Ev | Uncapped with Jack Altman Summary

1. Key Themes

The Anthropic IPO is a narrative-and-momentum event, and the numbers have been tracking the plan all along

Ev walks through the initial range of $70–$78 per share, roughly a $1.3–1.4 trillion valuation, with up to $60 billion raised at the high end. Pricing is set for November 18th with trading starting the 19th. The S-1 amendment disclosed run-rate revenue of $81 billion at the end of October, up $10 billion in a month. Jack's point is that the weekly whipsaws obscure a trend that was on plan all along.

Ev: "So 70 to 78 per share is about 1.3 to 1.4 trillion. And so, my guess is that they end up, um, you know, increasing the range and the final price ends up, you know, at least 1.5 if not 1.6" 00:01:19

Ev: "they amended their, uh, the other big, um, thing that happened today, they amended their S-1 with the fact that run-rate revenue was $81 billion at the end of October, which was up 10 billion from the previous month" 00:02:02

Jack: "if you didn't check the news, it would have just sort of been like what you thought it was going to be." 00:03:10

Ev ties the model cadence to the IPO: "they are going to, you know, almost have model releases in tandem with their IPO timeline." 00:02:45

Hedge funds love the frontier labs, and stocks go up because people want to buy them

The expected buyer base is hedge funds enamored with the labs' scale economics. Ev's hedge fund friend gave the simplest framing of price formation, and Jack extends it: narrative still dominates in public markets, not just in early-stage venture.

Ev: "basically, every single one absolutely loves Anthropic... it has like all these amazing properties of returning uh increasing returns at scale and a ton of operating leverage" 00:05:57

Ev, quoting his friend: "Stocks go up because people want to buy the stock. And ideally, people like me want to buy the stock." 00:05:57

Jack: "there's a misconception, I think, that, you know, when a startup is early-stage, it's narrative, and then as it gets later stage, it's numbers... the narrative is still pretty like surprisingly dominant." 00:07:30

Ev adds that cheap stocks stay cheap without buyers: "SK Hynix or Micron trade at three times earnings. Like, how could that happen?... if people wanted to buy the stock, it would be worth more." 00:07:01

Frontier labs will structurally trade with more volatility and at a discount to SaaS

Ev argues that low API switching costs make lab revenue less sticky than SaaS, which had capacity-modeled sales teams. Jack explains why volatility is rational: the outcome distribution is far wider than a Coca-Cola's.

Ev: "OpenAI, you know, reduced prices for um on the API, and very successfully, for a few months, very successfully took share from Anthropic... it's very clear that, especially in the API business, the switching costs are not high." 00:08:01

Ev: "they'll probably always trade at a discount to what software like pure software businesses would have traded on a like-for-like basis." 00:09:11

Jack: "Maybe it's going to be 5, maybe it's going to be 9, maybe something terrible's going to happen. And so, you have these extreme outcomes that can happen." 00:10:03

The AI infrastructure trade unwound because it was priced on one metric, and that metric got haircut

Ev quantifies the drawdown and the mechanism. Analysts converged on gigawatts of data center CapEx as the barometer. When lab ARR growth disappointed relative to the extrapolation, the gigawatt estimate dropped, and the highest-beta, most levered names fell hardest.

Ev: "the SOX or SOXX, um, Semiconductor ETF, is down over 20% over the last month... stocks like CoreWeave... are down more than 40%." 00:18:23

Ev: "folks were pricing in something like 70+ gigawatts for next year outside of China of new AI data center capacity... Now, uh, it looks more like, um, 40 to 50 gigawatts" 00:21:07

Ev: "the revenues from the AI ecosystem just aren't big enough at this point, and they're not growing fast enough at this point, to justify the amount of CapEx that's being contemplated for the next 2 years." 00:21:07

Ev points to leverage as the amplifier: "if you have a neo-cloud that is maybe financing a lot of their data centers with debt... and the capital markets start seizing up, and your cost of debt starts going up, then everything sort of gets questioned." 00:22:03

A CapEx bust is inevitable, but the bull case rests on a capability unlock like robotics

Jack states flatly that an infinite buildout cannot end smoothly. Both hosts still see a plausible re-acceleration if a research breakthrough lands within 12–24 months.

