Prof G Markets · · 1h 2m
Aswath Damodaran On Nvidia’s Collapse & Anthropic’s $1.5T IPO
With Aswath Damodaran
ChorusBring the market's conversation to your AI. Learn more
Summary
Aswath Damodaran joins Ed Elson on Prof G Markets to value Anthropic's proposed $1.5 trillion IPO and to assess Nvidia after its market cap fell from about $4.5-4.8 trillion to about $3.5 trillion. On Anthropic, he puts fair value near $1.475 trillion on a trillion-dollar revenue target, cuts it to about $77 a share after raising compute costs and regulatory drag, and says he would only buy after a large drop, given the founder-controlled 50.01% voting share, roughly $900 billion of planned compute commitments, and slowing revenue growth (47% in Q1). On Nvidia, he sees the China export and scandal risk as more serious than the CEO succession, expects structural margin pressure from hyperscaler custom chips, and has placed a limit buy at $140 while warning against large position sizes. He widens the view to macro, arguing AI capex has driven much of recent GDP and earnings growth, so a 2027 AI slowdown could trigger recession and a broad correction, while advising investors to add tech selectively rather than all at once.
Key takeaways
- Aswath Damodaran (bearish on NVIDIA): China scandal is the real risk to Nvidia; it may force giving up China sales and operating value regardless of CEO.
- Aswath Damodaran (bearish on NVIDIA): Nvidia's 70% operating margin is unsustainable; pressure from TSMC and hyperscalers will compress it over time.
- Aswath Damodaran (bearish on Anthropic): Anthropic's founder-held voting control is a governance risk that makes the shares unattractive without a deep discount.
- Ed Elson (bearish on Anthropic): Anthropic's revenue growth is slowing sharply, making the roughly $1 trillion revenue path unconvincing.
- Aswath Damodaran (bullish on Anthropic): Anthropic is one of two or three likely winners in enterprise AI agents, where stickiness favors early entrants.
What was said, by company and theme
25 views from 2 speakers. Every quote is checked against the episode's transcript (how).
Anthropic
Across the shows: 149 speakers in other episodes, 264 bullish and 123 bearish calls, past 30 days
Where everyone stands on Anthropic
Aswath Damodaran · NYU Stern School of BusinessBearish
Anthropic is worth about $77 a share, below the $83 IPO price, with equity value near $1.35 trillion.
Damodaran valued Anthropic at $1.475 trillion on November 2, using a trillion-dollar revenue path and a 27.5% margin. In the last 20 minutes before the interview he raised compute costs, since building data centers will get more expensive and each dollar buys less revenue. He also added a higher regulatory drag. That cut equity value by about $100 billion to roughly $1.35 trillion. Adding the updated share count and the large option overhang gives $77 per share. The IPO was priced at $83, which terrifies him because it sits so close to a banker's number. Bankers repriced from around $1.1-1.25 trillion after institutional enthusiasm, and they usually aim for a 10-15% opening pop. He would watch where the stock closes, not the offer or open. A close around $71-72 would make it interesting to him, but he would need a large drop because of the governance structure.
“My estimated value, if you bring in the share count and the fact that they have a big chunk of options outstanding, came up with a value of $77.”at 24:07
Aswath Damodaran · NYU Stern School of BusinessBearish
Anthropic's founder-held voting control is a governance risk that makes the shares unattractive without a deep discount.
Damodaran says Anthropic's structure is designed badly for shareholders. Seven founders hold a single class of founder share with 50.01% of voting rights, far beyond the dual-class structures seen at Google. He compares it to the golden shares governments kept when they privatised companies in the 1980s and 1990s, which gave them veto power over big decisions. Anthropic's founder share is a golden share with more power. An activist regulator could try to expropriate it and pay only the Class A price, even though the share might be worth hundreds of billions. Founders could challenge that in court, but the structure makes Anthropic a target. This is why he would not buy the stock without a large drop.
“Since Google, we were at voting shares and non voting shares. We're used to that. Right? This is a single [share] held by seven founders with 50.01% of voting rights.”at 26:37
Aswath Damodaran · NYU Stern School of BusinessBearish
Anthropic's multi-year compute commitments are a mismatch for an unprofitable company and could sink it without revenue.
