KEY POINTS
- AMD stock has gained 76% over the past 12 months, outperforming Nvidia's 30% rise, as the AI chip market expands beyond training into inference workloads where Nvidia's CUDA advantage narrows.
- AMD secured GPU capacity commitments from OpenAI and Meta totaling 6 gigawatts each, with both customers receiving warrants worth up to 10% of AMD's equity — a deal structure that signals long-term strategic partnerships.
- May 5 earnings will be the next proving ground — Wall Street wants to see MI400 order book depth and whether AMD's data center revenue can sustain the trajectory that lifted 2025 sales to $34.6 billion.
AMD has spent the last twelve months doing something it has not done in years: outrunning Nvidia in the stock market. Shares are up 76% over the trailing year, compared to Nvidia's 30% gain, and the outperformance reflects a growing market consensus that the AI chip race is no longer a one-horse show.
The catalyst was a pair of blockbuster partnerships. AMD locked in GPU capacity commitments from OpenAI and Meta Platforms, each structured around 6 gigawatts of compute power — a scale metric that underscores just how massive the infrastructure buildout has become. Both customers received warrants worth up to 10% of AMD's equity, a deal structure that turns major AI labs into long-term shareholders with skin in the game. It is an aggressive move, and it is working.
The MI400 Changes the Conversation
AMD unveiled the MI400 series at its June 2025 event, with deployment scheduled throughout 2026. The specifications are hard to dismiss: 432 GB of HBM4 memory and 19.6 TB/s bandwidth put the MI400 in direct competition with Nvidia's highest-end offerings. For training workloads, Nvidia's full-stack advantage — hardware, NVLink networking, and the CUDA software ecosystem — still creates significant switching costs. But the inference market tells a different story.
Inference, the process of running trained models in production, is less technically demanding than training. It does not require the same level of software optimization that makes CUDA a moat. As enterprises deploy AI at scale, inference workloads are growing faster than training workloads, and that shift plays directly to AMD's strengths: competitive silicon at lower price points with growing software support through its ROCm platform.
CEO Lisa Su characterized 2025 as a transformational year and told analysts that AMD entered 2026 with substantial forward momentum. The numbers support the claim. Full-year 2025 revenue hit $34.6 billion with approximately $4.3 billion in net income and a 50% gross margin. The EPYC processor family continues to take server CPU market share, and the data center segment — including both CPUs and GPUs — is now the primary growth engine.
Risk-Reward Favors AMD, Some Analysts Say
The valuation gap matters here. Nvidia trades at a higher forward multiple, justified by its dominance and margins. AMD trades at a discount that some analysts argue does not reflect the improving competitive position. Meyka's April 10 analysis called AMD the better risk-reward trade for 2026, citing the MI400 ramp, diversified revenue streams across CPUs and GPUs, and lower concentration risk than Nvidia's data-center-heavy model.
Both companies face headwinds from China export controls. Nvidia's Q1 FY2027 guidance of $78 billion explicitly excludes China data center compute revenue. AMD's Q1 2026 guidance of $9.8 billion includes only $100 million in MI308 China sales. The export restriction regime remains a known ceiling on both companies' addressable markets, but it hits Nvidia's absolute dollar exposure harder.
What Comes Next
AMD reports Q1 2026 earnings on May 5. Nvidia follows on May 20. The market will be watching AMD's data center revenue run rate, MI400 backlog commentary, and whether the OpenAI and Meta partnerships are translating into shipments or still sitting in the pipeline. For Nvidia, the focus will be on Blackwell supply constraints, gross margin sustainability above 70%, and any updates on the next-generation Rubin architecture.
The broader implication is structural. The AI chip market is expanding fast enough — Deloitte now estimates it at roughly $500 billion for 2026 — that it can support more than one dominant supplier. AMD does not need to beat Nvidia. It needs to capture enough of the incremental demand to justify the stock's repricing. So far, the market is betting it will.