Jack: "it's predestined that there will come a moment when everybody screeches to a halt, and they say, 'We have just like We got to stop. We've overbuilt.'" 00:27:35

Ev: "I think it's very possible that that with how quickly models are getting better, that there is some major breakthrough in the next 12 to 24 months, as a lot of researchers believe" 00:29:08

Jack on robotics: "one day, there will be a demo where one of the labs will show that they have perfect control of a robot... The companies are going to rip that day." 00:29:52

AI token demand will diffuse at a fast-but-finite pace as buyers rationalize

Jack describes the lifecycle inside customers: exponential spend, then cost rationalization when a task is saturated, then another leap when a new frontier model unlocks the next task. Growth comes from penetration of the rest of the economy plus frontier leaps.

Jack: "within an account, let's say at, I'm just making up, Stripe, you would see a huge exponential curve on AI consumption. And then at some point, you'd see it start to rationalize." 00:24:02

Jack: "that will happen at a fast, by any normal standards, rate, but it won't happen at this infinite exponential speed we were looking at." 00:25:05

The Jensen Huang episode reshapes US-China AI dynamics and strengthens the case for American open source

The hosts treat the alleged-spy scandal as a structural shock to open-source and chip-export advocacy. Ev explains Nvidia's incentive through "commoditize your complement," and the hosts name likely winners.

Ev: "Jensen has been one of the biggest proponents of open-source software" and "he's been sort of the key figure to arm the rebels against the frontier labs" 00:12:38

Ev on the economic logic: "there's a very basic economic principle called commoditize your complement... you're the one that can charge the rents, whereas everyone else is competing on price with each other." 00:12:50

Jack on lost credibility: "You can't break through the meming that would happen anymore." 00:15:02

Winners: Jack says "the frontier labs trade up because of just closed source" 00:15:31 and "non-Nvidia chips companies would do well." 00:15:40 Ev: "American open source has just become immensely critical." 00:16:16

Tech has a sentiment crisis with America, and it is now a geopolitical liability

After the Democratic midterm sweep, the hosts argue the industry has bridged to DC but not to the country. The Gallup data point is the anchor.

Ev: "America hates AI, like big time, and America hates data centers." 00:31:42

Ev on the Gallup poll: "In China, it got 93% said mostly help. And in the US, get 36." 00:38:10

Ev: "when people dislike something in broad swathes, my reaction isn't, 'Oh, they don't understand.' It's, 'Oh, there must be some threat.'" 00:38:42

Jack: "China loves that the public in America hates AI. They love that the public in America are anti-data center... you could not draw up better US... politics if you're China." 00:40:33

Data centers are flipping from a seller's market to a buyer's market for communities

Jack uses the Amazon HQ2 auction as a historical mirror. Municipalities now hold leverage, which is not necessarily bad, since done right it is symbiotic.

Jack: "Amazon in 2017 basically like put their HQ2 up for auction, and you had like over 200 cities and municipalities basically bidding" 00:35:57

Jack: "Now there's this big role reversal where I think all these cities and municipalities now feel like... you can actually bid to us and you can figure out how to make it worth the public's time and public's money." 00:35:57

On Meta's Louisiana Hyperion project: "that project alone doubled the local tax receipts for that county and gave teachers bonuses that were equal to like their salary in many cases." 00:37:06

Venture is absorbing an unprecedented flood of mega-fund capital, squeezing the middle and inflating seed

Ev and Jack tallied roughly $50–60 billion being raised by the largest firms over the coming months, from firms that raised just 12–18 months ago. The frontier labs may have been a one-time subsidy for capital absorption.