Damodaran says unprofitable companies should not borrow against future cash flows, and Anthropic is doing exactly that with contractual compute commitments. Ed Elson pointed out that Anthropic plans to spend about $900 billion on compute and infrastructure by 2030, much of it contractually fixed, versus tens of billions today. Damodaran says that would put most young companies out of business. Two things cushion Anthropic. Its cash balance after the IPO should be about $200 billion, which covers the large commitments for the next two or three years. He also puts the debt value of its leases at close to $500 billion and subtracts it from equity. The breathing room lasts only if revenue ramps up quickly. Compute spending is not the problem in itself; the problem is failing to generate the revenue needed to justify it.
“And in effect, Anthropic is violating that principle by entering into contractual commitments without the earnings to back it up.”at 30:47
Aswath Damodaran · NYU Stern School of BusinessUncertain
Anthropic needs revenue to roughly double yearly to reach $1 trillion in ten years; revenue is the key uncertainty.
Damodaran's valuation rests on a trillion-dollar revenue target in ten years, which would make Anthropic larger than any company now earning over $800 billion. Reaching that means revenue must roughly double every year, at least in the near term. If revenue ramps quickly, the compute commitments, profits, cost of capital and default risk all work out. If it doesn't, the company is squeezed. He calls revenue the number to watch most in the financial statements, because it drives the rest of the valuation. Whether there will be enough demand for AI products and agents to sustain that growth is an open question for Anthropic.
“To me, the central input that drives this valuation, the number that I would watch the most, is the revenue number.”at 33:36
Ed Elson · Prof G MediaBearish
Anthropic's revenue growth is slowing sharply, making the roughly $1 trillion revenue path unconvincing.
Ed Elson notes Anthropic's Q1 revenue growth was enormous but has fallen to 47%. Reporting suggests it could drop to 15-25% in Q4. Doubling next year would only work if growth stops slowing. Combined with the roughly $900 billion of compute and infrastructure spending planned through 2030, and the need to either raise hundreds of billions or reach a trillion in revenue, he says he is not convinced the company can make the revenue ramp it needs. He treats the deceleration as an important warning.
“I [agree] 100%, which is why I was concerned to see that the revenue growth is slowing down as much as it is.”at 34:28
Aswath Damodaran · NYU Stern School of BusinessBearish
Anthropic faces price pressure from OpenAI and open-source models, and most businesses don't need its frontier models.
Damodaran says Anthropic's slowing growth comes from two causes: running out of compute and OpenAI underpricing it. OpenAI will keep underpricing, forcing Anthropic to cut prices to match. Open-source models that are less powerful but nearly as effective add more competition. Most businesses don't need Claude even if they adopt AI internally, since it's overkill for their needs. He thinks Anthropic must stop building state-of-the-art frontier models with every feature, like a car with every add-on, when customers just want a reliable vehicle. Both the compute shortage and the competitive pricing can be true at once.
“OpenAI is going to continue to be that. They keep underpricing. You're going to have an issue because you're going to have to bring prices down to match their prices.”at 35:29
Aswath Damodaran · NYU Stern School of BusinessBearish
Anthropic's drug-development push conflicts with its enterprise sales pitch and may not capture the value it creates.
Anthropic plans to move from selling AI tools into developing drugs the way young biotech firms do, selling them to large pharma. He sees a conflict: a consulting firm like McKinsey using Claude would worry that it learns its consulting resources and then faces Anthropic as a competitor in consulting and banking. That undermines the sales pitch to the rest of the market. He also argues that AI-driven cures would mostly create consumer surplus, and that a low-cost AI player could underprice pharma incumbents. He doubts that a large share of the value created goes to Anthropic.
“The problem with that story is if anthropic tells its story about the drug business, that they're going to be a player... It actually gets in the way of the sales pitch that anthropic is making to the rest of the world.”at 37:32
Aswath Damodaran · NYU Stern School of BusinessBullish
Anthropic is one of two or three likely winners in enterprise AI agents, where stickiness favors early entrants.
He expects enterprise AI agents to become a big business dominated by two or three players. Anthropic is the most focused of the AI labs on building for enterprises rather than offering a add-on feature. Once an agent is embedded in a firm and learns from its data and staff, switching costs make it sticky, and first movers can keep later entrants from underpricing them. He admits he could be wrong: firms may use agents without giving them proprietary data, which would remove the stickiness. He stresses his optimist hat must stay on for the bull case to work.
“It's going to be a big business with two or three players. And with Anthropic, I see it as one of the players. It's the most focus of the AI companies in terms of saying, look, we have a... We're building things for enterprises.”at 40:16
Aswath Damodaran · NYU Stern School of BusinessBearish
Anthropic's governance structure may leave its leaders unwilling to change course when strategy fails.
