Everything About AI
NextFin News - Nvidia Chief Executive Officer Jensen Huang dismissed the latest AI stock-market rout as a passing episode, telling an interviewer in Berlin that the technology's long-term impact will outweigh investors' near-term fears even as shares of the world's most valuable company trade more than 10% below their early-October peak.
"All of this is going to pass, and the reason for that is very simple. It's going to be transformative. It's going to impact in a very positive way every single industry and every single country. There's no question it will continue, and we're at the beginning of that."
Huang made the remarks during an interview with Emily Chang at an event in Berlin. His confidence stands in sharp contrast to the mood on trading desks, where the AI trade has become the market's most crowded position and the first to be cut when conviction wavers.
The comments arrive at a moment of maximum doubt for the AI trade. Nvidia shares closed Friday at $214.13, down 1.12% on the day and 10.5% below the $239.24 closing high set on October 6. The pullback has been broad and bruising: on the worst sessions, semiconductor peers including Micron, AMD, and Qualcomm have fallen by more than 9%, and memory-chip makers — the best-performing corner of the S&P 500 earlier this year — have surrendered much of their gains. The S&P 500 and Nasdaq posted back-to-back losses as the AI trade took the hit, a reminder that when the market's most crowded position wobbles, everything correlated to it wobbles with it.
The sell-off is not about one company. It is about a question the market is only now asking out loud: after three years of spending more on AI data centers, chips, and energy than it cost to build the U.S. interstate highway system over four decades, who actually earns the money back?
What Is Behind the AI Rout
The October weakness is a cluster of disappointments rather than a single shock, and that diffusion is precisely what makes it hard to dismiss. Dell gave a third-quarter outlook below Wall Street's estimates, a reminder that the server makers buying the chips are themselves facing margin pressure. A new report said Alibaba is testing its own AI-inference chip to compete with Nvidia and AMD inside China, where Nvidia already recorded no sales of its H20 China chip in the most recent quarter while awaiting clarity on export regulations. Meta Platforms is weighing a fresh equity sale to fund its AI buildout, following a similar move by Alphabet earlier in the week. Each headline alone is survivable; together, they read like a pattern.
Underneath those headlines sits a bigger anxiety: the return on investment. Amazon, Microsoft, Alphabet, and Meta are collectively on pace to spend roughly $757 billion on capital expenditures in 2026, according to Goldman Sachs, with much of that directed toward AI infrastructure. That spending wave has been Nvidia's revenue engine — data-center revenue alone reached $89.02 billion in the latest reported quarter, up 117% year over year. But the investment thesis requires the cloud providers to eventually monetize the capacity they are buying, and that is where the confidence is thinning. The AI boom has become a feedback loop in which infrastructure spending encourages even more infrastructure spending, and the market is beginning to wonder whether the loop is anchored to end-user demand or to its own momentum.
The market has begun to price that doubt. Nvidia's forward price-to-earnings ratio has compressed to roughly 23x, the lowest multiple among the Magnificent Seven AI names, even as the company delivered record fiscal second-quarter revenue of $96.2 billion — more than double the prior year — and guided the current quarter toward $108 billion. In other words, the stock is being re-rated not because the growth has broken, but because investors are no longer willing to pay the same premium for growth whose payoff date keeps moving. A company trading at 23 times forward earnings while growing revenue at triple-digit rates is not expensive by any historical measure of technology; what has changed is the certainty attached to those earnings.
There is also the question of circularity. Microsoft, OpenAI, Anthropic, and Nvidia have been funding one another through a web of equity stakes and purchase commitments. These arrangements can accelerate a buildout, but they also create skewed incentives: a company that has one of its suppliers as a major shareholder may keep buying its products whether or not the purchase makes commercial sense. If demand for AI fails to match today's expectations, those arrangements magnify losses rather than cushion them.
Cyclical Panic, Structural Shift — Which One Is This?
The central question for investors is whether this is a cyclical fluctuation that will mean-revert or a structural shift that will not. The answer is: both are happening at once, and they point in opposite directions across time horizons. Getting this wrong flips the conclusion, so the two forces have to be separated rather than blended.
On the cyclical side, the evidence for a mean-reverting panic is strong. Nvidia's 10.5% drawdown from its October 6 peak is modest relative to the stock's own history — this is a security whose two-year average closing price is $171.61 with a standard deviation of $34.96, and which has returned 56.80% over two years on an annualized basis of 25.22%. A single-digit pullback in a stock with that volatility profile is the normal texture of ownership, not a regime break. The triggers of this particular rout — a soft server-maker guide, a competitive-chip report, a funding headline — are the classic short-term noise of a momentum-heavy sector. History offers a longer comfort: the dot-com buildout of the late 1990s, the smartphone buildout of the late 2000s, and the cloud buildout of the 2010s each produced multiple 20%-plus drawdowns in the infrastructure winners before the trend resumed. The pattern is not reassuring because it is pleasant; it is reassuring because it is repeatable.
But the structural read deserves equal weight. The AI buildout is not merely a capex cycle; it is a re-platforming of computing itself. Huang has argued that AI will need "1,000 times more" energy than the world currently has, a statement that frames the opportunity in infrastructure terms measured in decades, not quarters. Every industry and every country will have to rebuild their computing stack around accelerated computing — a structural shift in the definition of the total addressable market that does not reverse because one quarter of hyperscaler spending looks lumpy. The companies buying the chips are not speculators; they are the largest, most cash-generative technology firms in history, and they are treating AI capacity as a strategic necessity rather than a discretionary experiment. That kind of demand does not vanish on a headline.
