Theme · Updated
What podcasters are saying about Frontier AI labs & model race
Summary
In the past 30 days, 27 speakers on 20 finance podcasts discussed Frontier AI labs & model race in 34 views. Chorus has followed Frontier AI labs & model race since September 25, 2026: 35 views from 28 speakers on 21 shows.
Who talks about it most
- Citrini · 3
- Steve Eisman · 2
- Paul Kedrosky · 2
- Naveen Rao · 2
- Eric Schmidt · 2
- Ben Inker · 2
- Jeetu Patel · 1
- Jack Altman · 1
- Steve Hou · 1
- Ethan Mollick · 1
The latest views
Newest first. Every quote is checked against the episode's transcript.
Jeetu Patel · Cisco
The AI market has many players, not one or two, so monopoly concerns are misplaced.
Responding to Tom Keene's question about roll-ups and monopoly or monopsony outcomes, Patel says the number of participants in the market today is not one or two but a fair number of players. He expects business models to shift as agents arrive, but he gives no evidence on competitive dynamics or consolidation. The claim is a brief, unquantified view on market structure.
“The number of players that are participating in this market today is not one or two. There's Yeah. There's a fair number of players.”at 9:01
Bloomberg Talks: Cisco Systems President & Chief Product Officer Jeetu Patel Talks AI Industry ·
Jack Altman · Benchmark
Customers pay for the leading-edge model even at higher prices, so frontier leadership carries outsized momentum and pricing power.
Labs have swapped the lead all year, and momentum follows whoever has the best model, even by a small margin and even when the older model handles most tasks. Users seem to pay for top intelligence despite price sensitivity because individual value is in the thousands or tens of thousands of dollars, and that surplus is large. For unbounded-upside tasks such as coding, sales, drug discovery and robotics, users will keep buying the better model. Jack also notes that price per gigawatt and the monetization of frontier tokens have surprised everyone.
“And the leading edge, even if it's by a little bit and even if the last model could do most of the tasks, it just seems like it's worth a crazy amount of momentum.”at 7:18
Uncapped with Jack Altman: Benchmark Breakfast #1 | Jack and Ev ·
Steve Hou · Silicon Data
Agents make models interchangeable, so the moat may lie in accumulated user context rather than the model.
Hou says agents make users almost model-agnostic. He uses agents without knowing which model powers them and cares only about whether the outcome is delivered. He still sees some moat, because over time an AI knows the user better, and the more it has learned, the harder it is to leave. He is explicit that he is not making a bullish or bearish case, and says the moat is an open question.
“the funny thing about agents is that agents actually are making you almost model agnostic... All I care about is whether or not it's actually delivering an outcome.”at 53:46
Monetary Matters: Why Compute Prices Are Still Rising | Steve Hou of Silicon Data on GPU Rental Indices, Fed Meeting, and Anthropic S-1 (LIVE) ·
Ethan Mollick · Wharton School, University of Pennsylvania
AI labs' market has grown faster than they expected; demand exceeds supply and all are compute constrained.
Mollick reframes the surprise around AI progress. The labs were not surprised by capability gains so much as by market demand, which has exploded faster than they expected. Demand has been higher than they anticipated, and they are all now compute constrained. He adds that Ilya Sutskever's data-wall warning was not a claim that forward progress would stop.
“Their market has exploded [better] than they thought, which is different. Demand has been higher than I think they anticipated. Right? And they're all compute constrained at this point.”at 33:26
Big Technology Podcast: This AI Expert Argues We're Underestimating AI's Productivity Impact — With Ethan Mollick ·
Steve Eisman · The Real Eisman Playbook
Anthropic has no moat; it can only win share through price, which is a dangerous franchise.
Eisman argues that Anthropic's franchise has no moat. If the reported 50% price cut is true, it shows that price is the only lever for gaining share. He says a business that commits to about $800B of compute while its franchise rests entirely on price is dangerous. He hedges that he has not confirmed the price cut, but says that if it is true it reinforces the no-moat thesis. Chinese open-weight models are taking share, which adds to the pricing pressure.