Ev: "we believe that something like 50 to 60 billion of venture capital is going to be raised by the sort of largest firms over the next I don't know, 1 to 4 months" 00:41:11

Jack: "the frontier labs were sort of a one-time subsidy on this phenomenon of immense ability to suck in capital and produce great returns. And everywhere else that you can put in a lot of capital is completely unproven" 00:43:14

Ev on seed: "seeds are now permanently expensive because you have this oscillating traffic jam ahead of you." 00:51:27

Public market turmoil propagates backward through private markets like a traffic wave, and it now moves faster

Jack describes how crossover investors balk at pre-IPO prices, which in turn pressures every earlier round. Ev notes the lag has compressed since his Lattice days.

Jack: "it's like a traffic wave... when the public investors get walloped in the public markets and get scared, it begins to bleed backwards into private markets." 00:49:15

Ev: "I remember when I was running Lattice, I would like see this happening, but it would take 6 months... I think it goes faster than that now because everybody is just so hyperconnected." 00:50:45

Ev: "I think now we're kind of in the camp of everybody is on the fence." 00:54:25

Firm alignment (identity, strategy, fund size) matters more than any one "right" organizational design

On Sequoia's restructuring, Jack argues there is no absolute right way to organize a firm. Ev adds a bulk-and-cut lens that applies to YC, companies, and founders.

Jack: "What matters is the alignment. And the alignment of the firm with its fund sizes and the identity that it wants to portray to the world" 00:56:13

Ev: "it's like um like you with your training, you know, you're either bulking or cutting, and I do think that there's these bulking moments and cutting moments, and I think you kind of need both of those." 00:59:28

2. Contrarian Perspectives

Seed is the new "safe haven," which is why it is permanently expensive

The conventional view holds that seed prices are cyclical and will correct when markets chill. Ev argues the logic runs the other way: the traffic-wave dynamics make seed the one stage insulated from public-market contagion, so capital rushes there during turmoil, and great founders skip straight to Series-A-sized rounds.

Ev: "What the seed round has that no other round has is, unless the company fully doesn't work, it will always be the cheapest round available. And so, there is always some safe haven there." 00:51:51

Ev: "these great founders are not saying, 'I'm going to do my angel round and dilute 15%'... They're saying, 'I'm basically skipping to my Series A.'" 00:52:27

Mega-funds raising now to "have dry powder for the downturn" is a narrative, not a plan

The widely repeated LP pitch is that firms need dry powder for the coming slowdown. Jack skewers the logic: they are deploying fast in the potential bubble and then expecting to buy the same assets at a discount.

Jack: "this last year, which could totally been a bubble, we're going to invest the entire fund, but then we're going to be able to buy all the same things at 50% off." 00:46:43

Ev adds that raising now is still rational: "it probably is smart of all these firms to raise this money right now because you still can." 00:45:55

The mega-funds are suffocating the mid-sized funds, and the small early-only funds are the safest

Most commentary frames mega-funds as competition for deals. Ev frames the pressure as LP-wallet squeeze: the biggest firms are approaching the total desire for venture exposure, and mid-sized funds get crushed in the middle.

Ev: "there is a group of funds that's sort of mid-sized that I do think gets a lot of LP squeeze as a result of these other funds just being so big that they truly are starting to push up against the total desire for venture capital investment." 00:47:23

Jack: "the mega funds are suffocating the mid-sized funds." 00:47:44

Do not dismiss public skepticism of AI as ignorance

The standard Silicon Valley reflex is that people outside the coasts "don't get it." Ev argues the opposite, grounded in his Midwest roots: broad dislike signals a real threat, and the industry's own narrative made it worse.

Ev: "it drives me crazy when people in Silicon Valley are like, 'Oh, people in the middle of the country don't understand.' I'm like, 'No, screw you. I'm from the middle of the country, and I understand.'" 00:38:57

Ev: "we've gone around telling people, 'We're going to take all their jobs, we're going to plop down a data center that they have to work in, and then we're going to use that data center to automate away the work that they've learned for the last 20 years.'" 00:39:28

Narrative drives public-market price more than numbers, even for the biggest, most scrutinized companies

Contrary to the "numbers matter at scale" assumption, both hosts treat narrative as the dominant force. That has direct implications for how to read cheap multiples on semis.