The tension between the two is the whole story. In the short run, the market is right to demand evidence of returns; in the long run, Huang is right that the destination has not changed. The mistake investors make is applying a cyclical verdict to a structural trend, or a structural valuation to a cyclical timing problem. The first error produces premature exits; the second produces painful holding periods. Both are expensive.
The Second-Order Question Nobody Is Asking
The first-order effect of the rout is obvious: chip stocks fall when hyperscalers look like they might slow spending. The second-order effect is subtler, and it cuts the other way. If the AI buildout slows because the economics are not working, the pain concentrates at the infrastructure layer — Nvidia, AMD, memory makers, and the data-center real-estate trusts. But if the buildout continues because the economics are working, the value migrates to the application layer: the software companies and enterprises that actually monetize the intelligence. Either way, the current rout is a signal that capital is beginning to rotate from "who sells the picks and shovels" to "who finds the gold."
That rotation is already visible. Software stocks have rallied on the view that they are oversold relative to the chip names, and companies once perceived as AI laggards, such as Apple and Alphabet, have rebounded while the pure-play infrastructure names have given back ground. The market is not abandoning AI; it is repricing which layer of the stack captures the profit. This is the same rotation that played out in the late 1990s, when the early winners of the internet buildout — the router and server makers — gave way to the application companies that figured out how to charge users. The lesson is not that infrastructure investors lose; it is that the margin of safety migrates as a cycle matures.
There is also a third-order expectation gap. The consensus embedded in Nvidia's 23x forward multiple assumes continued acceleration in data-center demand. If that growth merely decelerates rather than collapses, the stock can still underperform — because the expectation was for acceleration, not merely for growth. This is the trap of a company that has trained investors to expect perfection: good news that is less good than expected trades like bad news. Nvidia's own guidance for the current quarter, around $108 billion, is a number so large that beating it requires not just execution but the continued absence of any friction — from export rules to power constraints to customer budget cycles.
The Strongest Case Against Huang
The bear case is not that AI is a fraud. It is that the financing model is. The strongest counter-thesis, anchored in the $757 billion capex figure from Goldman Sachs and the circular-deal structure documented by market analysts, is that the AI buildout is being funded by a feedback loop in which infrastructure spending encourages more infrastructure spending, detached from end-user demand. If cloud providers cannot turn GPU capacity into profitable services, the spending stops abruptly — and because the commitments are front-loaded while the revenues are back-loaded, the stop is sudden, not gradual.
This counter-thesis attacks the core of Huang's statement. "All of this is going to pass" is true only if the demand side eventually catches the supply side. If it does not, the turmoil is not a passing episode; it is the beginning of a multi-year digestion period, the way the fiber-optic buildout of the late 1990s was followed by a decade of excess capacity that crushed returns for the infrastructure investors who bought at the peak. The analogy is not perfect — today's buyers are far more capitalized than the leveraged startups of the dot-com era — but the mechanism is the same: when capacity arrives faster than the revenue it can generate, the price of that capacity falls until the marginal builder quits.
The falsifying signal is concrete: watch hyperscaler capex guidance over the next two quarters. If Amazon, Microsoft, Alphabet, and Meta collectively cut their 2027 AI infrastructure commitments by more than 15% while AI revenue growth at those same companies stays below 30%, the structural-demand thesis is wrong and this rout is the first chapter, not a blip. Conversely, if capex holds and AI revenue accelerates, the current weakness is the cyclical panic Huang describes. There is no middle ground that resolves the tension — the numbers will force a verdict.
What Comes Next
The near-term path runs through Nvidia's next earnings report, due November 17, and the company's GTC Berlin keynote on October 21. Investors will be looking less at the headline revenue number — which is likely to be large — and more at three things: the trajectory of data-center growth, the margin profile as the company navigates China restrictions, and any commentary on customer capex discipline. Management's own forecast of roughly $108 billion for the current quarter sets a bar that leaves little room for a guide-down; anything less will be read as evidence that the feedback loop is slowing.
The scenarios are clear:
- Base case: capex guidance holds, data-center revenue continues to grow at a high rate, and the stock stabilizes in the low $200s while the market digests the valuation reset.
- Upside case: AI revenue at the hyperscalers accelerates, circular-deal concerns fade, and Nvidia re-rates back toward its prior multiple as the rotation back into infrastructure begins.
- Downside case: capex commitments are cut, China restrictions tighten further, and the 23x forward multiple compresses toward the broader semiconductor average.
For investors, the practical implication is asymmetry rather than direction. The beneficiaries of a structural outcome are the companies with pricing power at whatever layer of the stack captures the profit — today that is still the infrastructure layer, but the margin of safety has narrowed. The exposed are the pure-play capex beneficiaries whose valuations assume uninterrupted acceleration. The forward look splits by horizon. In the short term, sentiment and liquidity dominate, and the stock can remain volatile regardless of fundamentals. In the medium term, earnings and capex guidance will decide the direction. In the long term, the structural question — whether AI re-platforms the global economy — will be answered by productivity data, not stock prices.
Jensen Huang's confidence rests on a bet that history will judge kindly: that every transformative technology looks like a bubble on the way up, and that the investors who confuse the two pay the price. The counter-bet is equally historical: that not every boom survives contact with its own financing. The signal that decides between them is not a CEO's conviction. It is whether the companies writing the checks can turn GPU cycles into profits before the credit cycle turns.
Explore more exclusive insights at nextfin.ai.