“it just seems, you know, that there are no moats around this business... if it's the case that they actually did cut prices by 50%, that would reinforce the concept that there are no moats and that the only way you're gonna gain share is price.”at 11:32
Prof G Markets: Why Steve Eisman Isn't Buying Anthropic's IPO ·
Cliff Sosin · CAS Investment Partners
Frontier labs' pricing power erodes as models commoditize; only a narrow set of hard tasks needs the frontier.
Each new frontier model unlocks a larger addressable market, but once evaluations saturate, the tasks commoditize. Customer-service chatbots, for example, would never use a frontier model. A wave of hard tasks needs the frontier, and behind it, most work does not. Cheap Chinese models have shown rapid commoditization at every step, and token costs have collapsed. Even if a lab is the cost leader, the market for tokens behind the frontier may be small, and the margin for a cost leader is set in a shrinking commodity market. Demand may also become rarefied: few use cases need very elite intelligence, and many need good intelligence, which is competitive and commoditized. Sosin also says his friends who are frontier bulls may have an information advantage, and he finds the claim that labs can keep producing cheaper tokens hard to judge.
“Certainly, the trend has been rapid commoditization of these things if you were to look at the rise of Chinese models in every step.”at 14:33
Invest Like The Best: Cliff Sosin - Owning the Complements to Intelligence - [Invest Like the Best, EP.497] ·
Satya Nadella · Microsoft
Full vertical integration from silicon to apps is unlikely; competition is intensifying at every layer of the stack.
Nadella says he once thought one vertically integrated company might control everything from models and apps down to silicon and systems, but he now thinks that is unlikely. Instead, competition is growing at every layer: silicon, systems, the model layer and the app layer. He expects roughly five or six closed and five or six open-weight models to compete. A builder can route each task to the best model through a decision engine. He concedes the eventual outcome depends on which dominant design emerges.
“If anything, it's going the other way, which is every day you wake up and you say, Wow, there's more competition in every layer of the stack, whether it's in silicon, whether it's in systems”at 16:28
20VC: 20VC: Satya Nadella on Open vs Closed Models: Who Wins | Whether to Trust Frontier Model Providers | Why Companies Need to Own Their Own Intelligence | The Importance of Verticalisation: From Chips to Models and More ·
Paul Kedrosky · SK Ventures
Labs can't slow spending; the US-China prisoner's dilemma locks in the race.
Each firm believes that if it stops spending, competitors will win a market likely dominated by a few players in inference and token supply. That makes continued spending mandatory. Anthropic's call to 'pace the frontier' is a prisoner's dilemma: even if labs agree to slow, China may not, and then China wins. A West Coast culture that treats technological progress as inexorable and anti-slowdown reinforces the dynamic. Capital markets, collecting fees, add more money to keep it going.
“What if we all agree but China doesn't? Well, then we're into this problem where China wins again, and we're back to this existential crisis. And so the race dynamic is baked into the very nature of what we're doing.”at 17:01
Hidden Forces: The Most Dangerous Investment Cycle In History | Paul Kedrosky ·
Paul Kedrosky · SK Ventures
Frontier model gains are slowing: composite benchmark improvement fell from ~18% to ~5-6%.
Paul argues that model progress is becoming asymptotic. On composite benchmarks, year-over-year improvement fell from about 18-19% three years ago to about 5-6%. Many benchmarks have also been ingested by the models, so headline results overstate progress. Labs keep launching new models and citing benchmark wins to maintain the story that justifies funding. If training returns had not declined, the first lab to stop training would win on costs, but the race makes that move impossible.
“while models are improving, it's becoming somewhat asymptotic... We went from, like, say, around 18, 19% performance year over year on composite benchmarks three years ago to more like five or 6%.”at 28:46
Hidden Forces: The Most Dangerous Investment Cycle In History | Paul Kedrosky ·
Citrini
Lab revenue lags capability gains; combined lab ARR at $300B instead of $500B would be a disaster.
