CNBC Halftime Report · · 57 min
Brad Gerstner on NVIDIA, Anthropic IPO and the AI Trade
With Brad Gerstner, Jim Lebenthal, Joe Terranova, Sam Altman
Summary
Brad Gerstner of Altimeter argues Nvidia's 2027-2028 delivery schedule is locked in, but that CUDA's moat has been dented, as shown by Anthropic's weekend work optimizing AMD GPUs, and that the leadership succession must be handled free of political pressure. He says the AI trade is overdone with roughly 20% downside, and that the consensus 2027 data center buildout of 43 GW is too high, with his estimate near 30 GW due to regulatory and capital constraints. He is bullish on Anthropic's IPO, where Altimeter plans a $5 billion order, and on OpenAI as leading labs capture most of the AI value. Jim Lebenthal's stress test suggests Nvidia's China exposure is priced for a worse case than is realistic, while Jenny Harrington is bearish on Nvidia's leadership decision and bullish on Disney's theme parks and Cisco's networking franchise. Joe Terranova sees hedge funds reducing risk and quant models amplifying the chip sell-off.
Key takeaways
- Brad Gerstner (bullish on NVIDIA): Nvidia's 2027 and 2028 deliveries and fab schedules are locked in, so near-term execution is not a worry.
- Brad Gerstner (bullish on Anthropic): Anthropic is a cornerstone compounder; Altimeter plans a $5B IPO order, and pricing around $1.4–1.5T.
- Joe Terranova (uncertain on NVIDIA): Nvidia looks incredibly cheap, but quant selling and a complicated story keep pressure on it.
- Jim Lebenthal (bullish on NVIDIA): Nvidia's stock prices in a China-sales worst case far beyond what is realistic.
- Jim Lebenthal (bullish on Anthropic): Anthropic's IPO adds $60-80B of AI-trade capital and shows strong demand.
What was said, by company and theme
17 views from 3 speakers. Every quote is checked against the episode's transcript (how).
NVIDIA
Across the shows: 152 speakers in other episodes, 140 bullish and 117 bearish calls, past 30 days
Where everyone stands on NVIDIA
Brad Gerstner · Altimeter CapitalBullish
Nvidia's 2027 and 2028 deliveries and fab schedules are locked in, so near-term execution is not a worry.
Gerstner says that for 2027 and 2028 it is all baked: the company knows what it is delivering, the fab schedule and the delivery schedule are set. So he is not worried about execution over the next six quarters. He says the company's leadership is capable of making the trains run on time day to day. He is an Altimeter Nvidia shareholder, though he has cut his exposure from large to small this year.
“I said in that post, for 2027 and 2028, it's all baked, right? We know what, the company knows what it's delivering, the fab schedule's all baked, the delivery schedule's all baked, So that execution for the next six quarters, I'm not worried about.”at 11:29
Brad Gerstner · Altimeter CapitalUncertain
Nvidia's leadership gap is about long-term vision after Huang, not near-term execution; Kress is capable as interim.
Gerstner says Colette Kress is very capable of running the company day to day, but he does not think she has said she wants the job permanently. He calls Jensen Huang an extreme visionary with extraordinary channel relationships, and says the open question for a board is what vision takes Nvidia beyond the locked-in schedule. He also argues the board should not decide under political pressure from the White House, Congress and the Treasury Secretary, and should take time to gather the facts before choosing a successor.
“What I worry about if I were on the board is what's the vision to take us beyond that? And that's where Jensen had extraordinary vision and also extraordinary relationships in the channel to help deliver the vision.”at 11:29
Brad Gerstner · Altimeter CapitalUncertain
Claude-built AMD kernels dent Nvidia's CUDA moat, but CUDA is not Nvidia's only moat.
Gerstner cites a leak from the Anthropic investor day that Anthropic used its own Claude models to build GPU kernel integrations that optimize AMD GPUs over a weekend, work he says would have taken six to nine months before. He calls that a real penetration of the CUDA moat that people have discussed. He says it is not the only moat Nvidia has, pointing to the Amazon and Google chip alternatives as the wider competitive context. The net view is that the moat is weakened but not broken.