Jack: "it's amazing the degree to which the narrative is driving the market right now." 00:07:45

Ev: "you can look into the numbers and say, 'Well, this is so cheap on this metric.' ... SK Hynix or Micron trade at three times earnings... people don't want to buy the stock" 00:07:01

3. Companies Identified

Anthropic

Frontier AI lab maker of Claude, now going public. Mentioned for run-rate revenue of $81 billion at end of October (up $10B month-over-month), an IPO range of $70–78 per share, and as the hedge fund community's favorite name. Ev: "run-rate revenue was $81 billion at the end of October, which was up 10 billion from the previous month, which was a much better growth month sequentially" 00:02:02 Ev: "every single one absolutely loves Anthropic." 00:05:57 It also drove the AI trade: "the AI trade was basically built on the month-to-month net new ARR estimates from Anthropic and OpenAI" 00:25:17

OpenAI

Frontier AI lab expected to IPO possibly next year. Mentioned for aggressive API price cuts that took share from Anthropic, evidence of low switching costs. Ev: "OpenAI, you know, reduced prices for um on the API, and very successfully, for a few months, very successfully took share from Anthropic." 00:08:01 Jack: "OpenAI will maybe IPO next year." 00:45:09

Nvidia

GPU maker, down about 30% from peak in the last month amid the Jensen Huang scandal. Mentioned as the central funder and enabler of open source and neo-clouds, using the commoditize-your-complement strategy. Ev: "he's been sort of the key figure to arm the rebels against the frontier labs" 00:12:38 Ev: "Nvidia is so structured around Jensen... an Nvidia without Jensen, it's sort of hard to even imagine" 00:14:02

CoreWeave

Neo-cloud public stock. Used as the example of a higher-beta, more levered AI infrastructure name. Ev: "stocks like CoreWeave, um, and others that are a little bit more on the higher beta range of the AI, um, infra trade in public markets, are down more than 40%." 00:18:23

SK Hynix and Micron

Memory semiconductor makers. Cited as examples where cheap multiples don't translate to price strength without buyer demand. Ev: "SK Hynix or Micron trade at three times earnings. Like, how could that happen?" 00:07:01

Microsoft, Meta, and Google

Hyperscalers with strong balance sheets. Contrasted with levered neo-clouds as the safer players during the CapEx questioning. Ev: "Microsoft, and Meta, and Google, who have, you know, billions and billions of dollars of free cashflow and really strong balance sheets" 00:22:03

Meta (Louisiana Hyperion project)

Cited as the exemplary case study of a data center done right for a community. Jack: "Meta's Louisiana Hyperion project... that project alone doubled the local tax receipts for that county and gave teachers bonuses that were equal to like their salary in many cases." 00:37:06

Sequoia Capital

Venture firm that announced restructuring: closing the London office, pausing Arc and its scout program, merging early and growth, and ending the Monday partner meeting. Praised for aligning identity, strategy, and fund size. Jack: "for the first time in the last few years, in my mind, Sequoia has a very consistent identity, strategy, fund size, everything, like everything's nice and aligned." 00:58:35 Ev: "Those are smart, makes sense." 00:59:12

Andreessen Horowitz

Venture firm cited as the exemplar of a maximalist, "more is more" strategy where fund size, strategy, and identity align. Jack: "they're the preeminent example of more is more... They've done an amazing job of that." 00:56:13

Y Combinator

Accelerator cited for its bulk-then-cut cycle under successive leaders. Ev: "when my brother was running YC, that was a massive bulking situation, and then Garry came and he did a cut. And I think they were both smart." 00:59:46

SpaceX

Cited as both a liquidity engine for LPs and as an example of a business that absorbs massive capital. Also referenced as dependent on the government relationship. Ev: "SpaceX has been a huge liquidity boom." 00:44:22

Anduril and Palantir

Defense/government-dependent technology companies. Mentioned as examples of companies that require strong government relationships, at risk if DC polarization shifts. Ev: "think about SpaceX, and Anduril, and Palantir, and these companies that really require the government relationship to be strong." 00:34:13

