CNBC Closing Bell · · 48 min
Nvidia Slumps, Anthropic IPO Looms and Markets Test Key Levels
With Dan Niles, John Kolovos, Saira Malik, Saket Kalia, Satya Nadella, Willie Lee
ChorusBring the market's conversation to your AI. Learn more
Summary
CNBC's Closing Bell weighed a slump in Nvidia and tech against the coming Anthropic IPO. Sarah Malik argued that 2027 earnings growth, led by non-tech sectors, should support equities unless the Fed tightens too hard, and that the Fed should hike gradually given the difficulty of separating temporary war-driven inflation from chronic inflation. Dan Niles was bearish on frontier AI lab economics, citing slowing revenue growth and widening operating losses at OpenAI and Anthropic, falling token prices, and a sector-wide semiconductor drawdown of about 35%. John Kolovos saw a head-and-shoulders top on the S&P 500 and a breakdown in tech, while Willie Lee argued private-market value is shifting to closed-source frontier labs such as Anthropic and OpenAI. The episode also covered Palo Alto Networks' strong next-gen security ARR, AI-driven patch-related outages, and a weak US housing market.
Key takeaways
- Saira Malik (bearish on NVIDIA): Nvidia's strong numbers can't offset slowing customer capex growth and the loss of Asia, so the stock lags.
- Dan Niles (bearish on Anthropic): Combined OpenAI and Anthropic growth is slowing and losses are widening; the valuation looks hard to justify.
- Saket Kalia (bullish on Palo Alto Networks): Palo Alto's next-gen security ARR beat expectations by a wide margin, showing demand for its security platform.
- John Kolovos on US stock market outlook: S&P 500 is forming a head-and-shoulders top; a move down to about 6,300-6,600 is possible.
- Willie Lee (bullish on Anthropic): Public investors show strong demand for Anthropic's shares even at the expected IPO valuation.
What was said, by company and theme
29 views from 8 speakers. Every quote is checked against the episode's transcript (how).
Anthropic
Across the shows: 132 speakers in other episodes, 225 bullish and 110 bearish calls, past 30 days
Where everyone stands on Anthropic
Saira Malik · NuveenUncertain
Anthropic is not the focus here; Malik only notes tech and IPO risk.
placeholder
“returns on these mega IPOs tends to be mixed.”at 3:10
Dan Niles · Niles Investment ManagementBearish
Combined OpenAI and Anthropic growth is slowing and losses are widening; the valuation looks hard to justify.
Niles says combined sequential revenue growth for OpenAI and Anthropic fell from about 75% in Q2 to about 47% in Q3. Combined operating losses widened from negative $7.9bn to negative $8.4bn as revenue rose, because of price competition between the two and from open-source models. Their operating margins fell from 19% to 15%, excluding items. He says a roughly $1.5tn valuation for Anthropic is about 28x sales, and the market is becoming more selective and won't give the AI labs the benefit of the doubt when losses rise.
“growth, if you combine both OpenAI and Anthropic together, are slowing. It's gone from 75% growth sequentially when you add the two together in q two to about 47% in q three.”at 14:50
Willie Lee · Neostellar CapitalBullish
Public investors show strong demand for Anthropic's shares even at the expected IPO valuation.
Lee runs a publicly traded venture capital firm whose thesis is that much value is now accruing in private markets. He points to the expected Anthropic pricing as evidence that public-market demand for these names is extreme, even at very high valuations. He expects companies like SpaceX, Anthropic and OpenAI to keep growing above and beyond their current scale, so he sees room to grow even though the profile is very different from earlier IPOs.
“there is extreme amounts of demands within the public markets for some of these types of names, even at these valuations.”at 41:37
Willie Lee · Neostellar CapitalBullish
AI spending is rotating from infrastructure toward Anthropic and closed frontier labs, which will capture value.
Lee sees a rotation away from AI infrastructure, which he links to news about Nvidia and Jensen Huang's turnover. He argues the money is moving toward Anthropic and direct LLM opportunities. Jensen was a strong champion of open source, and Lee thinks that without him pushing open source, power will accrue to closed-source frontier companies such as Anthropic and OpenAI. He says the IPO shows these next kingmakers can capture that value, even though Nvidia has reaffirmed its support for open source.