The pattern has been that model capability improves and lab ARR follows, but recently there is a big jump in capability without a commensurate increase in ARR, or a moderating ARR increase. The AI infrastructure build is being underwritten on assumptions that combined lab ARR reaches about $500B, which is what funds reinvestment into infrastructure. Consensus estimates reportedly imply roughly $1T in AI ARR by 2028, and Citrini says current run-rates need to grow sharply to get there. If combined lab ARR ends up around $300B, the shortfall would be a disaster for the financing assumptions behind the buildout.
“if you're relying on the combined lab ARR to be $500,000,000,000 and that's where you're underwriting your kind of assumptions of what gets funneled back into infrastructure, If it ends up being 300, then that's a disaster.”at 7:13
Monetary Matters: Citrini: Monkeys Are The Next Bottleneck | Citrini on AI Drug Discovery Experiment & Reflexive Financing Risk to AI Buildout & Hardware Trade ·
Citrini
AI labs and AI startups will become major buyers of experimental data to improve their models.
Before AI, demand for experimental work came mainly from pharma and biotech companies. Now AI labs and AI startups in biology become customers because they need experimental data to improve their models. Anthropic is a leading example, with biology a major focus. Citrini sees AI-generated experiments feeding back into models as a source of new demand for lab work, and the model-improvement loop cannot be sped up past the time experiments take.
“the AI labs themselves and AI startups in this area become customers because they need to generate the data, the experimental data, to improve their models, right?”at 51:26
Monetary Matters: Citrini: Monkeys Are The Next Bottleneck | Citrini on AI Drug Discovery Experiment & Reflexive Financing Risk to AI Buildout & Hardware Trade ·
Citrini
AI labs have weak moats because switching costs are near zero; defensibility will come mainly from capability.
Citrini does not accept the view that AI labs have no moat, but he argues the switching costs are close to zero. Users get used to how a model behaves, yet a competitor such as ChatGPT can release a model that is much better overall or for a specific task, and users can move. He sees specialisation as a possible source of moats, for example a biology-focused model that clearly beats a rival's. He expects any moats that do develop to be offshoots of capability, so the competition still comes down to which model is best.
“But the switching costs are zero, right? If [ChatGPT] comes out with a model that's way better or is way better for a specific thing.”at 1:24:05
Monetary Matters: Citrini: Monkeys Are The Next Bottleneck | Citrini on AI Drug Discovery Experiment & Reflexive Financing Risk to AI Buildout & Hardware Trade ·
Naveen Rao · Unconventional AI
Reinforcement learning is more compute-inefficient than pretraining, but will stay central for several years.
Measured by how much the model improves per unit of compute or per dollar, RL is less efficient than pretraining. It still keeps moving capability forward, and he expects it to stay very important for the next several years, with specialization building in the middle layers and perhaps pretraining.
“Well, I don't know that it's more efficient. Actually, I think RL is more inefficient. If you look at it in terms of how much better the model gets per unit compute or per unit dollar, it's actually more inefficient.”at 33:39
20VC: 20VC: Why Robotics Will be 100x Bigger than Coding Agents | Inference Will Be a $10TRN Market: What No One Sees | Why Software Investors are F***** Today with Naveen Rao, Unconventional ·
Naveen Rao · Unconventional AI
Big labs cannot afford long-term paradigm bets; startups take the big risks and big labs buy them.
Labs such as OpenAI, Anthropic, xAI and Google are racing, and being a month late on a model is a disaster, so they cannot take big bets on new paradigms. Deep innovation is being pursued by startups, such as the efforts of Yann LeCun and Fei-Fei Li. Big companies then take the incremental work and buy startups when they show promise, a pattern similar to drug development.
“Like, all the big risks are taken by startups, and all the incremental stuff is done by big companies. The big companies buy the startups when they show promise.”at 35:01
20VC: 20VC: Why Robotics Will be 100x Bigger than Coding Agents | Inference Will Be a $10TRN Market: What No One Sees | Why Software Investors are F***** Today with Naveen Rao, Unconventional ·
Nikesh Arora · Palo Alto Networks
Frontier labs aren't causing the cyber problem; they're compute-constrained and rate-limiting while grappling with power.
