Nvidia and AMD sell the same thing to the same customers: AI accelerators for the largest cloud companies and AI labs. Measured against their own histories, the market prices them at opposite extremes. As of October 5, 2026, AMD trades at the 99.6th percentile of its own five-year EV/Sales range. Nvidia trades at the 20th. On P/E, Nvidia sits at the 3rd percentile.
If you only read one thing
The short version: AMD's sales multiple more than doubled in 2026, from 10.3 times revenue in January to 25.2, as it signed multi-gigawatt GPU deals with OpenAI and Anthropic. Nvidia's fell, from 23.9 to 18.7, because its revenue grew 83% and the stock rose less. AMD is priced for its AI business to grow into the multiple. Nvidia is priced as if its growth will fade. Each could be wrong in either direction.
The numbers side by side
| Nvidia (NVDA) | AMD | |
|---|---|---|
| EV/Sales (TTM) | 18.7 | 25.2 |
| EV/Sales 5y median | 25.8 | 8.9 |
| EV/Sales 5y percentile | 20 | 99.6 |
| P/E (TTM, GAAP) | 29.6 | 163.0 |
| P/E 5y median | 57.2 | 106.4 |
| P/E 5y percentile | 3 | 71 |
| Revenue growth (TTM) | 83% | 40% |
| EV/Sales, January 5, 2026 | 23.9 (35th) | 10.3 (72nd) |
Where these stand now
Current readings update separately from the article’s dated analysis.
Within its usual range
AMD's P/E of 129.8x is higher than 63% of its weekly readings over the last five years.
Historically low
NVDA's P/E of 25.9x is the lowest of its weekly readings over the last five years.
We lead with EV/Sales (enterprise value over trailing revenue) because it compares the two on the same footing; our guide on when to use P/E, P/S, EV/Sales, or EV/EBITDA explains why. The P/E row needs a caveat, covered below.
The takeaway: AMD costs nearly three times its own five-year median relative to sales. Nvidia costs about a quarter less than its median. In January neither was at an extreme: Nvidia sat at the 35th percentile and AMD at the 72nd.
How AMD got to the top of its range
AMD spent the year collecting proof that it can be the second supplier. In October 2025 it signed a deal to supply OpenAI with 6 gigawatts of GPUs across several generations, starting with 1 gigawatt of MI450 chips in the second half of 2026. Its second-quarter 2026 results added a partnership with Anthropic for up to 2 gigawatts of MI450 GPUs, with data center revenue up 107% to $6.7 billion and total revenue up 50%. AMD guided to about $13 billion of third-quarter revenue. In September its stock rose about 29% and its market value crossed $1 trillion, per Yahoo Finance on the September chip rebound.
The multiple prices most of that in. At today's price, AMD's trailing revenue would need to nearly triple to bring its EV/Sales back to its own five-year median. The MI450 ramp is what has to deliver it, and the first gigawatt for OpenAI only starts this half. AMD's EV/Sales history shows how sharp the 2026 move was.
The takeaway: AMD's multiple re-rated on contracts that have not yet turned into revenue. That is a reasonable thing for a market to do, and it leaves little room if the MI450 ramp slips.
How Nvidia got to the bottom of its range
Nvidia's problem, if it is one, is that its revenue outran its stock. Its second fiscal quarter brought $96.2 billion of revenue, up 106%, and a $108 billion guide that assumes no data-center sales to China. In September it added a record $150 billion to its buyback. The stock was up a bit over 20% for the year at that point, well behind its revenue.
The result is a P/E of about 30 against a five-year median of 57, which puts it at the 3rd percentile of its own range. The only weeks in the last five years with a lower P/E came this summer, mostly in the weeks after its August report. We looked at the forward version of this in August, in what Nvidia's lowest forward P/E in a decade does and does not mean. The conclusion still holds: a compressed multiple on accelerating revenue means the market expects the growth to slow sharply, or doubts it will last. It is not a forecast that the stock goes up. Nvidia's P/E history shows the full band.
The takeaway: Nvidia's trailing earnings more than doubled in a year while its P/E fell from about 52 to about 30. The market is pricing Nvidia as if today's profits are near a peak.
Why the P/E comparison is messier
AMD's P/E of 163 looks absurd next to Nvidia's 30, but part of the gap is accounting. We use GAAP earnings as filed, and AMD's GAAP profit carries large amortization charges from its 2022 purchase of Xilinx. Data vendors that report adjusted earnings show a lower AMD P/E than ours. That is why EV/Sales is the cleaner comparison here, and why AMD's P/E percentile (71st) looks less extreme than its EV/Sales percentile.
Nvidia has no such distortion: last quarter its GAAP earnings per share ($2.46) were close to its adjusted figure ($2.22). Its P/E reading is as clean as these get.
The takeaway: Compare the two on EV/Sales. AMD's GAAP P/E overstates how expensive it is relative to Nvidia, but not enough to change the picture.
What would change the picture
- MI450 shipments. Revenue from the OpenAI and Anthropic deals has to show up in AMD's next few quarters for the multiple to hold.
- Nvidia's growth rate. If the $108 billion guide is met and growth stays high into 2027, the low percentile looks like a mistake by the market. If growth slows sharply, it was a correct read.
- Cloud spending plans for 2027. Both companies depend on the same handful of buyers, covered in our ranking of AI infrastructure stocks by valuation percentile.
- Custom chips. Broadcom's AI chip revenue rose 221% last quarter, per Broadcom's third-quarter release. Every custom chip a cloud company designs is one it does not buy from either of them.
Data: SEC EDGAR XBRL filings and market prices, computed by StockResearch as of October 5, 2026. P/E uses GAAP diluted earnings as filed. Company figures come from the linked releases. Jake is the founder of StockResearch.app, where he writes data-first research on valuation dislocations. Nothing here is investment advice or a recommendation to buy or sell any security. Do your own research.