“And I think that where it's going and seems like it's going is it's moving towards Anthropic and some of these direct LLM opportunities.”at 42:55
Palo Alto Networks
Across the shows: 7 speakers in other episodes, 5 bullish and 1 bearish calls, past 30 days
Where everyone stands on Palo Alto Networks
Saket Kalia · BarclaysBullish
Palo Alto's next-gen security ARR beat expectations by a wide margin, showing demand for its security platform.
Kalia says the key metric, next-gen security ARR, came in at $630m against his estimate of about $450m, which is a bigger beat than in prior quarters. He is unsure how much is organic versus acquired. He says Palo Alto and CrowdStrike are the two closest things to true security platforms, and security buyers in a tough environment are moving toward platforms, which helps explain the demand.
“There you had a very strong result, $630,000,000. We were at about $450,000,000. That's I think a bigger beat than we've seen in prior quarters.”at 24:03
Saket Kalia · BarclaysBullish
Applied Compute deal lets Palo Alto tap AI budgets, which grow faster than security budgets.
Applied Compute helps enterprises build their own models more securely. Kalia says Palo Alto can use its security leadership to win a larger share of enterprise AI budgets. He thinks those budgets are probably growing faster than overall security budgets, so the deal gives Palo Alto leverage on both the security and AI sides.
“I actually think that Palo can leverage their leadership and security to actually start accessing the AI budgets even more, which are probably growing faster than overall security budgets.”at 26:09
NVIDIA
Across the shows: 170 speakers in other episodes, 185 bullish and 155 bearish calls, past 30 days
Where everyone stands on NVIDIA
Saira Malik · NuveenBearish
Nvidia's strong numbers can't offset slowing customer capex growth and the loss of Asia, so the stock lags.
Nvidia's results looked strong, but investors are pricing in the slowing rate of change in capex growth from AI customers. Nvidia also has to exclude Asia from its numbers, which made the result less positive than expected. Those two factors explain the weak reaction, and Malik expects the tech leadership to keep facing turmoil.
“even though the numbers look strong, people are starting to price out the rate of change of CapEx growth, which is slowing. Also, Nvidia particularly needs to take Asia out of its number.”at 3:07
OpenAI
Across the shows: 42 speakers in other episodes, 20 bullish and 36 bearish calls, past 30 days
Where everyone stands on OpenAI
Willie Lee · Neostellar CapitalBullish
Anthropic's IPO valuation gives credibility to a higher future valuation for OpenAI.
Lee's firm, which is publicly traded and holds OpenAI in its portfolio, argues that Anthropic being able to list at this valuation lends credibility to where OpenAI could be valued in the future. He ties this to the view that the leading closed-source model companies are accruing the value as AI infrastructure loses its open-source champion.
“So Anthropic being able to go out at this valuation gives a lot of credence to, I think, where OpenAI could go out in the future.”at 42:34
US stock market outlook
Where everyone stands on US stock market outlook
Saira Malik · Nuveen
Most 2027 S&P 500 earnings growth should come from non-tech sectors like industrials and financials.
Consensus 2027 S&P 500 earnings growth is not concentrated in tech. Up to three quarters of it could come from non-tech sectors such as industrials and financials. US GDP growth above 3% supports cyclical sectors that depend on the economy. Malik uses this to argue for owning more than tech.
“They are not totally concentrated in tech stocks. Up to three quarters of the earnings growth could come from non technology sectors. So areas like industrials, financials.”at 6:09
Saira Malik · Nuveen
Bull market can run into 2027 if earnings grow double digits; multiples are not stretched.
Malik notes that a fifth year of a bull market has usually looked good. The key is earnings growth. Multiples are not expensive, so inflated valuations are not the main worry. If the S&P can deliver double-digit earnings growth again next year, which she thinks is plausible, equities should do well. The threat is a Fed that tightens too much and slows the economy, which would bring a recession or bear market.
“usually, actually, a fifth year looks good... if we can put up double digit earnings growth again next year, and so far it looks like we might be able to, then that's gonna be great for equities.”at 6:41
Dan Niles · Niles Investment Management
The market is becoming more selective and cash flow now matters much more than before.