“So that definitely is a penetration of the CUDA moat, you know, that people had been talking about. But I don't think that was the only moat that Nvidia had.”at 23:33
Joe Terranova · Virtus Investment PartnersUncertain
Nvidia looks incredibly cheap, but quant selling and a complicated story keep pressure on it.
Terranova acknowledges Nvidia is incredibly cheap on valuation, but doubts the market is paying attention to that. He says the stock is in the hands of quantitative models that are amplifying the downside, and capital is flowing to the most expensive Magnificent Seven name, Apple. He sees the story as complicated, with the CFO's earnings call unlikely to resolve it, including the unanswered question of whether China received advanced AI chips.
“we all acknowledge it's incredibly cheap. But does the market care about that?”at 38:44
Jim Lebenthal · Cerity PartnersBullish
Nvidia's stock prices in a China-sales worst case far beyond what is realistic.
Lebenthal runs a worst-case stress test. Even if up to 20% of sales go to China, including indirect illicit sales, and treating all of that as profit, next year's net income would fall from about $450B to about $310B, on projected sales near $700B. He argues the stock, down roughly $1.5-2T, is pricing in a worst case far worse than any he can realistically construct. He is a long shareholder.
“Let's use the higher number of the range of how much sales go to China, including indirect illicit sales that has been estimated up to 20%.”at 40:34
Anthropic
Across the shows: 118 speakers in other episodes, 175 bullish and 90 bearish calls, past 30 days
Where everyone stands on Anthropic
Brad Gerstner · Altimeter CapitalBullish
Anthropic is a cornerstone compounder; Altimeter plans a $5B IPO order, and pricing around $1.4–1.5T.
Gerstner says Altimeter is placing a $5 billion order in the Anthropic IPO, more than five times the largest order it has ever placed in an IPO, and that it would become Altimeter's largest position if it gets its allocation. He calls Anthropic a cornerstone compounder and says it is where the AI trade is going. He says Anthropic has presented its financials well despite the Nvidia turmoil. He expects the IPO to price at about $1.4 to 1.5 trillion, and says pricing near $1.75 to 1.8 trillion would be risky.
“our order book on Anthropic, we're putting in for $5,000,000,000. That's more than 5x the largest order we've ever placed in an IPO. I think this is a cornerstone compounder for a long period of time.”at 25:45
Jim Lebenthal · Cerity PartnersBullish
Anthropic's IPO adds $60-80B of AI-trade capital and shows strong demand.
Lebenthal argues an Anthropic IPO this week would be a shot in the arm for the AI trade because it brings roughly $60-80 billion of new capital. He expects much of that to flow into the chip complex, including Nvidia and Broadcom. He also cites the deal being about six times oversubscribed as evidence of continued appetite for the AI trade despite the sell-off.
“Anthropic coming public, assuming it does this week, is a shot in the arm for the AI trade.”at 45:05
Meta Platforms
Across the shows: 67 speakers in other episodes, 61 bullish and 35 bearish calls, past 30 days
Where everyone stands on Meta Platforms
Brad Gerstner · Altimeter CapitalBullish
Zuckerberg's founder-led cuts and AI pivot transformed Meta, with the stock rising from $90 to $650.
Gerstner says he wrote an open letter to Mark Zuckerberg on October 22 urging him to cut back on Reality Labs and double or triple down on AI. He credits Zuckerberg with acting on that. Zuckerberg wrote a January letter saying the flatter, faster, leaner approach was better. Gerstner says the stock went from $90 to $650 and that Zuckerberg totally transformed the company around AI. He uses this as evidence that founders have the authority to make bold changes that corporate CEOs find hard to make.
