The AI Trade Has a Nvidia Problem
Nvidia reported another quarter of record results on 21 May. Shares fell anyway.
That is the headline for anyone watching the AI trade right now. The numbers were strong by any conventional measure — another surge in data-centre revenue, guidance that beat the analyst consensus, margins that would make most companies weep with envy. And yet the market responded with what amounts to a shrug: after-hours trading showed the stock moving lower, investors apparently deciding that even another set of extraordinary figures was not enough to justify where the shares had been priced.
The gap between Nvidia's operational performance and the market's verdict is the most revealing thing in technology right now. It is not that the company is struggling. It is that the market has begun to question whether the company's next chapter can possibly match the story it has already told.
The results themselves warranted closer attention than the reaction suggested. Nvidia's data-centre segment — which now effectively is the business, given how the gaming and professional graphics arms have faded in relative importance — posted another period of explosive growth. Demand for the GPUs that power large language model training and inference remains robust. The company's pricing power persists. Gross margins stayed at levels that most semiconductor manufacturers would consider a fantasy. In any other context, this would be celebrated as a blowout quarter.
The problem is that the bar has been set somewhere else entirely. Markets have spent the past two years treating AI infrastructure spending as the defining secular growth story of the decade. Nvidia sits at the centre of that bet. But the question animating investors now is not whether the buildout is happening — clearly it is — but whether the spending can sustain itself, and whether Nvidia's position is as unassailable as the valuation implied.
There are reasons to wonder. The hyperscalers — Microsoft, Google, Amazon, Meta — are all reporting record capital expenditure. They are buying GPUs by the tray. But at some point, the accounting has to work. AI services need to generate revenue that justifies the infrastructure cost. The enterprise software revenue attributable to AI features has so far been more modest than the infrastructure spend would suggest. That gap is where the uncertainty lives.
There is also the competitive picture. Nvidia's dominance is real, but the ecosystem around it is shifting. Custom silicon — the chips that Alphabet's Google, Amazon, and others have built for their own workloads — is gaining share in the data-centre mix. That does not threaten Nvidia's core position overnight, but it does suggest that the gravitational pull of the market's expectations may need to be recalibrated. A company that once could count on a near-monopoly in cutting-edge training hardware is now managing a more complex competitive environment while the market watches every margin compression for signs of stress.
The valuation question is the hardest to answer. Nvidia trades at levels that imply not just current dominance but a specific vision of future growth — one in which AI demand remains robust, competitive threats remain manageable, and the gross margin profile holds. That vision may be correct. The results from this quarter suggest it is not obviously wrong. But when a stock falls on record results, it is telling you that the bar has moved. The market is no longer willing to extend credit for the story alone. It wants the numbers to keep arriving, and it wants them to arrive faster than they already are.
The irony is that Nvidia remains a remarkable business in a category almost by itself. The company has executed with precision, built a software ecosystem that creates genuine switching costs, and captured a moment in technology history that may not repeat. But markets are not in the business of rewarding the past. They price the future, and right now the future is asking harder questions of the AI trade than it has at any point in the past three years.
The consequences extend beyond Nvidia. If the market begins to pressure the name that has anchored the entire technology sector's narrative — if even Nvidia's results cannot satisfy — then the question of whether AI spending can be justified by commercial returns becomes unavoidable for every other company in the space. That conversation is coming. Thursday's after-hours move suggests it may have already started.
This publication's wire sources covered the Nvidia results as a straightforward earnings story. The market-reaction angle — the disconnect between performance and investor sentiment — received less attention in the initial framing.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- https://t.me/BBCWorldoffl/38471
- https://t.me/france24_en/28512