AI Faces Another Big Test 11/10/26

Anthropic

Across the shows: 110 speakers in other episodes, 147 bullish and 89 bearish calls, past 30 days

• Rohit Kulkarni · Burnham Park Capital MarketsBullish

  Anthropic expects compute costs to fall enough for margins above 35%, a strong long-term profit potential.

  From the IPO roadshow, Anthropic believes it can bring down the cost of compute enough for margins to exceed 35%. Its newer models are better, cheaper and drive more revenue, which supports that margin path. The margin trajectory is lumpy, but he calls it encouraging, a sign of long-term profit potential. That profile, combined with recent revenue momentum, is the main reason he sees the IPO as likely to succeed despite a cloudy AI stock backdrop.

  “One is this company thinks that they can get the cost of compute down so much so that their margins could be more than 35%. That's very encouraging.”at 35:54

• Rohit Kulkarni · Burnham Park Capital MarketsBullish

  Anthropic's October revenue jump and margin profile make a successful IPO likely despite cloudy AI stocks.

  Anthropic added $10B in annualized revenue in a single month, which was above his expectations, and that new-model-driven growth supports the case. He sees that October trend and the margin profile as good enough for a successful IPO. He acknowledges the backdrop for AI stocks is cloudy, so the strength of the company's own numbers is what carries the case.

  “in a pretty cloudy backdrop of AI stocks, but there is, this October trend and the margin profile, which I think should be good enough for this company to have a successful IPO.”at 36:30

• Rohit Kulkarni · Burnham Park Capital MarketsBullish

  Anthropic can fund its $900B in commitments if it keeps shipping pricier, better models every few months.

  Kulkarni argues that with about $200B in cash after the IPO, Anthropic should be well set over the next eighteen months. The model depends on a compounding cycle: new models built on AI, getting faster, cheaper and more efficient, which lets Anthropic keep charging a premium for newer models. That cycle is what makes the $900B in signed commitments serviceable. He flags the failure case: open models and large big-tech LLMs catching up and competing on a slightly lower price. Price competition was the issue in Q2 and part of Q3, but he sees no sign of it in October. He also notes that Anthropic's investor deck relies on four infrastructure suppliers (Amazon, Google, NVIDIA and AMD), so it is hedged to some extent.

  “I think, at least over the next eighteen months, with $200,000,000,000 in the bank, they should be pretty well set given the compounding nature of new models... So long as that cycle keeps flowing that they're able to get newer models and newer models charged at a premium, this entire economic model with them signing up $900,000,000,000 in commitments should be doable.”at 37:18

CoreWeave

Across the shows: 89 speakers in other episodes, 25 bullish and 97 bearish calls, past 30 days

• Brent Thill · JefferiesBullish

  CoreWeave's $132B backlog is many times its market cap; upside if it converts that backlog into margin-rich revenue.

  Backlog is $132B of signed, non-cancelable contracts from high-quality customers such as Meta and Microsoft, and it is many times CoreWeave's roughly $26B market cap. Revenue beat and margins came in well above expectations, and customers prepay 20–30% upfront, which eases funding strain. The key question is whether CoreWeave can convert that backlog into revenue, which requires building capacity. If it keeps converting at high margins, the stock has more upside. The bear case is not yet coming true, though many bears remain.

  “If they can show that they can convert and they can close that gap, then we think, again, the stock seems to have more upside if they can keep converting on that.”at 13:37

NVIDIA

Across the shows: 97 speakers in other episodes, 78 bullish and 50 bearish calls, past 30 days

• Dmitri Alperovitch · Silverado Policy AcceleratorBearish

  Jensen Huang cannot survive as Nvidia CEO; export-control violations and governance lapses are serious.

