2025-11-20 21:35:03
Well, they did it. I'm happy to report that Jensen Huang saw his shadow and so we will have at least another three months of AI euphoria. I'm "happy" about this because it means that NVIDIA's stock won't collapse when it opens today and thus risk bringing down the entire stock market which is so heavily wrapped around the Big Tech bet. At some point, this seems inevitable, but not today. Phew.
The reason why is because NVIDIA continues to bring in boatloads of money selling their AI wares. They even managed to defy the Law of Large Numbers this quarter and saw that mythical uptick in revenue growth – the first on a sequential basis in a couple of years. Even more impressively, they're projecting another uptick next quarter. No matter how you choose to look at this situation, it's insanely impressive growth.
At the same time, it's growth that is largely fueled by a handful of players. This makes sense, NVIDIA's Big Tech peers are really the only companies that can afford to spend the amounts of money that makes up most of what NVIDIA is now bringing in. This is obvious, but often overlooked: much of NVIDIA's revenue is coming from Big Tech.1
2025-11-20 04:14:08
It's November 22, 1996 and Steve Jobs has a cold. But unlike Frank Sinatra 30 years earlier, Jobs actually sits down for his interview,1 because it's the one year anniversary of the release of Toy Story, Pixar's first film released in collaboration with Disney. And now, to mark the 30th anniversary of the movie, the Steve Jobs Archive is sharing this never-before-seen footage.
Jobs is just 41 years old at the time but it has been 11 – eleven! – years since he was pushed out of Apple. That was the same year that he started NeXT, which he's still the CEO of in this video, not yet reunited with his first company (that would happen just weeks later). He's also been the CEO of Pixar for just over a decade here, having bought the business from George Lucas, out of Lucasfilm – for $10M. It took that entire decade to make their first feature film. But it worked. Boy did it:
Toy Story was the world’s first entirely computer-animated feature-length film. An instant hit with audiences and critics, it also transformed Pixar, which went public the week after its premiere. Buoyed by Toy Story’s success, Pixar’s stock price closed at nearly double its initial offering, giving it a market valuation of approximately $1.5 billion and marking the largest IPO of 1995. The following year, Toy Story was nominated for three Academy Awards en route to winning a Special Achievement Oscar in March.
The largest IPO of 1995 would be a seed round for an AI startup today – and in some cases, not even that outlandish of one.
The entire 22-minute video is obviously worth the watch. There are a lot of good nuggets about both Pixar and Disney, but the one part that really stood out to me was about how he thought about building teams – in particular at Pixar, which wasn't necessarily a world, filmmaking, that he was close to. But he did know technology and knew this would have to be a sort of hybrid of Silicon Valley and Hollywood – even though he also knew previous attempts to fuse those worlds never worked. The cultures were simply too different. Or were they...
When asked about how he's different than the studio heads of old, Jobs takes a moment to think about it – a very Jobs-ian move in such interviews – and he starts to talk through how they're building something different with Pixar. But as he talks through it, he seems to realize that actually, it's not that different from what he's done in technology when it comes to building teams:
"This is the same in technology as it is at Pixar and on the creative side. Which is: where you’ve got incredibly talented people that are also rare and in-demand, if you don’t treat them right, they can go get another job in 10 minutes. So this strange thing happens, which is sort of the hierarchy of power inverts. And the CEO is actually at the bottom [laughs]. So I sort of feel like I work for most of these people. Because they’re the ones that are doing all the brilliant work. And it’s the same in software. It’s the same thing. The best people are very hard to come by. And so it’s managements job to support them. Because they’re on the front lines doing the work. So that’s kind of how we run the place."
"I play the orchestra" anyone? But really, there are similarities to the answer he gave in a Q&A all the way back in 1983 about attracting and retaining talent at Apple. And obviously this will resonate with anyone building teams – in particular, perhaps those building teams working on AI at the moment. "They can go get another job in 10 minutes."