Even in the best case, Anthropic will likely need to try a path that fails and then adapt. He worries the way the company is structured gives its leaders little incentive to change direction when they do not want to. The only remaining check would be the market, pushing down the share price, which he calls a blunt instrument. He compares it to Facebook, where heavy losses were needed before Zuckerberg abandoned the metaverse, and says the AI version would need about ten times that damage. This governance concern is why he treats Anthropic as something to buy only at a discount.
“I think the way Anthropic is structured, worry about whether the people running Anthropic will be willing to change because what's the incentive?”at 42:17
Aswath Damodaran · NYU Stern School of BusinessBullish
Anthropic is worth about $1.475T, and he would buy only at a discount to that value.
His valuation of about $1.475 trillion assumes the optimistic case, in which AI delivers at least half its promise and Anthropic becomes one of a few dominant enterprise AI players. He says Anthropic is not a stock for a short list of five holdings, because the risks are large: disruption, governance, and the need to change strategy. He is still willing to consider it as one company in the AI space, but only at a meaningful discount to his value.
“That's why I said this is not an investment if you're thinking about five stocks to have in your portfolio. Right. But it's an investment I'm willing to consider at a discount on that 1,475,000,000,000 as one company in this space.”at 43:03
NVIDIA
Across the shows: 178 speakers in other episodes, 186 bullish and 171 bearish calls, past 30 days
Where everyone stands on NVIDIA
Aswath Damodaran · NYU Stern School of BusinessBearish
China scandal is the real risk to Nvidia; it may force giving up China sales and operating value regardless of CEO.
Damodaran calls the China issue more dangerous than the CEO question. He expects congressional hearings and a closer look at whether Nvidia evaded laws or restrictions. At the end, Nvidia may have to give up operating value to regain good standing. That could mean losing some or most of its existing China sales, which already exist under its current structure, and damaging its relationships with Asian partners, which would hit margins. Bringing Huang back would not make this go away and might make it worse. Market cap has dropped from about $4.5-4.8 trillion to about $3.5 trillion, and Damodaran thinks investors are right that this won't blow over. He would only see an overreaction if the China issue fizzles. The investment becomes a bet on how serious and how lasting the China problem is.
“And I think that at the end of the process, it might very well mean that NVIDIA has to give up some operating value to get back into the good graces.”at 4:57
Aswath Damodaran · NYU Stern School of BusinessNeutral
Replacing Jensen Huang as CEO alone would have a small value effect, roughly 3-5% short term.
Damodaran separates a pure CEO replacement from the China issue. A new CEO, probably from inside Nvidia, could take over while Huang stays on as a strategic adviser, and the second management layer can carry operations. He notes Huang's value lies in his culture of spotting big markets early (gaming, crypto, AI), which will be lost, but the effect is smaller now that Nvidia is a $4 trillion company rather than a $100 billion one. He puts a short-term value effect at about 3-5%. A long-term effect would be larger if revenue growth or margins fall, and he doubts Huang will return. Recovery from a scandal like this is hard. His own spreadsheet lets users flex revenue growth and margins to size a long-term impact.
“So let's put the China story to the side. If it's just a CEO replacement, the effect is going to be small.”at 7:10
Aswath Damodaran · NYU Stern School of BusinessBearish
Nvidia's 70% operating margin is unsustainable; pressure from TSMC and hyperscalers will compress it over time.
Nvidia designs chips, has TSMC make them, and sells them at an 800-900% markup to Meta, Alphabet and Microsoft. Typical chip companies earn about 40% operating margins, while Nvidia earns about 70%. That premium is possible only because hyperscalers are desperate to build data centers and have no alternative. Both ends are under strain. TSMC wants a bigger share, and Meta and Microsoft are building their own chips to avoid the premium. Damodaran says the question is not whether margins will fall but what will force them down sooner. He stops short of predicting a collapse. The margin decline is structural, so the 70% cannot persist. Huang's personal relationship with TSMC's management may also come under strain.
“So it's a model that's destined to see a margin decline over time. I'm not saying it's going to collapse.”at 15:12
Aswath Damodaran · NYU Stern School of BusinessBearish
Cheaper non-Nvidia chips running AI models would threaten Nvidia's business model.
