWN): go
- 1:10:04
- Unidentified speaker (speaker_3): we haven't
- 1:10:05
- Unidentified speaker (speaker_1): talked about robots at all, which is
- 1:10:07
- Unidentified speaker (speaker_1): that part of the term for anthropic? Well yeah kind of, but not really.
- 1:10:12
- Unidentified speaker (speaker_2): Well I mean it's such a a big and fascinating subject and it's great to get your perspective. We probably need to do this in six months, twelve months time and and revisit it because it's we'll
- 1:10:23
- Unidentified speaker (UNKNOWN): see if any of these
- 1:10:23
- Unidentified speaker (speaker_2): questions have
- 1:10:24
- Unidentified speaker (UNKNOWN): changed.
- 1:10:25
- Unidentified speaker (speaker_3): We'll see if
- 1:10:25
- Unidentified speaker (UNKNOWN): some of these they
- 1:10:26
- Unidentified speaker (speaker_3): haven't changed at all.
- 1:10:27
- Unidentified speaker (speaker_2): I bumped into a a a friend of mine, a client who um works for one of the big pod shops. And I said, Oh, I'm I'm I'm interviewing Benedict on Monday. And I said, What what should I ask him? And he said, Well, he said, uh I don't know. He said it changes every week. And that you know, the at the sharp end in the stock market, the issues are literally changing
- 1:10:47
- Unidentified speaker (speaker_2): week by week. But it's so kind of you to come on. Tell um listeners and viewers where they can find you.
- 1:10:55
- Unidentified speaker (speaker_1): Sure. So I do a weekly newsletter that's got 140, 150,000 subscribers, and there's a free version you can sign up for. And then I write essays and I do a presentation twice a year, which I'm actually about to publish my my new presentation, and then people pay me to come and give presentations, give that presentation to them their. investors,
- 1:11:15
- Unidentified speaker (speaker_1): their board, their clients, and so on. So yes, I've doing I'm I'm on airplanes a lot this year because a lot of people want to know what this AI thing is.
- 1:11:23
- Unidentified speaker (UNKNOWN): So yes.
- 1:11:24
- Unidentified speaker (speaker_2): And it's g Google Benedict
- 1:11:26
- Unidentified speaker (speaker_3): Evans. Google Benedict Evans and then
- 1:11:28
- Unidentified speaker (speaker_1): you'll
- 1:11:28
- Unidentified speaker (speaker_3): find me.
- 1:11:29
- Unidentified speaker (speaker_2): So listen, thanks very much. I wish you safe travels and we shall reconvene. Yep.
- 1:11:36
- Unidentified speaker (speaker_3): See you soon. Thanks a lot.
- 1:11:39
- Unidentified speaker (UNKNOWN): I
- 1:11:39
- Unidentified speaker (speaker_0): hope you enjoyed that conversation with Benedict Evans. What struck me most was the distinction between the potential for the enormous economic value AI could create and the much more difficult question of who will capture that. We could continue to see extraordinary growth and demand for AI, with falling Bowling costs and new applications
- 1:11:59
- Unidentified speaker (speaker_0): transforming entire industries. Yet the companies building the models and infrastructure might still struggle to earn attractive returns. I thought Benedict's comments on the comparisons with the early internet and mobile telecoms industries were particularly interesting. And of course, his thoughts on the economics of anthropic, useful lives of
- 1:12:18
- Unidentified speaker (speaker_0): GPUs, and the potential commoditization of AI models were all really insightful. You can find Benedict's work at benans.com benvins.com and I'll put the details in the show notes. If you enjoyed the episode, please subscribe, share with a friend or colleague, and leave us a review. Thanks for listening. I'll see you next time.
- 1:12:45
- Unidentified speaker (speaker_1): For your education and entertainment only. Nothing in this podcast should be construed as investment advice. or relied upon for investment decisions. Always do your own research.
- 1:12:56
- Unidentified speaker (UNKNOWN): Thank
- 1:12:56
- Unidentified speaker (speaker_3): you for listening. I hope you enjoyed this episode.
- 1:12:59
- Unidentified speaker (speaker_2): It would mean a great deal to me if you could please share this with just one friend. And feel free to spread the word by subscribing or even leaving a review on Apple Podcasts or Spotify. Thank you for your support.