When pressed how he's been able to build such a culture to foster success:
“We do that by creating the right environment. A very rich environment. And removing obstacles out of their way. And putting together the right teams of people together. And getting the right projects. Keeping the quality levels very high, both in terms of people and strategy and projects. Those kinds of things.”
He keeps going:
“As an example our studio has grown tremendously this year. We started off at about 175 people and we’ll be at about 300 people by the end of this year. And one of our major challenges was not to lower the quality bar one iota. And we managed to accomplish that. We put together a very strong recruiting group. We did a lot of other things that I’d rather not talk about cause they’re competitive advantages. And we were able to hire you know, 125 A+ people this year.”
He played the orchestra, but he also assembled that orchestra.
It would be interesting to know what Jobs would have thought about our current Age of AI – in particular, AI as a tool to augment creative endeavors, such as filmmaking. Hollywood is obviously up-in-arms about it right now, but there was a time when digital animation was the new technology cast aside. The same has been true of basically all new technology.
While many would undoubtedly jump to say that Jobs might hate something as seemingly "soulless" as some of the creative output we get right now from AI, I'd like to imagine he would find a way to fuse the creative talent with the technical talent to make something magical. Something like Toy Story.2
One more thing: the other stand-out bit is when Jobs is explaining how Walt Disney realized that to make filmmaking work with this new technology at the time, animation, he would actually have to invert the way movies were typically made – something which Pixar had to do as well: "You have to edit your film before you make it."
M.G. Siegler
M.G. Siegler
M.G. Siegler
1 Issey Miyake shirt and all. ↩
2 Incidentally, we just showed our youngest part of Toy Story this past weekend as her very first movie. Have to gear her up for Toy Story 5 with it's antagonist... the um, tablet! ↩
2025-11-19 20:55:03
My god what a colossal waste of time. Here's Cecilia Kang for The New York Times:
Meta did not break the law when it bought its nascent rivals Instagram and WhatsApp, a federal judge said on Tuesday, handing a major win to the $1.51 trillion company and dealing a blow to the government’s efforts to rein in the power of tech giants.
Judge James E. Boasberg of the U.S. District Court of the District of Columbia said in an 89-page ruling that Meta did not create a monopoly in social networking through the acquisitions and that the market has continued to expand with rivals including TikTok and YouTube. The Federal Trade Commission had sued the company, accusing it of breaking antitrust law by acquiring Instagram and WhatsApp in a “buy or bury” strategy to cement its social networking dominance.
The F.T.C. “continues to insist that Meta competes with the same old rivals it has for the last decade, that the company holds a monopoly among that small set and that it maintained that monopoly through anticompetitive acquisitions,” Judge Boasberg said, adding that the agency needed to prove that argument. “The court’s verdict today determines that the F.T.C. has not done so.”
This case was so obviously going to fail from the get-go – and actually, it did fail from the get-go, but the judge allowed the FTC to rework it and refile it – that it should call into question the government's judgement and competence in going after any such cases.
Even a recent case in which they "won", against Google in Search dominance, they completely and utterly failed to secure remedies that will change much of anything. And they shouldn't have gotten any such remedies, because that entire premise is fatally flawed in the age of modern technology – and the Age of AI, where things are moving faster still, will only make such issues more obvious.
To put it simply and plainly: the government litigates antitrust in hindsight, but by the time they're able to bring these cases forward to trial, the technology at the center of these debates is already in the rearview mirror, making many of the arguments and debates moot. That was the issue with the Google remedies, and that was the problem with this Meta case. I made the longer version of this argument back in April's Antitrust's Hindsight Problem:
If anything, I suspect these cases – as well as the myriad other antitrust cases against Big Tech™ – will ultimately point to how ineffectual current antitrust law is for our modern age (apologies, Lina Khan). But that will likely only be obvious in hindsight, which is interesting because that itself is a big part of the problem with these cases: that they're being litigated with hindsight...