Arora rejects the theory that frontier labs gatekeeping their most powerful cyber-defense tools is causing the chaos. He says most lab leaders are researchers who are surprised by how powerful the models are and are asking partners what to do with them. The labs are compute-constrained and rate-limit even enterprise customers who have access to the newest models. He sees them trying to figure out how to release and monetize the models, rather than causing the problem.
“So I think they're not causing the problem. They're just trying to grapple with the fact that these things are extremely powerful.”at 28:23
All-In Podcast: AI Cyber Panic, Ken Griffin's Bubble Warning, The Fall of France? ·
John Belton · Gabelli Funds
OpenAI and Anthropic combined revenue should grow six to seven times this year, which is encouraging.
Belton concedes Anthropic looked to have slowed in the third quarter. He argues that does not change the picture, since OpenAI and Anthropic together will probably grow revenue six or seven times this year, reaching a massive scale. He presents this as evidence that AI is generating real monetization outside the infrastructure layer.
“Anthropic looks like they slowed down in the third quarter. It doesn't change the fact that combined OpenAI and Anthropic probably get a six or seven X revenue this year to a pretty massive scale, so that's very encouraging.”at 8:25
CNBC Closing Bell: Closing Bell Overtime: Markets Search for Opportunity After a Wild Week 10/23/26 ·
Warren Pies · 3Fourteen Research
Combined annualized revenue of the two leading AI labs should reach $170-190B by year-end, up from $30B.
Pies projects that combined annual revenue run rate for the two labs will reach $170 billion to $190 billion by the end of the year, starting from $30 billion at the beginning of the year. He calls this huge growth and uses it as evidence that the AI investment story is intact. The transcript does not name the two labs.
“I think lab ARR when we get to the end of the year combined for the two labs is gonna be, you know, one seventy to one ninety starting the year at 30,000,000,000. That's huge growth. I think it's intact.”at 45:31
George Noble · Noble Capital Advisors
AI labs' revenue growth is slowing rapidly, and the market rejected OpenAI's accounting explanation.
George points to OpenAI postponing its IPO to 2027 and to talk of slowing revenue growth at Anthropic. He cites a recent OpenAI revenue miss, which the company blamed on a gross-versus-net accounting change, and says the market was not convinced. He argues growth rates of 45% don't matter when a company is valued at $1.5 to $2 trillion on $50 to $70 billion of revenue, and that the slowdown is what drives capital flows.
“And if you look at these companies, the revenue growth is slowing rapidly.”at 8:25
Monetary Matters: "Coreweave's a Zero" | George Noble on Why He's Short Neoclouds, Why AI is a Shale 2.0 Bubble Headed for Debt/Depreciation Brick Wall, and Why He's "Coming For The Momentum Bros" ·
Public listings of frontier AI labs will accelerate the AI infrastructure arms race.
Ives expects Anthropic to go public in November and OpenAI to follow. As these companies become public, the arms race gets accelerated. He says tech companies see roughly twelve to eighteen months to build out before the window closes, and customers are waiting in line for capacity.
“You talk about Anthropic that will go IPO in November. You know, as they become public, as OpenAI goes public, whether it's 1,200,000,000,000 or 1,400,000,000,000, the arms race just gets accelerated.”at 17:32
Bloomberg Surveillance: Bloomberg Surveillance TV: October 23rd, 2026 ·
Jeff Snider · Atlas Financial
Large language models are prediction systems with hard scaling limits, so their progress will plateau.
Snider argues AI models are prediction systems, not intelligence, and that they face real technical constraints. He cites the Chinchilla scaling laws: performance gains from bigger models depend on ever more training data, and the internet's high-quality human text is running out. As models train on increasingly AI-generated content, errors compound, and paying humans for high-quality books and journals has diminishing returns. He says there are reports the plateau in productivity was reached last year.