Niles argues the market is more discerning and selective about AI companies. Cash flow matters far more than it used to. He contrasts Meta, which has a large cash-generating social media business, with Anthropic, which doesn't. Operating losses at OpenAI and Anthropic are rising while revenue grows, which is the kind of thing investors now penalise.
“this market's getting a lot more discerning and selective around this, and we've seen cash flow matter a whole lot more than it has in the past.”at 16:01
John Kolovos · Macro Risk Advisors
S&P 500 is forming a head-and-shoulders top; a move down to about 6,300-6,600 is possible.
Kolovos sees a classic head-and-shoulders top forming on the S&P 500, with the left shoulder in June and the head in September. An oversold rally would create the right shoulder, and then the index could fall to about 6,300-6,600. His base case is a drop to about 7,000, which most people already cite. The pattern is distributive, so the market needs to prove itself. He says sellers are in control.
“then ultimately that creates a move down to about [6,600] to 6,300.”at 35:49
John Kolovos · Macro Risk Advisors
A sustained break below July lows in XLK implies about 15% more downside; treat tech as a bear market.
Semis have already topped, and the broader tech sector is showing a similar breakdown. A sustained break below the July lows in XLK would imply another 15% drawdown. Kolovos says the sentiment on tech is bad even though clients believe valuations are good, and low-beta names are outperforming within tech. He advises treating tech as if it is in a bear market.
“A sustained break underneath the July lows for the XLK would imply about another 15% drawdown from that breakdown point.”at 36:31
John Kolovos · Macro Risk Advisors
Short-term bearish on the S&P; breadth lagging with low-vol leadership resembles past market tops.
Kolovos says low-beta stocks are already outperforming within tech, so investors should be defensive in tech and the broader market. He links this to his short-term bearish view on the S&P 500. In his view, breadth leading on low volatility is a bad combination for the market. He says this happened near the 2021-2022 peaks and in 2018, which he calls an analog many have been discussing. He adds that animal spirits need to return, and that whether the top pattern resolves to the downside is still open.
“And that also leads into why I'm also short term bearish on the S and P.”at 37:18
Cybersecurity
Where everyone stands on Cybersecurity
Saket Kalia · Barclays
Frontier AI has made bad actors more productive; the threat environment is much more dangerous than six months ago.
Kalia says frontier models help good actors do great things, but they also make bad actors more productive. Compared with six or nine months ago, he argues the threat environment is far more dangerous. He links this to higher demand for security products, which he says is part of what is showing up in Palo Alto's results.
“it's also unfortunately made bad actors more productive... I would argue it's head and shoulders much more dangerous than it was before.”at 24:57
Eamon Javers · CNBC
Recent outages likely come from rushed AI-accelerated patching, not hostile attacks.
Javers reports a working theory, shared by Democrats and Republicans, that AI is speeding up discovery of holes in big systems. Defenders have to patch those holes on an aggressive timetable, racing attackers who use AI too. The patches are causing trouble because they are rushed or create unintended effects elsewhere. He expects things to keep breaking until AI defence catches up with AI attack and the two reach a rough equilibrium.
“the patches themselves that are actually causing some of the trouble here because they are being done too quickly or they're setting off unintended consequences elsewhere in these systems.”at 29:01
Satya Nadella · Microsoft
AI bug discovery is driving patch churn that causes instability on older infrastructure.
Nadella explains recent tech outages with two forces. First, AI has helped find and fix more software bugs, but each fix has to be patched in, and that patch velocity creates instability. He says the outages look like old infrastructure running new patches. Second, he says a wave of cyber activity is happening at the same time, which he says has not yet hit Microsoft in the same way.
“First is the patch velocity because the thing that AI has done is is allowed us to find these bugs and fix these bugs.”at 39:25
US rates & the Fed
Where everyone stands on US rates & the Fed
Saira Malik · Nuveen
The biggest 2027 risk is the Fed tightening too quickly and crushing earnings growth.
Malik sees the main risk for 2027 as the Fed tightening too quickly. That would hit earnings growth, the main driver of the market's rise. Markets have not relied on multiple expansion this year, so earnings are the key support. She also says the Fed faces an inflation puzzle: it is hard to separate how much inflation comes from the war and higher oil prices, which may be one-time, from chronic inflation in shelter and airfares. That is why she favours raising rates gradually.