I mean, that's sort of true about *every such case*. You have to go after what is viewed to be problematic elements of the past, obviously. We don't have Minority Report-style pre-crime – yet. But these Big Tech™ antitrust cases in particular seem to look back upon the past with a different set of glasses than how things were viewed in real time. Nowhere is that more evident than in Meta's ongoing antitrust case. The FTC wants the company to sell-off Instagram and WhatsApp, without really acknowledging that the common consensus at the time – in particular with the former – was that the deals were silly ones in which Facebook, as the company was then known, was drastically overpaying.
Essentially, the government is going after Meta for doing deals that are only deemed savvy and anti-competitive in hindsight. It's sort of the opposite of pre-crime! It's making things crimes in the present that were not crimes in the past.
Worse, had the judge ripped WhatsApp and Instagram away from Meta over a decade after those deals closed, it would have certainly had a chilling effect on basically any deal – in a time when most deals are still mainly frozen because all of these companies still have PTSD in dealing with the wrath of Lina Khan. Again, she was largely ineffectual in action, as we're seeing now with cases playing out after her tenure is over, but the government also made getting deals through such a pain in the ass that it literally wasn't worth the time. Remember when they tried to block Meta from buying a VR fitness app? Yeah...1 Talk about waste of time...
So instead, we saw the rise of "hackqusitions". Deals so convoluted and so clearly set up to get around such nonsense (in many cases, less about the fear of being blocked but the fear of how long such deals would take to get through – basically death in the Age of AI).
Speaking of death, I also might point out that if the government had forced Meta to spin out Instagram, it very well could have killed the company. That sounds like hyperbole, but I'm not sure that it is – by most accounts, Instagram is now the main driver of Meta's all-important ad revenue, more so than Facebook itself. Certainly, it would have ensured that Meta had no shot to compete in AI because they wouldn't have the money to do so. Mark Zuckerberg's case to would-be hires that Meta is better positioned for the long term in AI than, say, OpenAI, because they have the profits to pour back into the projects would have been all-but destroyed.
Hard to debate "buy, build, or bury" when you can't buy, build, or bury...2
So does this mean M&A meat is back on the menu, boys? Maybe, but probably not. Again, "hackquisitions" work so much faster than traditional M&A since they're not subject to, well, much of anything aside from an agreement between two parties. So I wouldn't be shocked to see them continue. Unless the government wakes up to that reality. Which they might... in 2045.
That's around when the government might also realize that Meta's real competition here was TikTok and YouTube, you morons. Any child would know that. I mean that very literally! It took one simple, quick experiment to highlight this fact to the judge in this case, as Casey Newton rightly zeroes in on in writing up the verdict today on Platformer. The government? They were too busy on something called 'MeWe', apparently. Maybe we should look into that. Of course, even those undoubtedly won't be the competition by then because,3 well, see: above.
It's like trying to play Whac-A-Mole where the moles are high on amphetamines. And your mallet weighs a thousand pounds.
It's almost as if the government has spent the past 25 years congratulating themselves for the Microsoft antitrust case that they also actually lost in the end, that they're blinded to the reality of how markets actually work in the modern age. And that's really going to screw them in the Age of AI.
With this Meta case, at least we'll always have the discovery documents and testimony here. That was fun, albeit not for Meta. The same will likely be true for Apple and Amazon with their own antitrust cases still making their way to trial. The Apple case in particular also seems awfully weak, so that mixed with the fact that Apple hates to disclose anything about their business (not to mention public embarrassment) might lead to some sort of settlement before that happens. We'll see.
I'll close with my closing from my post about antitrust back in April (a longer excerpt, since it was paywalled):
Now, that is the one part where the government may have a point and a role here. Rather than worrying about the dominance of Google Search, if they were focused on making sure Google doesn't leverage their market position to control the future of AI in an anticompetitive way, that's the correct framing here. And you're starting to hear them talk in such a way. The problem goes back to pre-crime. No one can know for sure that AI will be the key to dominating the future right now. Many believe that, and for good reason, I think. But it's trying to litigate the future.
So yeah, that puts the government in a tricky spot between the past and the future. Still, that future should clearly be the focus versus the past. And so if there's a way to keep an eye on ensuring that a market position isn't leveraged in an anticompetitive way – *while still allowing Google to compete* – that's the angle.
