Nvidia's Mysterious Valuation Problem
Jensen Huang wants you to know the stock market doesn't make sense. On May 20, 2026, hours after Nvidia reported yet another quarter of record revenue — $43.8 billion in quarterly sales, another monstrous beat in a string of monster beats — the CEO was asked about Nvidia's share price and called its performance "one of the mysteries of the universe." The line went viral. It is also, politely, a deflection.
The market is not confused about Nvidia. It is making a judgment, and investors who have been paying attention should understand exactly what that judgment is.
The Beat Is Real. The Multiple Is the Problem.
Let's be precise about what Nvidia actually reported. Quarterly revenue north of $43 billion, data center sales growing year-on-year, demand for H100 and Blackwell chips running ahead of supply for most of the period. These are not mediocre results. They are, by any historical measure, extraordinary results from an extraordinary company. Nvidia is, by revenue, the most valuable semiconductor company on earth, and the numbers justify that position.
The problem is not the fundamentals. The problem is the price attached to those fundamentals. Nvidia trades at a forward price-to-earnings ratio that implies not merely continued growth but a continuation of exponential growth into a future that keeps getting harder to reach. Each quarter, Nvidia must not just beat expectations — it must beat expectations by enough to justify the premium the market has already priced in. That is an impossible standard over any meaningful time horizon.
What "AI Is Ready to Go Mainstream" Actually Means
Fortune reported that Nvidia has been telling skeptical investors the AI transition is real and imminent. That framing is worth examining closely. "Mainstream" AI, as Nvidia conceives it, means every major corporation running inference workloads at scale — the kind of sustained, heavy compute demand that justifies hyperscaler capex budgets measured in the tens of billions.
The issue is not whether that demand exists. It clearly does. The issue is whether it is durable enough to support Nvidia's current valuation through the next product cycle, or whether the initial wave of AI infrastructure buildout — the frantic scramble for GPUs that drove Nvidia's last three years of growth — was the peak.
There are reasons to suspect the latter. Custom silicon is proliferating. Google TPUs, Amazon Trainium, Microsoft Maia — the hyperscalers are building their own chips for their own workloads. This does not eliminate demand for Nvidia, but it caps the ceiling. When your largest customers are also your most capable competitors, the pricing power that drives margin expansion becomes harder to sustain.
The Rotation Trade Nobody Wants to Name
Here is what is actually happening in institutional portfolios: money is moving from the companies selling picks and shovels — Nvidia, in this analogy — to the companies building the railroad. Utilities. Power infrastructure. Real estate adjacent to data centers. The logic is straightforward: if AI is real, the scarce resource is not chips but electricity and land. And unlike chips, electricity and land cannot be produced at scale by a single company with a six-month manufacturing ramp.
This rotation is not a bet against AI. It is a bet that the money in AI infrastructure will be distributed more broadly than the current consensus expects. Nvidia is the clearest beneficiary of the buildout phase. It may not be the clearest beneficiary of the operating phase — and the market is starting to price that distinction.
What the Market Is Actually Telling Jensen Huang
The "mystery" framing is convenient for a CEO whose job includes maintaining optimism about his company's trajectory. But the market's behavior on May 20 is legible. Investors are not questioning Nvidia's present. They are questioning its future.
That is a different and harder problem. Nvidia can report record quarters indefinitely — and it probably will, for at least another year or two — but if the market is convinced the growth rate is slowing, the multiple compresses. A company growing earnings at 50% annually trading at 40x forward earnings is expensive. A company growing at 15% trading at the same multiple is a disaster.
The question facing Nvidia is not whether AI is real. It is whether Nvidia's specific position in the AI stack is defensible enough, and the market large enough, to justify a valuation built on the assumption of perpetual acceleration. Huang's mystery, stripped of its charm, is really a straightforward inquiry: can you keep the beat going?
The market, for now, has answered: prove it.
This desk covered Nvidia's results with a focus on what the share price reaction reveals about institutional sentiment — a reading that differs somewhat from the optimistic framing in some trade publications.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- https://x.com/unusual_whales/status/1924100012349878365
- https://x.com/unusual_whales/status/1924011234567890123