Databricks and Stripe

Mature private companies cited to show that outside the frontier labs, good businesses don't need massive capital. Stripe was also used as the hypothetical AI-consuming customer. Ev: "how much money do Databricks and Stripe really need?... It's not $150 billion." 00:43:50

Cognition

AI coding company, described as rhyming with the frontier labs but not needing tens of billions. Jack: "there's some that maybe look a little bit like it, like a Cognition" 00:45:22

Tesla (Model Y L)

Ev's product recommendation, called the best car of all time. Jack: "Yes. Best car of all time." 01:02:05

Citrini Research

Research outfit whose chart juxtaposes hyperscaler CapEx against AI lab ARR. Ev: "There's a chart from um Citrini Research that shows hyperscaler CapEx versus AI lab ARR." 00:26:14

Amazon (HQ2)

Referenced for the 2017 HQ2 sweepstakes, which exemplified the old power dynamic between companies and municipalities. Jack: "you had Amazon sitting there basically like waving a dollar bill, being like, 'We have $5 billion that we're going to invest in the community and 50,000 jobs'" 00:35:57

Lattice

Ev's former company, referenced for his experience seeing public-private market contagion at a slower cadence. Ev: "I remember when I was running Lattice, I would like see this happening, but it would take 6 months." 00:50:45

Astra and Fable (robotics reference)

Mentioned by Jack as examples that could benefit if compute and model scale eventually crack robotics. Ev separately mentions Fable and Mythos 5 2 as fresh Anthropic model releases timed with the IPO. Jack: "if it just turns out that you just throw enough compute and enough, you know, size of model eventually at robotics, then, you know, Astra, and Fable, and all these guys, all of a sudden you've got good robotics." 00:29:36

METR (Meter)

AI evaluation organization that released findings suggesting signs of recursive self-improvement (RSI). Ev: "Meter also had a release that they believe that they're seeing more and more signs of RSI." 00:28:46

Buffer

Tool through which Ev now accesses Twitter after deleting the app. Ev: "I went cold turkey. So, now I only access Twitter through Buffer." 00:04:38

Apple (iPhone Duo)

Ev's recommended product from the prior episode, described as a hit. Ev: "The Duo is unbelievable." 01:02:00

4. People Identified

Jensen Huang

CEO of Nvidia. Discussed at length as the central figure in open-source and neo-cloud ecosystem support, now embroiled in an alleged Chinese-spy scandal. Ev: "if Jensen is removed from the CEO role, that is a huge headwind... Nvidia is so structured around Jensen." 00:13:35 Jack: "he can't vocally be, you know, like, you know, pro-China, pro-open source, pro-exporting chips." 00:14:44

Dario Amodei

CEO of Anthropic. Referenced jokingly as a hypothetical tail risk to illustrate volatility. Ev: "maybe Dario gets caught with a with a Chinese spy, you know." 00:10:21

Ben Thompson

Stratechery analyst and Sharp Tech co-host. Cited for the argument that selling chips to China lets them build on CUDA, and for surfacing the Amazon HQ2 analogy. Ev: "many smart analysts like Ben Thompson that would argue that there's very rational reasons why Jensen would want to sell chips into China" 00:14:50 Jack: "Ben Thompson on his show Sharp Tech recently brought up..." 00:35:23

JD Vance

US Vice President. Cited for acknowledging the midterm loss and the voter message on war and inflation. Jack: "JD Vance even like, you know, he had that big tweet... 'We heard loud and clear.'" 00:31:05

Gina Hinojosa

New Democratic governor of Texas. Cited for a measured, partnership-oriented quote on AI data centers rather than hostility. Jack: "'that work with AI companies has to be a partnership that benefits Texans. And right now, what has happened is that data centers have moved into mostly rural communities and there are no rules and laws to protect Texans'" 00:35:23

David Cahn

Sequoia partner whose post on the gap between AI CapEx and revenue is recalled. Jack: "I think maybe it was David Cahn at Sequoia wrote that post... the revenue's got to catch up at some point." 00:27:09