Admittedly, it's a tough needle to thread.
But even without that, it feels like the market is moving in ways right now that are naturally starting to erode what all of these antitrust cases are focused on. The humorously defined social media dominance that Meta enjoys seemingly matters less by the day. And same with Google when it comes to Search. As noted, the ad case against Google is probably the strongest, but it's perhaps also under attack as AI starts to change the way the web actually works – or, at least, is served.
Meanwhile, the cases against Apple and Amazon are yet to begin. But there are already clear issues with the Apple case as well, largely in trying to cram their market position with the iPhone into older antitrust definitions around market dominance. There as well, there's perhaps a case to be made against the App Store in particular as a form of distribution, but it's going to require the DoJ lawyers to yes, think different. About the future, not about the past.
Good luck.
M.G. Siegler
M.G. Siegler
M.G. Siegler
M.G. Siegler
M.G. Siegler
M.G. Siegler
M.G. Siegler
1 Shoutout though for successfully torpedoing the Amazon/iRobot deal – though Europe largely gets the credit for that particular one. We almost fell victim to Big Tech owning Big Vacuum! Instead, iRobot is about to die, having had to lay off nearly everyone. What a win for everyone! ↩
2 Here's where I'll again note the irony that TikTok rose to power largely on the back of Meta's ad platform. It allowed Bytedance to turn the acquired startup Musical.ly into the hit product that Vine should have been, had Twitter not so pathetically fumbled that bag! And now TikTok's success effectively saved Meta here. Funny that. ↩
3 It's still not entirely clear that TikTok will even exist next year, let alone in a decade or two, thanks to the continued bungling of the "deal" to sell the service. ↩
2025-11-19 06:04:52
When I was a kid, I distinctly remember the phrase "collect them all" as a marketing slogan used to upsell toys and other collectables. As a completist, this always worked on me – honestly, still does. It also sure seems to work on Big Tech when it comes to AI.
With the news today that Microsoft and NVIDIA are now making investments in Anthropic – naturally alongside a pledge from Anthropic to spend at least $30B on Azure, because you can't have a tit without a tat – I'm honestly sort of at a loss for how to frame this. Everything is just so tangled and interwoven at this point. I mean, this is a company which owns 27% of OpenAI investing in their main rival. And, to be fair, it comes after OpenAI signed deals with all of Microsoft's main rivals in the cloud. So instead, I'll just assume from now on that all bets are off because all bets are on.
Still, I feel the need to list out some of this conflicted mess just for my own sanity and so I can keep referencing and checking it as I write about tangential topics. Really, it's just an expansion of a list I made in July when noting just how hedged Big Tech was now thanks to their massive investments in these AI companies.
This is just the latest such deal, but obviously won't be the last.