Damodaran cites reporting that Anthropic got Claude running on traditional non-Nvidia chips, though he is unsure whether it appeared in a prospectus or was overstated. If AI can design or run itself on cheaper chips, Nvidia's pricing power and margins fall apart. He sees this as the main hidden threat under the 11x forward earnings multiple. Investors are holding their breath over what could cause the model to implode. Earnings look cheap only if that doesn't happen, so a risk discount has to be built into the valuation.
“That is a terrifying thought. Because if AI can become self perpetuating and it can replace NVIDIA chips with much cheaper chips, there goes your business model.”at 16:38
Aswath Damodaran · NYU Stern School of BusinessBullish
Nvidia is a buy at about $140 a share, despite uncertainty, if the business model keeps its power.
Damodaran has placed a limit buy on Nvidia at $140. He has sold all the shares he bought in 2018 over the past two years, so he holds none now. At that price he will accept the many unknowns, including the China risk, the CEO question and the chip-model threats. He believes the business model still has enough strength left for him to make his money back and more. He says the stock is too risky to be a third or half of any portfolio, so position size should be limited. Ed Elson argued the 11x forward earnings multiple looks cheap, and Damodaran agreed only if the business model holds up.
“But at 140, I'm willing to live with the uncertainties because I think there's enough power left in the business model for me to make my money plus more.”at 18:36
AI infrastructure & data centers
Where everyone stands on AI infrastructure & data centers
Aswath Damodaran · NYU Stern School of Business
A more hostile Congress will make data centers and AI more expensive to build.
Damodaran says the November midterm election produced a legislature much more hostile to both data centers and AI as a technology. Building data centers and compute capacity will become more expensive. That is a key input in his Anthropic model, because the more Anthropic must spend on compute to earn each dollar of revenue, the lower its value. He adjusted his model to reflect higher near-term build costs and a larger regulatory drag on margins.
“You have an election that's brought in a congress, a legislature that is going to be much more hostile to both data centers and to AI as a technology. It's going to get more expensive to build data centers.”at 22:01
Aswath Damodaran · NYU Stern School of Business
Compute power will become the scarce resource in AI, so locking up capacity now makes strategic sense.
Damodaran thinks compute will be the binding constraint in the AI business. Anthropic was far behind OpenAI on compute a year ago and now holds about a third of the compute and has moved ahead with long-term commitments. He reads those commitments as Anthropic's way of ensuring it can deliver growth. Even if demand is there, Anthropic may fail to deliver without reserved compute. The commitments carry risk, but he understands the strategic logic behind them.
“But I can see why they did it. In this business, the AI business, I think compute power is going to become the scarce resource.”at 31:49
US stock market outlook
Where everyone stands on US stock market outlook
Aswath Damodaran · NYU Stern School of Business
The index held up despite the tech selloff, showing the US economy has growth beyond AI.
The S&P 500 is up about 7% this year while AI and tech stocks have dropped by trillions of dollars. He had expected the market to be down about 10% if you had told him tech would fall this much. Investors rotated out of tech into other sectors, which absorbed the losses. The US has about 6,000 publicly traded companies spread across sectors, unlike concentrated markets where tech losses cannot be offset. He warns that this can only continue if tech bottoms out.
“And it shows you that The US economy has legs that go beyond AI.”at 48:49
Aswath Damodaran · NYU Stern School of Business
Index gains mask weak breadth; most stocks and investors will underperform this year.
A month ago, when the market was up 11-12%, 52% of stocks were still down for the year. The market is not one where stocks rise together. Returns depend heavily on sector and stock selection. He expects around 80% of investors to underperform the index and post negative returns, because many portfolios were heavy in the Mag Seven and tech, which had a damaging year.
“52% of stocks this year were still down for the year. It's not a market where stocks are all climbing.”at 56:56
US growth & recession risk
Where everyone stands on US growth & recession risk
Aswath Damodaran · NYU Stern School of Business
The AI boom has added roughly 1% a year to US real GDP growth for four years.
He cites an estimate that about one percentage point of annual real GDP growth over the last four years came from the AI boom. The spending on data centers and AI capex has supported the economy and kept it afloat. Other sectors have benefited because people have money to spend, which is why consumer staples and other areas have held up.
“It's estimated that one, you know, 1% of real GDP growth each year for the last four years has come from the AI boom.”at 51:38
Aswath Damodaran · NYU Stern School of Business
A substantial AI slowdown could trigger a recession and broad market correction in 2027.
