Nvidia Corp. earned more than $70 billion in data center revenue last fiscal year and posted GAAP net income north of $55 billion, yet the stock trades at roughly 26 times forward earnings. Advanced Micro Devices, Inc., whose data center GPU business is a fraction of that size, commands a forward multiple above 50. That gap, where the dominant AI chipmaker is priced at about half the P/E of its closest rival, captures one of the sharpest debates on Wall Street heading into mid-2026: Is Nvidia too cheap, or is AMD too expensive?
The numbers behind the split
Nvidia’s annual report for the fiscal year ended January 2026 shows a company generating profit margins that most industries would consider extraordinary. Data center sales, driven by the Blackwell GPU architecture and sustained demand from hyperscalers like Microsoft, Amazon, and Google, now account for the vast majority of total revenue. Gaming and automotive contribute meaningful but comparatively modest sums. Diluted earnings per share climbed sharply year over year, reflecting both top-line growth and operating leverage.
AMD’s 10-K for the fiscal year ended December 2025 tells a different story: a company still in the early innings of monetizing AI. Its MI300X accelerator gained traction with cloud providers and enterprise buyers, and data center revenue grew at a healthy clip, but AMD’s total earnings base remains a small fraction of Nvidia’s. The company’s product mix also spans PC processors, embedded chips, and console silicon, giving it diversification that Nvidia lacks but diluting the pure AI narrative that commands the richest multiples.
When analysts divide each company’s share price by its projected earnings for the next twelve months, the result is a forward P/E near 26 for Nvidia and above 50 for AMD, based on consensus estimates compiled in April 2026. In plain terms, investors are willing to pay roughly twice as much per dollar of expected AMD profit as they are for a dollar of Nvidia profit. That premium reflects a bet that AMD’s earnings growth rate will outpace Nvidia’s on a percentage basis, even if Nvidia’s absolute profit stays far larger.
Why Nvidia’s multiple stays compressed
A forward P/E in the mid-20s might sound modest for a company growing this fast, but the math works against Nvidia in a specific way. When a business already earns tens of billions of dollars, each incremental percentage point of growth requires enormous new revenue. Nvidia’s 10-K risk factors spell this out plainly, flagging intense competition, customer concentration among a handful of hyperscalers, and the possibility that AI infrastructure spending could prove cyclical rather than linear.
Export controls add another layer of uncertainty. U.S. restrictions on shipping advanced AI chips to China have already cost Nvidia billions in potential sales, and the regulatory landscape continues to shift. Any tightening of those rules would hit Nvidia harder than AMD in absolute dollar terms simply because Nvidia holds the larger share of the market being restricted.
There is also the question of whether current hyperscaler capital expenditure is sustainable. Microsoft, Google, Amazon, and Meta have collectively signaled hundreds of billions in AI-related spending over the next several years, but those budgets are not guaranteed. If cloud providers slow their buildouts, or if they shift spending toward custom in-house chips, Nvidia’s revenue growth could decelerate faster than the market expects, and a mid-20s P/E would suddenly look less like a discount and more like fair value.
The bull case for AMD’s premium
AMD bulls argue that the higher multiple is justified precisely because the company is earlier in its AI growth curve. The MI300 series gave AMD a credible foothold in data center GPUs for the first time, and the upcoming MI350, built on a new architecture, is expected to narrow the performance gap with Nvidia’s Blackwell chips. If AMD can capture even mid-teens market share in AI accelerators, the earnings impact would be transformative relative to its current base.
Diversification helps the case as well. AMD’s client processor business benefits from a PC refresh cycle, and its embedded segment serves automotive and industrial markets with long design-in cycles. These businesses provide a revenue floor that pure-play AI exposure does not, reducing the downside if AI spending disappoints in any given quarter.
AMD’s corporate structure also gives management flexibility to pursue acquisitions or joint ventures. The company’s governance documents authorize equity issuance and other capital tools that could fund strategic moves, though any share dilution would need to be weighed against the growth it enables. For investors paying a 50-plus P/E, execution on capital allocation is not optional; it is the entire thesis.
What the market is really pricing
Strip away the jargon and the P/E gap comes down to a single question: How durable is the AI spending cycle, and who benefits most as it matures?
If AI infrastructure demand stays strong through 2027 and beyond, Nvidia’s current multiple could prove to be a bargain. The company’s software ecosystem, particularly its CUDA platform, creates switching costs that no competitor has fully replicated. Customers building on CUDA face real friction in moving workloads to AMD or other alternatives, which protects Nvidia’s pricing power and margins even as competition intensifies.
If, on the other hand, the market is right that AMD’s growth rate will outstrip Nvidia’s in percentage terms, then paying a higher multiple for AMD makes mathematical sense. A company growing earnings at 40% annually will double its profit base faster than one growing at 20%, and the P/E ratio should reflect that difference. The risk is that AMD’s growth projections depend on winning design wins against an entrenched incumbent with deeper pockets and a more mature software stack.
Neither valuation exists in a vacuum. Both are shaped by the same macro forces: Federal Reserve policy, geopolitical tensions over semiconductor supply chains, and the pace at which enterprises move AI workloads from experimentation to production. A broad market selloff or a shift in interest rate expectations could compress both multiples simultaneously, narrowing the gap for reasons that have nothing to do with chip performance.
Where the filings meet the forecast
The most reliable anchor in this debate is the audited data sitting on SEC EDGAR. Nvidia’s and AMD’s 10-K filings provide GAAP earnings, segment breakdowns, and risk disclosures that are signed by executives under Sarbanes-Oxley and reviewed by independent auditors. Those numbers are not projections; they are legal commitments about what each company actually earned.
Everything else, including the forward P/E ratios that drive most of the valuation conversation, rests on analyst models built from assumptions about AI chip demand, pricing trends, competitive dynamics, and regulatory outcomes. Those assumptions vary widely across Wall Street. Some firms model Nvidia’s data center revenue growing 30%-plus through fiscal 2028; others see deceleration to the low teens. AMD estimates are even more dispersed, reflecting genuine uncertainty about how quickly the company can scale its GPU business against a dominant rival.
For investors trying to make sense of the gap, the disciplined move is to start with what the filings confirm, then stress-test the assumptions that drive the forward estimates. Nvidia trading at roughly half AMD’s P/E is not a verdict on which stock is the better buy. It is a snapshot of how millions of market participants, armed with the same public data but very different views of the future, are placing their bets on the next phase of the AI chip race.