“ultimately the stock went from $90 to $650 and look what he did. He totally transformed that company around AI.”at 19:40
OpenAI
Across the shows: 41 speakers in other episodes, 16 bullish and 35 bearish calls, past 30 days
Where everyone stands on OpenAI
Brad Gerstner · Altimeter CapitalBullish
OpenAI and Anthropic are pulling away from the field and will capture most of AI's economic value.
Gerstner compares the current moment to Google in 2005, after about 20 search engines were consolidated into one that captured nearly all the upside in search. He says Anthropic and OpenAI are now separating dramatically from the pack and will capture the lion's share of economic value. Altimeter is a major shareholder in both, and he wants to double and triple down on those bets. He acknowledges that the outcome still has to be proven.
“I think this is that 2005 moment where Anthropic and OpenAI are separating themselves dramatically from the pack. I think they will capture the lion's share of the economic value. We're major shareholders in both.”at 27:28
The Walt Disney Company
Across the shows: 6 speakers in other episodes, 5 bullish and 1 bearish calls, past 30 days
Where everyone stands on The Walt Disney Company
Jim Lebenthal · Cerity PartnersBullish
Disney's cost cuts reflect a new CEO running the business, not financial engineering.
Lebenthal argues Disney's cost reduction is a new CEO cutting fat and running the business smartly. He distinguishes that from financial engineering, which he defines as taking on debt to fix problems, buybacks, or off-balance-sheet debt. None of that is happening at Disney. He also says the company must show streaming profitability offsets the decline in legacy revenue and that its theme parks and cruise ships are performing well.
“I think that's a new CEO coming in and running the business and cutting out fat.”at 50:59
Palo Alto Networks
Across the shows: 5 speakers in other episodes, 4 bullish and 1 bearish calls, past 30 days
Where everyone stands on Palo Alto Networks
Joe Terranova · Virtus Investment PartnersUncertain
Palo Alto will beat with 33%+ revenue growth, but the stock reaction is the worry.
Terranova expects Palo Alto to beat, with revenue growth above 33%, triple-digit earnings growth, strong margins and good guidance. He is worried about how the market will react. The implied move is about 9% against an average move of about 5%, and he says the cyber trade must prove itself after cybersecurity stocks cushioned a depressed software market.
“I think you're going to get the beat. It's the reaction I'm worried about.”at 51:54
US–China AI & chip race
Where everyone stands on US–China AI & chip race
Brad Gerstner · Altimeter Capital
Chip export restrictions and three new bills are a headwind to the AI semiconductor trade.
Gerstner points to three bills introduced in the past week to further restrict the diffusion of American chip technology abroad, and to a new Commerce and Treasury framework issued that morning. The framework slows chip deliveries or requires more licenses for exports to Singapore, Malaysia and Indonesia. He argues that the existence of new rules suggests the old ones did not exist, and he thinks all of this has to be treated as a headwind for AI semiconductors. He says the market would weigh it against the sector.
“I think there have been in the last week three bills introduced to further restrict the diffusion of American technology around the world... All of that has to be perceived as a headwind to the AI semis trade.”at 22:10
Brad Gerstner · Altimeter Capital
The US is ahead of China in AI and extending its lead through export limits and slower model releases.
Gerstner says the US is ahead of China on AI and extending that lead. The first reason is that making it harder for China to distill US models slows Chinese models, which he says draw heavily on US capabilities. The second is that US labs are releasing models less quickly, and the gap between unreleased and released models is widening, which makes copying harder. He cites a new US framework that requires Commerce approval before a frontier model is released and adds liability and audit requirements. He says Chinese progress is being slowed by those measures.
“And so we are ahead and I think we're extending our lead. Now, why are we extending our lead? In part because we're making it harder for the Chinese to distill our models.”at 32:31
AI bubble & tech valuations
Where everyone stands on AI bubble & tech valuations
Brad Gerstner · Altimeter Capital
The AI trade is overdone with about 20% downside; Altimeter cut its exposure in response.