“I think the biggest risk for 2027 is whether the Fed tightens too quickly... That could destroy the number one thing that's driving these markets higher, which is earnings growth”at 3:07
Saira Malik · Nuveen
Fed will raise rates 25 basis points in December; consensus expects the same.
Malik's base case is a 25bp hike in December, and she expects consensus to stay there. She argues that the Fed's hawkish comments from Governor Waller, whether they mean a 50bp move in December or 25bp followed by multiple hikes in 2027, may not differ much in practice. Inflation remains well above target, and the Fed is struggling to tell temporary inflation from chronic inflation, so it should move gradually.
“Well, our view is that the Fed raises 25 basis points in December. I think consensus will stay there.”at 4:42
AI chips & semiconductors
Where everyone stands on AI chips & semiconductors
Dan Niles · Niles Investment Management
Memory makers' long-term supply agreements will be torn up as big customers renegotiate.
Niles pushes back on the argument that memory makers are protected by long-term supply agreements through calendar 2030. He expects big customers to come back and renegotiate those contracts, which would undercut the supply-agreement safety cushion for memory producers.
“those long term supply agreements will get torn up because their big customers come back and renegotiate them.”at 18:19
Dan Niles · Niles Investment Management
Semis are not game over; drawdowns could be deep like 1997-98, but the year can still finish up.
The semiconductor index is down about 35% from its highs, and Nvidia is only a bit worse, so the selloff is sector-wide. Niles argues the market may be repeating 1997-98, when the semi index had 40% drawdowns yet both years finished higher. He says drawdowns could be more severe than what has happened so far, and that the sector could still rise later as the agentic AI cycle plays out.
“I think it's nineteen ninety seven, nineteen ninety eight when you had 40% drawdowns in the semiconductor index... Both of those years finished the year up.”at 20:07
International equities
Where everyone stands on International equities
Saira Malik · Nuveen
Developed Europe, emerging markets and non-tech S&P stocks should outperform; keep tech but don't rely on it.
Malik says investors should look beyond technology for 2027. Developed Europe, emerging markets and non-tech S&P 500 stocks are performing well. Investors should keep some tech but build balanced portfolios rather than rely on tech as their main bet, since tech has led at the expense of everything else in past cycles and that pattern is unlikely to repeat.
“you need to keep looking outside of technology. You want to keep some tech in your portfolio, but acknowledge that developed... Europe and emerging markets, as well as non tech in the S and P 500 is performing well”at 3:54
AI bubble & tech valuations
Where everyone stands on AI bubble & tech valuations
Saira Malik · Nuveen
Tech stocks are not incredibly expensive after the selloff, though they should not be the single bet.
After the tech reset, Malik argues tech stocks are not incredibly expensive, so keeping a stake is reasonable. But she thinks tech should not be an investor's single bet. Past periods when tech led at the expense of everything else are unlikely to repeat, given the broader earnings growth outside tech.
“Well, I admit that tech stocks actually are not incredibly expensive.”at 5:50
US consumer
Where everyone stands on US consumer
Saira Malik · Nuveen
Consumers keep spending despite gloomy sentiment, so retail sales show resilience.
Malik says consumer confidence surveys make the consumer look grumpy, but people keep spending. Retail sales and other data show this. The consumer drives the economy, and that spending is what would support earnings growth if the Fed doesn't tighten too much.
“consumer confidence often makes the consumer look grumpy, but the consumer keeps spending. We see that with retail sales and other areas.”at 6:41
AI pricing & unit economics
Where everyone stands on AI pricing & unit economics
Dan Niles · Niles Investment Management
Agentic AI needs 10-100x more tokens to offset a 60% drop in price per token since May.
Agentic AI is brand new and could take 10 to 100 times more tokens than earlier AI uses. Price per token has fallen 60% since May. That price drop is part of why growth is decelerating and losses are growing at OpenAI and Anthropic. Niles asks whether token volume growth can offset the lower prices. He treats this as the key question for whether the AI trade can continue.
“That does take 10 to a 100 times more tokens. So the question is, can that growth offset the fact that price per token has gone down 60% from May”at 19:47
AI regulation & safety
Where everyone stands on AI regulation & safety
Eamon Javers · CNBC
Pre-testing AI models can't eliminate exploit risk, since any model that codes can read existing code.