AI is now the undercurrent driving both the economy and the market. If data center building, the supply chain serving it, and the products and services it enables slow down materially, the economy takes a hit. Households would lack income, which could push the economy into recession and cause a broad market correction. He sees this as the big question for 2027, because a slowdown in AI is likely, and its earnings and GDP boost would then fade.
“If the AI story really slows down... then I think there's an economic hit we're going to take... and that's going to cause a market correction.”at 58:11
AI: overall outlook
Where everyone stands on AI: overall outlook
Aswath Damodaran · NYU Stern School of Business
Only a small share of value created by AI is likely to accrue to AI companies.
Many claims will be made on the value that AI creates, from customers, workers, and consumers. Cures and productivity gains spread across society as consumer surplus. Lower-cost AI entrants can also cut margins for the incumbents they compete with, as with pharma. He therefore doubts that the AI companies capture most of the value. He says people make too quick a leap from value creation to the claim that AI firms will keep it.
“I'm not sure what percentage of the value created from AI actually accrues to the AI companies.”at 39:07
Healthcare system & insurers
Where everyone stands on Healthcare system & insurers
Aswath Damodaran · NYU Stern School of Business
Healthcare stocks' 18% gain may signal strain in the healthcare system.
Healthcare has risen about 18% this year. He treats this as a warning sign rather than good news. Healthcare is something people must spend on, and a rally of that size suggests problems in the healthcare system. He frames it as part of a broader healthcare problem in the US.
“I'm not sure that's good news that healthcare is up 18% because it suggests that there's fraying in the healthcare system.”at 51:14
AI bubble & tech valuations
Where everyone stands on AI bubble & tech valuations
Aswath Damodaran · NYU Stern School of Business
AI and tech stocks are now cheaper than in recent years, so investors should start buying.
He thinks the AI and tech correction is nearing its bottom. Waiting for the exact bottom usually causes investors to miss the rebound. Investors who avoided AI and tech for years because the Mag Seven looked expensive can now buy these companies at prices well below those of two to four years ago. He advises adding a part of the portfolio to tech and AI, not all of it, and choosing companies whose prices look reasonable.
“So if you've never invested in AI and tech because you've always felt the max seven was too expensive... Now is your chance to start adding those companies at prices much better than you'd have got two years ago”at 52:54
Earnings & corporate profits
Where everyone stands on Earnings & corporate profits
Aswath Damodaran · NYU Stern School of Business
S&P 500 earnings growth of about 56% in 2026 will not repeat in 2027 as AI benefits fade.
Aggregate S&P 500 earnings growth this year, about 56%, is the highest he has seen in 40 years, and much of it came from AI. Companies recognize AI capex spending as revenue and earnings. As spending slows, depreciation on existing AI capex rises and the revenue boost from new capex falls. Earnings growth then fades, which can lift price-earnings ratios even if prices are flat. He says next year's growth will not be 56%, and the open question is whether it is 4-5% or 15-20%.
“The question is, what does the 56% growth look like next year? It's definitely not gonna be 56%.”at 59:55
Across the shows
What other episodes Chorus covered said about the same companies in the 30 days up to this one.
Topics
- AI slowdown and recession risk
- AI stock valuations after selloff
- AI-driven GDP growth
- Anthropic IPO valuation
- CEO succession
- China export and scandal risk
- IPO governance structure
- Nvidia entry valuation
- S&P 500 earnings growth 2027
- chip margins and business model
- competitive pricing pressure
- compute commitments without earnings
- compute scarcity
- custom chip competition
- data center regulation and costs
- drug development push
- enterprise AI agents
- governance and strategic flexibility
- healthcare system strain
- market resilience amid tech selloff
- revenue growth deceleration
- revenue ramp needed
- stock breadth and dispersion
- value capture from AI
More from Prof G Markets
- Anthropic’s $1.5 Trillion IPO Is A Trap · November 19, 2026
- Nvidia Is Better Off Without Jensen Huang · November 18, 2026
- Could Jensen Huang Actually Return To Nvidia? · November 17, 2026
- Nvidia’s $2 Trillion Scandal Is About To Get Much Bigger · November 16, 2026
- Here’s What It Will Really Take To Fix America’s Economy — ft. Kathryn Anne Edwards · November 13, 2026
Ask your AI about this episode
Chorus gives Claude, ChatGPT and other assistants every view from this episode and every other show Chorus follows. Ask who agrees, who changed their mind, and what else was said about the same companies.
Chorus reports what speakers said. It isn't investment advice. How Chorus works · About Chorus