Gerstner says he told CNBC on May 28 and again on July 2 that the AI trade was overdone and that there was about 20% downside. He cites Nvidia, which he says is down 36% top to bottom this year after rising about 18x over three years, as evidence that a roughly 30% retracement from the top is in line with that call. He says Altimeter took profits and reduced its AI exposure, cutting from large to small, and that its call on the AI trade was not a surprise to him.
“I came on on May 28 on your show and I said, the AI trade is overdone. We pulled back on our exposures. We're taking profits because we think there's 20% downside in the AI trade.”at 21:04
AI infrastructure & data centers
Where everyone stands on AI infrastructure & data centers
Brad Gerstner · Altimeter Capital
Compute buildout for 2027 will land near 30 GW versus 43 GW consensus, because of power and capital limits.
Gerstner says the consensus 2027 forecast was 43 gigawatts of compute and that it was priced into semis, including TPUs, Trainium and custom Broadcom ASICs. He said publicly he would take the under and expects about 30 GW. He says token demand is going parabolic, with agents and OpenAI handling a large share of search volume, so demand is not the constraint. The gap comes from regulatory issues, including in Texas, and capital constraints. He says he expected buildout to come in 5 to 10 GW below forecast all summer.
“I didn't think we were going to be able to build what was then the consensus forecast for 2027, which was 43 gigawatts of compute. ... I think it's probably going to be closer to 30.”at 24:10
Open-source AI models
Where everyone stands on Open-source AI models
Brad Gerstner · Altimeter Capital
Open source will struggle to scale because frontier labs' cheap models undercut its main cost advantage.
Gerstner says he worries about open source's ability to scale and compete because of network effects and scale advantages. He says frontier-lab economics let them offer very cheap models such as Haiku and Luna to enterprises, which undermines the main reason people use open-source models, which is that they are less expensive. He says the competition is more nuanced than Jensen Huang leaving the field, and that open source still does well in some cases.
“I do worry about open source's ability to scale and compete for natural reasons network effects, scale advantages.”at 28:08
Market sentiment, positioning & volatility
Where everyone stands on Market sentiment, positioning & volatility
Joe Terranova · Virtus Investment Partners
Hedge funds are pulling back on risk as the year closes, suppressing year-end optimism.
Terranova says conversations with hedge fund contacts over several days show most are reducing risk. They are reacting to what they see in front of them and to the calendar closing on 2026. He says this dampens the bullish sentiment he had expected for a higher year-end and undercuts the narrative that the market is absorbing the chip news resiliently.
“most people are pulling back on risk because they're... What they are seeing in front of them and because the calendar is closing on 2026 rather quickly.”at 46:39
Across the shows
What other episodes Chorus covered said about the same companies in the 30 days up to this one.
- NVIDIA152 speakers · 140 bullish, 117 bearish calls
- Anthropic118 speakers · 175 bullish, 90 bearish calls
- Meta Platforms67 speakers · 61 bullish, 35 bearish calls
- OpenAI41 speakers · 16 bullish, 35 bearish calls
- The Walt Disney Company6 speakers · 5 bullish, 1 bearish calls
- Palo Alto Networks5 speakers · 4 bullish, 1 bearish calls
Topics
- AI lead over China
- AI reorientation under founder
- AI trade valuation and downside
- Anthropic IPO
- CUDA moat and competition
- China sales exposure
- chip export controls
- chip stock valuation
- cost cutting and management
- data center buildout
- earnings reaction
- hedge fund risk reduction
- leadership and ecosystem risk
- leadership transition
- near-term execution
- networking backbone
- open-source competitiveness
- theme park growth
- value capture by leading labs
More from CNBC Halftime Report
- NVIDIA Fallout, Chip Volatility and the Case for 60/40 · November 13, 2026
- NVIDIA Leadership Shakeup, AI Trade Fallout and OpenAI Valuation 11/12/26 · November 12, 2026
- AI Relief Rally, Nvidia Governance and CoreWeave Earnings · November 11, 2026
- NVIDIA Fallout, Anthropic IPO and Hot CPI Hit Markets · November 10, 2026
- NVIDIA Fallout and AI Trade Rotation · November 9, 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