“These are prediction models. These are not intelligence. There's no thinking and thought behind them.”at 2:58
Eurodollar University: Anthropic Has $750 Billion in Commitments… and Just $17 Billion in Revenue ·
Jordi Visser · Visser Labs
Anthropic and OpenAI ARR could reach about $170B this year, up from $30B, with no bubble signal.
Anthropic and OpenAI combined had about $30B of ARR at the end of last year and are expected to finish this year near $170B, a jump Visser says nobody expected. Coding agents drove the growth, and the combined revenues this year exceed the entire software stack. The two companies are worth about $3.5T combined before any IPO, up from almost nothing about fifteen months ago. Despite this, people still debate whether ARR will fall back.
“At the end of last year, Anthropic and Open AI combined had 30,000,000,000 in ARR. They're gonna finish this year somewhere around a 170,000,000,000, so they've gone from 30,000,000,000 to a 170,000,000,000.”at 8:15
The Julia La Roche Show: #423 Jordi Visser: Why the AI Bears Are Wrong — and the One Asset He’d Bet On Right Now ·
Tobias Carlisle · Acquirers Funds / The Acquirers Podcast
OpenAI and Anthropic are already building AI that builds smarter AI, so we're in the singularity now.
Carlisle argues the singularity is defined by machines that make smarter machines. He says OpenAI and Anthropic are doing this now or will do so this year, so 'we're living in the singularity right now.' He frames this as a present-tense fact, with the timing set by 'this year'.
“OpenAI and Anthropic are making the machines that make the machines smarter. They're either doing that now or that's... They're gonna be doing that this year.”at 41:18
Brett Winton · ARK Invest
Closed frontier labs will pull away because their models run months ahead and cost less to run.
Brett argues the closed frontier labs can accelerate away from competitors. They have models about six months ahead of everyone else, and they can use those models internally at cost rather than paying market prices. If gross margins are true for Anthropic, competitors pay about four times more, so the lab's R&D dollars are worth roughly 100 to 1,000 times more than a rival's. He says performance per unit cost doubles every month or two, so the lead compounds. He also argues the frontier labs can distill their best models into cheaper ones and dominate the whole cost curve. This is like having a Ferrari that can also be compressed into a Honda Civic. He thinks SpaceX is a test case, since it currently rents compute and is spotting the leaders a lead.
“the reason the Frontier Labs, these closed labs, probably could accelerate away is because they have models six months in advance of everybody else, and they get to access those models at cost...”at 18:48
ARK Invest (FYI podcast, In the Know video): The Real Reason Tesla Announced The Roadster | The Brainstorm 154 ·
Christian Munafo · VanEck
Frontier labs keep most revenue despite open models taking token share, because enterprises pay for high-value use.
Open-source models now carry over 70% of token consumption, but Anthropic and OpenAI still take over 70% of revenue. High-impact use cases such as advanced coding, software engineering, forecasting and agentic workflows come at a premium and need enterprise security. Switching costs for enterprises like VanEck are high. Open models are about four to six months behind the frontier, a gap Epoch AI data shows has been consistent. Christian says the team lacks full conviction on ten-year outlooks, so this is an ongoing debate.
“the majority of the revenue generation continues staying with these frontier models... those high impact, high value use cases come at a price.”at 22:16
Thoughtful Money: The Big Squeeze: High Bond Yields Now Becoming A Problem For Stocks | Jan van Eck ·
Eric Schmidt · Relativity Space
Frontier models go obsolete within about five months, so release cycles are compressing toward months and then days.
Schmidt describes a compressed development cycle: a frontier model takes about six months to build and about two months to test, then is antiquated within five months of release. Each model costs hundreds of millions of dollars, by his conservative estimate. Competition between Anthropic and OpenAI, and Google's Gemini, shifts month to month, with the timing measured in months rather than years. He expects the next cycle to be measured in days.
