Nvidia stock is up roughly 2,000% over the past 5 years to a ~$4.2 trillion market cap, built almost entirely on data-center revenue that grew from under $4B to over $130B a year. That's the short answer. The longer answer is more interesting.
The hardest thing to accept about Nvidia is that the move makes sense. This is not a 1999 story of a company trading at 50x sales on a hope. Nvidia earns the numbers: ~75% gross margins, net income over $70B, and a forward P/E in the low 30s. The question for 2026 is not whether the past run was justified — it was — but whether the next leg has anything left, and what specifically could break it.
Nvidia stock in 2026: why it's still up 2,000% over 5 years
Nvidia stock is up about 2,000% over five years because its data-center GPU business went from a side segment to the single most important supply chain in technology. Revenue grew from roughly $27B in fiscal 2023 to over $130B at an annual run rate by 2026, with gross margins near 75%. The stock followed earnings, not hype — the multiple actually compressed as profits caught up.
That distinction matters. When a stock rises 20x, most investors assume the multiple did the work. With Nvidia, it was the opposite. The P/E ratio is lower in 2026 than it was at several points in 2023, even as the price multiplied. Earnings grew faster than the share price for stretches of the run. That is the rarest kind of bull market — one where the fundamentals are sprinting to keep up with the tape rather than the other way around. You can track how this flows through to the rest of the sector on the Big Tech Earnings dashboard.
The Numbers Behind Nvidia's 2026 Run
Here is the financial picture that supports the stock. Each row is a reason the run has lasted — and a variable to watch if it reverses.
| Metric | ~FY2021 | 2026 (run rate) |
|---|---|---|
| Market cap | ~$350B | ~$4.2T |
| Data-center revenue | ~$3.8B | $130B+ annual rate |
| Total revenue | ~$17B | $150B+ annual rate |
| Gross margin | ~62% | ~75% |
| Net income | ~$4.3B | $70B+ |
| Data center % of revenue | ~22% | ~88% |
| Forward P/E | ~50x | ~30x |
| 5-year price change | — | +~2,000% |
Figures are approximate run-rate estimates compiled from Nvidia filings and consensus for fiscal 2026; Nvidia's fiscal year runs roughly a year ahead of the calendar.
The bull case for Nvidia stock in 2026
The bull case is simple: AI compute demand still exceeds supply, and Nvidia sells the picks. Hyperscaler capex is running around $350B a year across Microsoft, Google, Meta, and Amazon, and a large share of it flows to Nvidia GPUs and systems. As long as that spend grows, Nvidia's order book grows with it.
The CUDA moat
Two decades of software, libraries, and developer mindshare lock buyers in. Switching costs are real and measured in engineer-years, not dollars.
Annual cadence
Blackwell to its successors on a one-year product cycle keeps Nvidia a generation ahead and resets the upgrade clock for every customer every year.
Full-system selling
Nvidia now sells racks, networking, and software — not just chips — pushing average deal size and margin higher per data center.
Inference tailwind
As AI moves from training to inference at scale, total compute demand rises rather than falls. More users means more chips, not fewer.
What could break the Nvidia run: the 3 real risks
The three risks that could break the Nvidia run are customer concentration, China export restrictions, and a slowdown in the AI capex cycle. Each is concrete and already partly visible in the numbers, which is what separates them from the vague "it's a bubble" argument. A real bear case names its mechanism.
Competition: who is actually taking share?
Nvidia holds roughly 90% of the AI training market, but 2026 is the first year the competitive picture has real teeth. The threat is less about a better GPU and more about custom silicon that does not need to be better — only good enough and cheaper for a specific workload.
Where Nvidia still dominates
- ✓ Frontier model training (~90% share)
- ✓ The CUDA software ecosystem
- ✓ Highest-performance single-chip and rack systems
- ✓ Networking via the ~$13B+ InfiniBand/Spectrum-X business
Where share is leaking
- ✕ Google TPUs for internal and Cloud workloads
- ✕ Amazon Trainium/Inferentia for AWS-native AI
- ✕ AMD MI300/MI400 closing the inference gap
- ✕ Hyperscaler in-house ASICs for cost-sensitive inference
The pattern to watch: every major customer is also building a competitor. That is not a coincidence — it is rational behavior when one supplier captures most of your fastest-growing cost line. Compare the full hyperscaler spend picture on the AI Spending dashboard.
Is Nvidia stock a buy in 2026? My honest take
I have watched a lot of "this time is different" stories end badly. Nvidia is genuinely different in one respect: the earnings are real, the cash is real, and the margins are real. At a forward P/E near 30x with earnings still growing double digits, the valuation is not the problem. The PEG ratio sits below 1, which is cheaper than the broad market on a growth-adjusted basis.
The problem is that the entire thesis rests on a single variable — AI capex — controlled by four or five companies. When the bull and bear case both hinge on the same number, you are not buying a diversified business; you are making a concentrated bet on the durability of one spending cycle. That can absolutely keep working in 2026. But size the position accordingly, and watch hyperscaler capex guidance more closely than you watch Nvidia's own earnings.
The 2,000% run was earned, not inflated.
But a $4.2T company whose bull and bear case both come down to hyperscaler capex is one spending decision away from a re-rate. Watch the buyers, not the chips.
Track AI infrastructure spending and earnings on the Big Tech Earnings dashboard at Value Add VC. Originally published in the Trace Cohen newsletter.