Amazon
Investor in: Anthropic • Databricks • Hugging Face • Scale AI
Investment from: N/A
Other deals: OpenAI
AMD
Investor in: Cohere • Hugging Face • Scale AI • Thinking Machine Lab • World Labs
Investment from: OpenAI
Other deals:
Anthropic
Investor in: N/A
Investment from: Amazon • Google • Microsoft • NVIDIA
Other deals: Apple
Apple
Investor in: N/A
Investment from: N/A
Other deals: Anthropic • Google • OpenAI
Cisco
Investor in: Anthropic • Cohere • CoreWeave • Mistral • Scale AI • Thinking Machine Lab
Investment from: N/A
Other deals: Microsoft
CoreWeave
Investor in: N/A
Investment from: Microsoft • NVIDIA
Other deals: OpenAI
Databricks
Investor in: N/A
Investment from: Amazon • Microsoft • NVIDIA • Salesforce
Other deals: Google
Google
Investor in: Anthropic • Hugging Face • Safe Superintelligence
Investment from: N/A
Other deals: Broadcom • Databricks • NVIDIA • OpenAI • Thinking Machine Lab
Hugging Face
Investor in: N/A
Investment from: Amazon • AMD • Google • IBM • Intel • NVIDIA • Salesforce
Other deals:
Intel
Investor in: Hugging Face
Investment from: NVIDIA • SoftBank • US Government
Other deals: Amazon • Google • Microsoft
Meta
Investor in: Databricks • Safe Superintelligence • Scale AI
Investment from: N/A
Other deals: Black Forest Labs • Broadcom • Midjourney • NVIDIA
Microsoft
Investor in: Anthropic • Databricks • Mistral • OpenAI
Investment from: N/A
Other deals: NVIDIA
Mistral
Investor in: N/A
Investment from: ASML • Cisco • IBM • Microsoft • NVIDIA • Salesforce • Samsung
Other deals:
NVIDIA
Investor in: Anthropic • Cohere • CoreWeave • Mistral • Nokia • OpenAI • Perplexity • Safe Superintelligence • Scale AI • Thinking Machine Lab • xAI
Investment from: SoftBank (just exited)
Other deals: Literally everyone, except maybe Apple
OpenAI
Investor in: AMD
Investment from: Microsoft • NVIDIA • Oracle • SoftBank
Other deals: Almost everyone, except Anthropic, Meta, and xAI
Oracle
Investor in: Ampere • Cohere • OpenAI
Investment from: N/A
Other deals: NVIDIA • SoftBank
Salesforce
Investor in: Anthropic • Cohere • Databricks • Hugging Face • Mistral • World Labs
Investment from: N/A
Other deals:
Scale AI
Investor in: N/A
Investment from: Amazon • AMD • Cisco • Meta • NVIDIA
Deals with:
SoftBank
Investor in: Ampere • Intel • NVIDIA (just exited) • OpenAI
Investment from: N/A
Deals with: Oracle
xAI
Investor in: N/A
Investment from: AMD • NVIDIA
Deals with: Black Forest Labs • Microsoft • Oracle
So there you go, while NVIDIA is obviously out there like Thanos collecting Infinity Stones, others are more subtle in their strategic bets. Special shout-out to Cisco, which kept unexpectedly coming up as I was putting together the list.
It also sure feels like with this latest news we're mere weeks away from Google and/or Amazon investing in OpenAI, completing the ouroboros. Apple? All totally cool, right? Since this isn't "zero sum" per Nadella...1
Anyway, I'm off to play Clue. Or Pokemon.
I can't believe they don't break into a rendition of "Kumbaya" to close...
M.G. Siegler
M.G. Siegler
M.G. Siegler
1 Narrator: as it turns out, it was not "totally cool" in the end, as many of these conflicted sides and companies would fight tooth and nail for market share. ↩
2025-11-18 22:49:57
Tomorrow is once again the big day. Will Jensen Huang see his shadow? NVIDIA's results will determine if utter AI exuberance gets to continue for another few months or, in a reversal with Punxsutawney Phil's model, if AI Winter is coming.
I make this joke every quarter, but it's also more of an acute risk every quarter. NVIDIA famously became the first $5T company in history a few weeks ago. Alongside that milestone, it surpassed accounting for 8% of the S&P 500 – the largest single-stock weight in the history of the index. Increasingly, as goes NVIDIA, so goes the stock market. So yeah, Jensen seeing his shadow is important.
As to if he will or not, there are conflicting signs. By all accounts, AI spend – and in particular, Big Tech CapEx spend – continues to surge. Given just how much of that spend runs directly to and through NVIDIA, it seems almost impossible to believe they could disappoint on earnings. At the same time, the Law of Large Numbers exists in business for a reason. The bigger you are, the harder it is to grow bigger still at the same rate of growth. And, of course, NVIDIA's rate has been falling for a while now.
In the chart below, you can plainly see the brief rise and fast fall of the crypto boomlet and then the utterly insane rise of AI starting in mid-2023.