“The models, for example, take roughly six months to build, two months or so to actually test, and they're released and they're antiquated within five months. The models themselves cost hundreds of millions of dollars to make.”at 4:24
In Good Company with Nicolai Tangen: Eric Schmidt: The Age of Agents, Superintelligence, and Why Europe Is Falling Behind ·
Eric Schmidt · Relativity Space
Western frontier models have outperformed expectations, and even stronger versions are coming soon.
Schmidt says the group of leading Western AI developers was surprised by how capable the latest model is, which he says is better than he expected. He adds that another version is coming that looks better still, and that other Western models also appear stronger than he had thought. He also flags that Chinese models may have similar capabilities, though he cannot verify this.
“I think all of us were surprised at the power of mythos... but it's better than I thought it would be. And now there's another version coming which looks like it's better than that.”at 26:37
In Good Company with Nicolai Tangen: Eric Schmidt: The Age of Agents, Superintelligence, and Why Europe Is Falling Behind ·
Max Wiethe · Monetary Matters Network
Most everyday AI work doesn't need superintelligence-level frontier models, questioning the race's payoff.
Max says the assumption was that everything would be done with frontier models solving harder problems. But most work does not require a superintelligence to get done. Reading emails or sorting podcast transcripts does not need it, and most people don't need to be at the cutting edge. He takes this as a reason to doubt the value of the frontier-model push.
“well, actually, maybe most of the work doesn't require, like, a 250 IQ superintelligence to get things done.”at 40:04
Monetary Matters: Is The AI Trade Over? | Jack & Max on Panic in AI Credit, SpaceX Earnings, and Semis ·
Steve Eisman · The Real Eisman Playbook
AI models have no moats; users switch freely between Claude, Gemini and ChatGPT.
Eisman says the AI business has no moat at all. Unlike Google, where users stayed for years, people move between Claude, Gemini and ChatGPT as the best model changes. He argues this is why the labs are spending so heavily on lobbying and doom messaging, since open-weight alternatives erode whatever pricing power they have.
“And the basic problem with the entire AI enterprise is that there are no moats. None. Right. Absolutely none... One model is good one day and then people switch to the next.”at 2:02
The Compound and Friends: How OpenAI Could Cause a US Recession With Steve Eisman ·
Josh Brown · Ritholtz Wealth Management
Enterprise standardization on one LLM is a sneaky moat, since staff stay on what they know.
Brown's firm standardized on Claude Enterprise, and he moved his ChatGPT work to Claude because the firm already pays for it. He thinks the enterprise LLM is a sneaky way to build a moat: once Fortune 500 workers are used to a model, they are unlikely to switch. Eisman disagrees, arguing that they would switch to an open-weight model if it were much cheaper.
“I sort of think the enterprise LLM is a sneaky way to build a moat because you get all the workers of the Fortune five hundred accustomed to using it, and then why would they go use another?”at 3:28
The Compound and Friends: How OpenAI Could Cause a US Recession With Steve Eisman ·
Ben Inker · GMO
AI compute and models look like competitive, low-switching-cost businesses, so high returns on capital are unlikely.
Winner-take-all businesses earn high returns on capital through network effects and returns to scale. AI compute and model provision instead look like businesses with low switching costs and comparable offerings. The host notes that users switch between models, including open-source ones, across queries and tasks, and Inker agrees. Broad adoption of AI requires it to be cheap, and something that is cheap is hard to make wonderfully profitable. Together these point to weak returns for the model and compute providers.
“What we see is a lot of this is about either the provision of compute, or the provision of AI models. And what we see in both cases is those look more likely to be competitive businesses, than winner take all businesses.”at 1:54
Monetary Matters: The AI Bubble Will Pop from Equity Supply Says This $80 Billion Asset Manager | Ben Inker | GMO ·
5 older views on Frontier AI labs & model race are on the episode pages, and all of them are a question away in your AI with Chorus.
Ask your AI about Frontier AI labs & model race
Chorus gives Claude, ChatGPT and other assistants every view on Frontier AI labs & model race from the market's podcasts. Ask who's worried, who isn't, who changed their mind, and why.
Chorus reports what speakers said. It isn't investment advice.