These days, there’s a specific type of discussion that takes place in portfolio review meetings, and it nearly always ends the same way. Nvidia is mentioned. Another person brings up AMD. The meeting continues after everyone frowns and looks at the spreadsheet. On the surface, the comparison seems straightforward: two semiconductor companies selling to the same hyperscalers while both riding the AI wave. It is practically a category error to treat them as substitutes because their investment cases are so dissimilar.
One version of the story is conveyed by the numbers, and it is not nuanced. With a data center division alone valued at $62.31 billion—more than ten times the size of AMD’s entire data center business—Nvidia’s most recent quarter revenue was $68.13 billion, up 73.2 percent year over year. The non-GAAP gross margin was 75.2%. The quarter’s free cash flow was $34.9 billion. That isn’t a business. That is the budget of a small nation. In contrast, AMD reported $10.27 billion in revenue, $2.08 billion in free cash flow for the quarter, and underlying gross margins close to 55%. Nvidia is losing the race for AI chips on every crucial metric. The field is being lapped by it.
| Detail | Nvidia (NVDA) | AMD |
|---|---|---|
| Current Price | $201.67 | $278.17 |
| Market Cap | ~$4.9 trillion | ~$454 billion |
| 52-Week Range | $95.04 – $212.19 | $95 – $290 (approx.) |
| Q4 FY2026 Revenue | $68.13B (+73.2% YoY) | $10.27B (+34.1% YoY) |
| Data Center Revenue | $62.31B | $5.38B |
| Data Center Share of Total | ~91% | ~52% |
| Non-GAAP Gross Margin | 75.2% | ~55% (underlying) |
| Non-GAAP EPS Beat | $1.62 vs. $1.52 | $1.53 vs. $1.32 |
| Free Cash Flow (Quarter) | $34.90B | $2.08B |
| Forward P/E | ~36x | ~33x |
| Software Moat | CUDA ecosystem, NVLink | ROCm, open-source focus |
| Next-Gen Architecture | Rubin (successor to Blackwell) | MI450 series on TSMC 2nm |
| Major Customer Deals | Meta, OpenAI, CoreWeave | OpenAI, Oracle, HUMAIN |
| 1-Year Stock Performance | +82% | +165% |
| Since-2023 Performance | +1,120% | +242% |
| Analyst Buy Ratings | 60 (1 sell) | 37 (0 sell) |
| Consensus Price Target (NVDA) | $268.22 | — |
| Business Diversification | Narrow (GPU-focused) | Wide (CPU, GPU, FPGA, gaming) |
| Key Risk | China export restrictions, valuation | Margin gap, execution risk |
However, over the past year, AMD’s stock has performed significantly better than Nvidia’s, rising 165% as opposed to Nvidia’s 82%. That is the tension that will never go away. It is incomprehensible to investors who only consider the financials. The opposite direction cannot be explained by investors who only look at the stock chart. The truth is that neither market is wholly incorrect and that both are pricing in different futures.
The structural advantage of Nvidia is sometimes hidden by the headlines. Not only does Nvidia produce superior GPUs, but many engineers contend that, at the silicon level, it doesn’t. It’s because AI research now uses CUDA, Nvidia’s software ecosystem, as its default language. CUDA compatibility is a prerequisite for all graduate students, inference frameworks, and open-source model releases. Price is not a factor in switching costs. When model training cycles are more important than hardware savings, no serious AI lab wants to spend months retooling. The majority of hardware companies don’t even acknowledge Jensen Huang’s software as a weapon.
The situation with AMD is different. It’s not attempting to unseat Nvidia. In order to prevent Nvidia from having monopolistic pricing power, the vendor hyperscalers stay in the mix and strive to be the essential backup option. Although it’s a more modest goal, it’s more achievable. Up to 6 gigawatts of GPU capacity from AMD are guaranteed by the OpenAI supply agreement, which was signed late last year, by 2030. Oracle has promised to supply 50,000 MI450 chips. These are not awards for consolation. They signify a significant change in the willingness of cloud providers to diversify their supply chains. Additionally, AMD’s business is already more varied, including gaming chips, embedded processors, FPGAs from the Xilinx acquisition, and CPUs for Intel-dominated markets. Nvidia fluctuates in relation to AI infrastructure. If one wager fails, AMD has more options.
This becomes philosophically intriguing in the valuation discussion. Nvidia’s market capitalization is close to $4.9 trillion, and its forward P/E ratio is close to 36. AMD has a $454 billion market capitalization and trades at about 33 forward. Nvidia appears to be the cheaper growth story on paper, with lower P/E, better margins, and faster revenue growth. However, there is a well-known market law that states that it is more difficult to double when you are larger. When Nvidia doubles, its market capitalization increases by an additional $4.9 trillion. When AMD doubles, $454 billion is added. At that scale, the math is just different, as one Reddit investor stated last month.
Speaking with fund managers in London and New York, it seems that the honest response for the majority of portfolios is that this isn’t an option. This is a question of allocation. The main holding is Nvidia, which is the safer of the two, the dominant leader, and the compounder. The torque is AMD, which is smaller, more erratic, and has more asymmetric upside if the MI450 series succeeds and margins rise to the management’s target gross margin of 55–58%. As this develops, it’s difficult to avoid the conclusion that the investors who do the best over the next three years will be the ones who did not attempt to choose between Lisa Su and Jensen Huang. Both will belong to them. When the hardware is no longer the only factor that matters, they will simply own them in varying amounts based on how strongly they believe the software moat will endure.
