For most of its modern life, NVIDIA traded at prices that made cautious investors wince. At its April 2023 peak, according to FullRatio’s compiled filing data, the market was paying nearly 139 times earnings for a chip company that had not yet reported a single quarter of AI-boom revenue. That was the mood at full fever. Today, with revenue more than quadrupling since that peak multiple was set, the market is paying roughly 40 times earnings, a figure our published methodology places at the eighth percentile of NVIDIA’s own decade-long range. The business got enormously bigger. The price tag got dramatically cheaper. That tension is the whole story.
The math
Start with what the filings actually show. NVIDIA’s annual revenue was $16.7 billion in fiscal 2021. By fiscal 2026, it had grown to a figure that would have sounded like science fiction four years ago, and the company has kept more than half of every revenue dollar as profit along the way. That is not a company getting bigger while getting thinner. It is a company getting bigger while getting richer on every dollar it collects.
The quarterly cadence confirms the trend rather than contradicting it. NVIDIA’s Q3 fiscal 2026 press release showed $57.0 billion in revenue, up 62% from the same quarter a year earlier, with a gross margin well above seventy cents on the dollar. The Q1 fiscal 2027 release then reported $81.6 billion, comfortably ahead of the company’s own guidance. NVIDIA has beaten its own conservative guidance in each of the last four reported quarters, which means the filed numbers have consistently surprised to the upside even as the stock has drifted down from its 52-week high. The data center segment, which is where AI infrastructure spending lands, generated $75.2 billion in Q1 fiscal 2027 alone, a 92% year-over-year increase. That single quarter’s segment revenue is larger than the entire company’s annual revenue was just three years ago.
The math section delivers one clear message: the fundamentals have not stalled. They have accelerated.
The mood
The mood, meaning what the market is currently willing to pay per dollar of earnings, tells a different story. A year ago, the trailing P/E ratio sat near 49. Today it sits near 40, and our published methodology puts that at the eighth percentile of the past decade. The price-to-sales ratio, which compares the company’s market value to its annual revenue and is useful when earnings are lumpy, has compressed from a peak of 45 to roughly 22, the twenty-first percentile of its own history. The market is paying less for each dollar of NVIDIA’s revenue than it has in roughly four out of every five years on record.
The mood has cooled for reasons the market can articulate. Export restrictions on H20 chips to China produced a large one-time charge in Q1 fiscal 2026, briefly dragging reported gross margin well below its recent run rate before it recovered. Custom silicon from Amazon, Google, and others is growing fast, with custom AI chip sales projected to grow at nearly three times the rate of standard GPUs in 2026. The concern is not that NVIDIA is losing today. It is that the moat looks narrower than it did when the P/E was 139.
The gap
Here is where math and mood pull in opposite directions. The math says a company compounding revenue at triple-digit rates, holding margins above half of every revenue dollar, and beating its own guidance every quarter is performing at a level that historically commanded premium multiples. The mood says the premium has already been paid, the risks are real, and the next chapter is harder to read than the last one.
The bear reading is coherent. AMD holds an estimated 5 to 7% share of the AI accelerator market, a small number today, but custom chips from hyperscalers are scaling fast. If the largest cloud buyers, who are also NVIDIA’s largest customers, redirect even a fraction of their spending toward chips they design themselves, the revenue concentration risk becomes visible quickly. The data on customer concentration from the most recent 10-K was not available in our research, which is an honest gap worth naming. What we cannot measure, we cannot dismiss.
The bull reading is equally coherent. NVIDIA has held roughly four-fifths of the AI accelerator market by revenue through a period when every major technology company was trying to build around it. The Q4 fiscal 2026 results showed $68.1 billion against guidance of $65 billion, and the pattern of conservative guidance followed by material beats has now repeated four consecutive times. A business that keeps more than half of every revenue dollar as profit and is still growing at a pace most companies never see is, by the filed evidence, genuinely unusual. Whether the eighth-percentile multiple represents a different risk-reward profile than the ninety-second is a question the valuation framework surfaces but does not resolve, the answer depends on assumptions about future margin and growth trajectories that the filed data do not yet confirm.
The gap exists because the market is pricing in a deceleration that the filed numbers have not yet delivered. Whether that caution is prescient or premature is the question the next several quarters will answer.
The evidence that would change the read arrives in stages. The Q2 fiscal 2027 earnings release will show whether the Q1 beat was a one-quarter event or a sustained pattern. Gross margin is the number to watch most closely: a sustained move below seventy cents on the dollar on a GAAP basis, excluding one-time charges, would signal that competitive pressure is landing in the income statement rather than just in analyst commentary. The export restriction environment is a policy variable that no filing can predict, and that uncertainty is real. What the filings can show, and have shown consistently, is that the underlying demand for NVIDIA’s products has not yet found its ceiling.
Reading the numbers
P/E ratio at the eighth percentile of its own decade. A price-to-earnings ratio tells you how many dollars the market pays for each dollar a company earns annually. NVIDIA’s current ratio of roughly 40 means investors pay $40 for every $1 of annual earnings. The eighth percentile means that in roughly nine of every ten years on record, the market paid more than this. Think of it like a house that sold for $900,000 at its peak now listed at $400,000 while the rental income has quadrupled. The price fell even as the income rose.
Net margin of 55.8% in fiscal 2026. Net margin is the share of revenue left after every cost, tax, and expense is paid. At 55.8%, NVIDIA keeps more than half of every dollar it collects as profit. For context, most mature technology companies operate at 20 to 25%. If a business collected $100 from a customer, NVIDIA’s margin structure would leave $55.80 in the bank. That is the filed number, not a projection.
Data center revenue of $75.2 billion in a single quarter. This is one segment’s revenue in three months. NVIDIA’s entire company revenue for all of fiscal 2021 was $16.7 billion. The data center segment alone now earns in one quarter what the whole company used to earn in four and a half years. That comparison is the clearest single measure of how much the business has changed since the AI spending wave began.
Sources
- FullRatio: NVIDIA P/E ratio history
- GuruFocus: NVIDIA price-to-sales ratio
- ValueInvesting.io: NVIDIA EV/EBITDA multiples
- NVIDIA Q1 Fiscal 2025 earnings release
- NVIDIA Q2 Fiscal 2025 earnings release
- NVIDIA Q3 Fiscal 2026 earnings release
- NVIDIA Q1 Fiscal 2027 earnings release
- Jon Peddie Research: NVIDIA Q4 Fiscal 2026 results
- PatentPC: AI chip market share data
- Silicon Analysts: AMD vs NVIDIA market share 2026