Again, this chart isn't surprising, it's that Law of Large Numbers at work. If anything, it's insanely impressive that NVIDIA is still seeing growth north of 50% given the scale they're at and the previous year's comps. But at some point, barring some new breakthrough – which obviously is certainly possible and perhaps even plausible in this world! – NVIDIA will fall back to more "normal" rates of quarterly revenue growth. While they're all obviously different businesses (with some overlap), the rest of Big Tech – the largest companies in the world – have yearly growth rates ranging from 8% (Apple) to Meta (26%).
NVIDIA itself is guiding towards around 54% growth this quarter, while Wall Street consensus has them in the 56% to 60% range. The latter would be rather incredible because it would be the first sequential uptick in that growth rate since the end of 2023/early 2024, nearly two full years ago. Regardless, this is undoubtedly going to be the first $50B revenue quarter in the history of the company.
And even if NVIDIA was able to pull off the accelerated growth, the real key, as always, will be what they're guiding towards next quarter (current Wall Street bets are in the 56% growth rate range there as well). And there are a lot of constantly-in-motion pieces there between the shipment of new chips and the China situation (if and when NVIDIA is able to do business in China again, the numbers will clearly surge and should easily beat any estimates since they're no longer accounting for any China sales in their models for the AI chip business, and it used to be a pretty good part of the business!).
There's also new wild cards potentially in play in NVIDIA's quarterly numbers thanks to their increasing number (and increasing in size) stakes in other AI companies – most notably, OpenAI itself. How will they account for the swings in value there? Will their accountants determine that they need to account for some of that startup's sizable losses on their own books as Microsoft does? Presumably not since the stake is still very small (~2%) and only growing to a potentially sizable one over time. So instead they'll have to mark to market which will impact their bottom line (though presumably not yet in a major way beyond a uptick in their previous smaller holding due to the new round of financing). And what about all the other stakes?
Also, what about those two key customers that we so key to last quarter?
Almost all of the above lays out a picture where NVIDIA's business is still booming for the foreseeable future. But that doesn't mean there aren't signs of concern – in the form of a few big bets against them (or no longer for them), in particular.
The biggest of these, of course, is the short position perhaps the most famous short-seller in history, Michael Burry, disclosed a couple weeks ago. To be clear, he's far more short on, and presumably bearish on, Palantir. But still, before he closed up his shop (or at least deregistered it with the SEC), Burry made two final short bets and one of those is on the most valuable company in the world.
Burry followed up with some hints that he's in particular worried about the depreciation schedule that the tech companies are using for their data center assets – the biggest aspect of that is clearly the lifespan of NVIDIA chips. He suggests that this element alone is a form of financial manipulation that helped inflate the current AI bubble.
At the same time, if that's his rationale – or even just a key part of his bubble thesis – why would he short Palantir so heavily versus, say, the "neocloud" companies? Their business models are heavily predicated around this depreciation element. As I wrote in (the paid version of) the newsletter last week:
I'm honestly not quite sure what to make of Michael Burry's move to shutter his fund (or at least deregistering it with the SEC) right after making his short bet on Palantir (and to a lesser extent NVIDIA). His stated reason that the value of securities is not "in sync with the markets" sort of implies he's giving up on the short game because the market is too irrational, and that this may be his swan song bet? But it's almost too obvious of one to make given where Palantir is valued relative to earnings. And ditto with NVIDIA given that it's, well, the most valuable company in the world on the back of AI. Why not short neoclouds or something else if you really believe AI is a bubble? Especially with his follow-up talk about potential nefarious business going on with the way companies are handling depreciation of these chips – this is clearly going to be an issue for someone at some point given how all over the board the timelines are: some 3 years, some 4, some 5, CoreWeave um, 6! And it is interesting that these shifts have happened while NVIDIA has also shifted from a two year release cadence to a one year (which Satya Nadella mentioned almost in passing on The Dwarkesh Podcast this week when talking about why they paused some data center spend earlier this year – yeah...) Anyway, Palantir is all kinds of complicated because of the government contracts if nothing else. And, of course, Burry was just wrong with his infamous last "Sell" tweet. So what's the game afoot here? I guess we'll know more on November 25! Thanksgiving stuffing coming early this year?