Javers responds to the idea of pre-testing AI models before release. He says any model that can write code can also be used to understand existing code, which is how attackers find weaknesses. So pre-testing won't remove the possibility of AI-enabled exploits, and the defence has to be about the pace of patching and alerts. Federal bodies such as CISA already act as clearinghouses.
“any model that can code can also be used to understand existing code and so therefore you're never going to be able to sort of eradicate this possibility.”at 31:37
Commodities (broad)
Where everyone stands on Commodities (broad)
John Kolovos · Macro Risk Advisors
Commodities face about 8% drawdown risk; correction within a secular bull, driven by growth concerns.
Kolovos reads the Bloomberg Commodity Index chart as evolving into a proper ABC-type correction. He calls it a correction within a broader secular bull market for commodities. He expects a bounce off the 200-day moving average and a relief rally into December that could help risk sentiment. He thinks a failure to hold would mean a lower low, and puts aggregate commodity drawdown risk at about 8% from here. The decline is led by oil, while copper, palladium and other commodities have rolled over, which he reads as a growth issue. Financial conditions are tight and tightening, and the dollar is making new highs. He says commodities are giving a bearish signal until silver starts to outperform gold.
“I would say commodities in aggregate have about an 8% drawdown risk from here.”at 38:23
Frontier AI labs & model race
Where everyone stands on Frontier AI labs & model race
Willie Lee · Neostellar Capital
The LLM market will support multiple winners, not winner-take-all, as users prefer different models.
Lee says many people expected winner-take-all in the LLM space, but that has clearly not happened. OpenAI is moving into consumer personal assistants, and Anthropic is building tools for knowledge workers. He argues people have different preferences for model personality and for how models are wrapped in tools to solve problems, so there is clearly room for more than one player.
“There's clearly different preferences on how people work, and there's clearly room for more than one player here.”at 44:23
US housing
Where everyone stands on US housing
Dan Olek · CNBC
Housing weakness persists; sales and builder sentiment are at record lows, and mortgage rates must fall to unlock.
Olek cites October pending home sales falling 4.7% month over month and 16.7% year over year, the lowest in the index's 25-year history. He notes two consecutive months of record-low contracts, which the Realtors' chief economist says could push annual home sales to the lowest since the mid-1990s. Housing permits missed badly even as starts beat, and homebuilder sentiment fell to a 14-year low. He says the outlook does not look like it will improve. Rates fell during October but moved back up, and he says they need to fall from the seven-percent range to a six-handle to unlock sales, as happened earlier this year before rates rose above 7% in September.
“To unlock, you've got to get back to a six handle. You've got to get out of the sevens and into the six.”at 46:47
Across the shows
What other episodes Chorus covered said about the same companies in the 30 days up to this one.
- NVIDIA170 speakers · 185 bullish, 155 bearish calls
- Anthropic132 speakers · 225 bullish, 110 bearish calls
- OpenAI42 speakers · 20 bullish, 36 bearish calls
- Palo Alto Networks7 speakers · 5 bullish, 1 bearish calls
Topics
- AI budget capture
- AI capex growth
- Fed rate path
- Fed tightening risk
- IPO demand for AI names
- LLM market competition
- S&P 500 earnings breadth
- S&P 500 technical top
- bull market durability
- cash flow focus
- commodity correction
- consumer spending resilience
- cyber threat level
- frontier lab growth and losses
- housing demand and mortgage rates
- index breadth and market tops
- late-stage valuations
- memory supply agreements
- next-gen security ARR
- non-US equity performance
- outages from accelerated patching
- patching and cyber threats
- placeholder
- pre-deployment model testing
- semiconductor drawdown
- tech sector breakdown
- tech stock valuations
- token prices vs token volume
- value capture vs AI infrastructure
More from CNBC Closing Bell
- Anthropic’s Record IPO, AI Selloff and Market Rout · November 20, 2026
- Anthropic IPO, AI Trade Reset and Market Sell-Off on Closing Bell · November 19, 2026
- Anthropic IPO, NVIDIA CEO Search and AI Trade · November 18, 2026
- NVIDIA Earnings Beat, SEC Probe and AI Trade Reset · November 17, 2026
- NVIDIA Earnings Stakes, AI Infrastructure and Tech Outages · November 17, 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