  Alperovitch says he sees no way Huang survives as CEO. An affair with an alleged Chinese intelligence officer is not illegal in itself, but violating US export-control laws by funnelling chips to listed Chinese entities is, with penalties of up to twenty years in prison. He expects more to emerge about what Huang knew about the Singapore shell company that funnelled chips to China. He argues the president's stance doesn't matter, since the president can pardon but cannot exempt anyone from the law. Huang could claim he was misled and blame someone inside Nvidia, but Nvidia has already been caught in two lies. Its spokesman denied the affair, and the board's Sunday statement prejudged an independent investigation by saying national security was not compromised, which the next day's New York Times and Bloomberg reporting contradicted. He also cites reports that Nvidia tried to influence the Commerce Department to shut down its investigation into the Megaspeed smuggling case.

  “I see no way in which Jensen can survive this as CEO of NVIDIA. ... In many ways, he is NVIDIA.”at 41:22

US rates & the Fed

• Stephen Parker · J.P. Morgan Private Bank

  Fed likely hikes once more in December, then one more in 2027 before stopping, as inflation eases.

  He expects inflation pressures to ease into next year. Brent is already down 20% this month on possible Middle East de-escalation, and a slowdown in AI capex would also take pressure off inflation. Financial conditions look okay on the headline, but excluding equities, bond yields make them restrictive, and he thinks the Fed will take that into account. Result: one hike in December, one more next year, then the Fed is done.

  “So I think the Fed is gonna take that into account. We think we get a hike in December and only one more for next year, and then we're done.”at 4:52

AI infrastructure & data centers

• Stephen Parker · J.P. Morgan Private Bank

  Hyperscaler AI capex will slow rather than stop, shifting JPMorgan toward hyperscalers and away from semiconductors.

  Approvals and regulation mean the 2027 AI capex is already locked in, and markets will start focusing on 2028. He expects capex to slow because of higher costs and more regulation, not to stop. He sees that as acceptable because hyperscalers were earlier criticized for burning free cash flow, and a slowdown could bring a nearer-term path to free cash flow positivity. That is why JPMorgan has moved its preference away from semiconductors toward hyperscalers. Upside from the infrastructure build-out is now seen as limited.

  “We think that the AI CapEx spend story is still very much in play. ... We don't think it's going to be about stopping this CapEx story. We think it's going be about slowing it down.”at 5:12

AI: overall outlook

• Brent Thill · Jefferies

  AI trade sentiment is the most bearish in years, a short-term rotation into safer names like Amazon and Microsoft.

  He describes an AI mini bear market in which investors are gravitating toward safer platforms like Amazon and Microsoft to hide in the short term. CoreWeave's good numbers did not lift the stock much because of the tape. Sustainability and infrastructure constraints such as building physical capacity fast enough remain the central worries. He says many bear theses are not coming true, but bears have not left the camp.

  “investors are incredibly bearish. I have not seen this magnitude of bearishness on the AI trade in a long time.”at 10:36

US growth & recession risk

• Robert Sockin · PGIM

  US recession risk raised to about 25%, as Fed hikes and AI-driven growth raise the chance something breaks.

  Recession risk was raised from a very low 10% earlier this year to around 25%. Each Fed hike raises the risk that something breaks, and concerns about AI are legitimate and material. Growth is heavily leveraged on the AI capex story and on wealth effects from higher equities for upper-income consumers. A real AI retrenchment could spread through the financial system in ways that are not yet visible, adding financial-stability risk. The outlook is a bimodal distribution, with overheating still the base case at about 40%.

  “So we raised recession risk from a very low 10% earlier this year to around 25%.”at 27:30

Oil & fuels

• Pippa Stevens · CNBC

  Refiners, not producers, drove energy's blowout quarters; upstream producers are keeping capex tightly controlled.

  Explaining why XLE fell less than crude after oil's late-October high, she says downstream operations have driven many of the blowout quarters, more than the upstream side. Producers are keeping capex under control, and many of the upstream gains already came earlier in the year. Crude has fallen about 22.5% from its recent high on diplomatic hopes, and Iraq is cutting prices. If prices stay in the $85–90 range, the open question is whether major E&P producers will raise drilling, since most of the extra output so far has come from private players.

  “it shows that the downstream division has really been driving a lot of those blowout quarters that we saw, not quite as much upstream side of things and this sense that producers are keeping their CapEx under control”at 30:39