But Burry also isn't the only one making a bet here. Or perhaps more accurately, there are others pulling their bets on NVIDIA.
SoftBank made headlines last week with the news that they were selling their entire multi-billion-dollar position in NVIDIA. They were quick to note that this was simply to get cash to be able to fulfill their commitments to fund OpenAI, but it's still a signal against NVIDIA. If SoftBank was super bullish about the company's growth prospects from here – as they clearly are on OpenAI – presumably they would have figured out a way to sell other assets and/or take on more debt to make those payments. It's not stated but certainly implied that they were okay selling their NVIDIA position right now at these prices.
And that's especially eye-opening since one of Masa Son's greatest regrets that he has often talked about in recent years was exiting the NVIDIA stake he previously held. Of course, that's also easy and obvious to say when you had a holding you sold for $4B (in 2019) that had you held would have been worth around $250B these days... Still, it's an interesting signal!
As is the move by Peter Thiel's hedge fund, Thiel Macro, to sell off their entire NVIDIA stake, as Bloomberg reported yesterday. Granted, the stake was "only" worth about $100M, but it was still one of the core holdings of the fund alongside Apple, Microsoft, and Tesla. They're holding on to those other holdings...
Even if the size of the sale wasn't going to move the stock, the news, especially in concert with the other items above, builds a narrative of negativity around NVIDIA at the moment. That might be fair or not (again, Burry's bet seemed more of a throw-in and SoftBank does need the cash), but it's out there.1 A big bear, circling the world's most valuable company...
The reality of the situation remains that while NVIDIA's numbers this quarter (and projections for next quarter) can certainly tank (or boost) the entire stock market given the sheer size and weight of the company, they're unlikely to lead to a full on bubble burst. If their numbers were ever bad enough, they could certainly kickstart such a sell-off (again, vice versa in the other direction if they're good enough), but bubbles are built on the back of irrationality.
I might argue that this particular bubble, the AI Bubble, is actually more rational than previous bubbles (and certainly the players involved are making that case). But that doesn't mean it's any less of a bubble. And bubbles always end. And that ending is even more likely to be for an irrational reason. That is, fear.
If and when something triggers a sell-off, what continues to fuel it won't be NVIDIA's numbers, it will be panic that the whole game is up and that the bets are now literally off. And that will feed back into the system, causing money to pull back which in turn will give companies less money to spend, which in turn will send less money to NVIDIA. And they'll be hit extra hard because of how entangled they are in the whole ecosystem at the moment.
I'm not saying this is about to happen. But I'm saying it will happen, at some point. Let's see if we see that shadow...
Update November 19, 2025: The numbers are out and NVIDIA has once again beaten across the board – including with that (slight) reacceleration of y/y revenue growth: 62%. Impressive. Shadow seen! 3 more months of AI!
M.G. Siegler
M.G. Siegler
M.G. Siegler
1 It's more tangential, but I might also just note the disclosure revealing Berkshire Hathaway's large new position in Alphabet (now they're 10th largest holding). That's not necessarily a bet against NVIDIA, but you certainly might view it as a bet on Google being one of the key victors of the AI race. And if that's true, a large part of it will likely be the success of their TPU chips vs. NVIDIA's chips... ↩
2025-11-18 03:18:21
Sometimes, things come in waves. Like in 1997 when we got not one, but two disaster movies about volcanoes in Dante's Peak and yes, Volcano. Or the following year when we got two movies about asteroids coming to destroy the world in Armageddon and Deep Impact. Now we seem to be having such a moment around the formation of companies to build "World Models" it seems.
Just last week, I wrote about the notion that Meta may soon find themselves in a battle with their (soon to be) former Chief AI Scientist if and when Yann LeCun breaks away to form a new startup focused on building World Models. I noted in passing that there were a number of other startups already working in the general space, but then came news today that another name is entering the arena, and it's a big one; an asteroid, as it were: Jeff Bezos.