Garry Tan

YC CEO, credited with a "cut" after expansion. Ev: "Garry came and he did a cut. And I think they were both smart." 00:59:46

Ev's brother (Y Combinator leadership)

Ran YC during its expansionary period, per Ev, who cites it as an example of a "bulking" phase. Ev: "when my brother was running YC, that was a massive bulking situation" 00:59:46

5. Operating Insights

Alternate between "bulking" and "cutting" phases, and don't confuse a cut with a mistake

Ev's framework: organizations, and the best founders, expand to try many things and then simplify to the core. Shutting down an initiative does not mean launching it was wrong, since the experiments may have produced the good outcomes.

Ev: "the best founders that, you know, we work with... they try a bunch of stuff, they're shipping a new thing every week, every month, and they cut. And I think that's a healthy thing." 01:00:32

Jack: "just because they did these things and shut them down, it actually does not at all mean they were wrong to do them in the first place." 01:01:09

Check that identity, strategy, and scale are aligned, because misalignment is when things break

Jack's diagnostic for any organization: your self-image, strategy, and size must be coherent. There is no universally right design, only alignment.

Jack: "you think you're one thing, but your operations, your strategy, your fund size, everything else is out of whack with that thing that you think you are." 00:58:56

Zoom out in volatility, and cut the information feed that distorts your sense of the trend

Jack's advice is to anchor on stated goals rather than weekly news. Ev's tactical move was deleting Twitter, consuming it only through Buffer.

Jack: "if you just didn't follow any of the news, and you just kind of were like, 'Well, you know, they said this year, the hope was to go from 10 to 100 in run-rate'... it would have just sort of been like what you thought it was going to be." 00:03:10

Ev: "when you're in the fog of war in times of volatility, it's very hard just to zoom out." 00:03:46

Founders and investors should model the next round's market, not just the current one

Because private markets lag public turmoil by months, and seed insulation changes pricing dynamics, operators should plan runway and fundraising windows around the traffic-wave dynamic instead of assuming a steady market.

Jack: "you want to think that you're going to be able to, one, get, you know, your next round done, and two, ideally, that like you see share price appreciation all the way to IPO." 00:50:27

Use narrative-aware analysis, not just fundamentals, when evaluating exposure to public markets

Treat narrative as a first-order input: ask who the marginal buyer is and why they would want to own the asset.

Ev's friend: "The way that stocks go up is that more people, more volume wants to buy it than sell it." 00:07:01

6. Overlooked Insights

Ev's tell on the "AI trade" structure: one metric, two leading indicators, and an engineered feedback loop

Buried in the unwind discussion is a precise description of how the whole AI infra complex became a single-variable trade: lab ARR is the proxy for CapEx dollars, and CapEx dollars are the revenue of every semis, power, and shell company. That means public equity prices across a dozen sectors were effectively derivatives of two private companies' monthly ARR adds. It explains the synchronized collapse and implies that the same chain reverses on the upside, so any lab ARR re-acceleration can re-rate everything at once.

Ev: "what does AI lab ARR represent?... how much CapEx can be built. What is CapEx?... It's GPUs. It's power. It's the shell... all of the semis stocks are the dollars of that goes into the CapEx figure. And so like you have like these two leading indicators" 00:25:27

"Frontier labs were a one-time subsidy" on venture's capacity to absorb capital, and the next capital-hungry categories are unproven

Said almost as a joke ("intelligent design"), Jack's point is a thesis about the entire fundraising supercycle: mega-funds scaled in lockstep with a unique, once-only outlet for deploying tens of billions at venture-like returns. With the labs maturing into IPOs, the remaining capital-hungry categories (neolabs, robotics) are unproven asset classes, while the rest of the great companies simply don't need that much money. That creates a structural mismatch between fund sizes and investable opportunity, which could surface as compressed returns, not just a squeeze on mid-sized funds.

Jack: "there's like two or three categories that are very capital hungry that are completely TBD on if they're great returns as like a asset class or like a sub-asset class." 00:45:34