Summary: On May 5, 2026, AMD reported $10.25 billion in quarterly revenue, up 38%, with data center revenue of $5.8 billion, up 57%. CEO Lisa Su raised her three-to-five-year server CPU market growth outlook from 18% to 35% annually, projecting a market above $120 billion by 2030. The quarter also pointed to MI450 and Helios ramps for Meta and OpenAI deployments.
AMD closed the books on its first quarter after the bell on May 5, 2026: total revenue of $10.25 billion, a 38 percent year-over-year gain, with the data center segment contributing $5.8 billion of it, up 57 percent. Shares added more than 14 percent in after-hours trading. The more consequential news landed on the earnings call, where CEO Lisa Su all but doubled her growth outlook for the server CPU market over the next three to five years — from 18 percent a year to 35 percent — a pace that would carry the market beyond $120 billion by 2030.
Her verdict on the quarter: “a clear inflection in our growth trajectory and a structural shift in our business.” The meaning is that data center has become the core force driving AMD’s revenue and profits. For two years the AI infrastructure story has been told almost entirely as a GPU story; this report moves CPUs back to the middle of it.
Reading the Key Figures: Data Center Is Now Half of AMD
- Data center revenue landed at $5.8bn, 57 percent above a year ago, and now accounts for more than half of company revenue, with Epyc server CPUs and rising Instinct GPU shipments doing the lifting
- Net income came to $1.4bn — 95 percent higher than last year, though 8 percent under the prior quarter; non-GAAP gross margin held at 53 percent, a point beneath Q4, while operating income fell 16 percent QoQ
- Operating expenses climbed 34 percent YoY; CFO Jean Hu highlighted a record for quarterly free cash flow
- Q2 guidance calls for $11.2 billion in revenue — about 46 percent YoY and 9 percent sequential — with non-GAAP gross margin near 56 percent
The quarter-over-quarter softness is nothing to read as trouble: spending on inventory build-up, capacity expansion, and the AI software ecosystem all lands in the expense lines, and investor attention has already moved on to how fast MI450 ramps in the second half.
The CPU Story: Agentic AI Rewrites Server Demand
Su had already called it in March: server CPU demand “far exceeded” expectations, and agentic AI applications were the main force behind it. The message this time was much the same — inference and agentic workloads keep raising the appetite for server CPU compute, and AMD is coordinating with supply chain partners to “meaningfully increase” wafer and back-end capacity. She expects server CPU revenue to rise more than 70 percent YoY in Q2, and sees that momentum running through 2027.
The analysts went further still. Per Stephen Sopko of HyperFrame Research, inference and agentic AI “aren’t just incremental — they’re structurally shifting the CPU-to-GPU ratio higher,” which widens the total addressable server CPU market. Matt Kimball of Moor Insights adds that once inference becomes the center of gravity, “that CUDA moat becomes less treacherous to cross” — buyers increasingly judge chips on token economics and performance-per-dollar rather than raw performance.
MI450, Helios, and the Two 6GW Deals
- Meta (the 6GW agreement signed in February 2026): shipments start in the second half — custom Instinct GPUs built on the MI450 architecture, running alongside sixth-generation Epyc “Venice” CPUs in the Helios rack-scale architecture; analysts interviewed put power density at 120–140 kW per rack, liquid cooling required
- Helios timing: first volumes in Q3, the ramp turning substantial in Q4, with the climb continuing into Q1 2027
- OpenAI (the 6GW deal from October 2025): Su describes progress as “going really well”; OpenAI has previously said it will stand up a 1GW data center on MI450
- Tata Consultancy Services will design Helios systems aimed at enterprise and sovereign AI; AWS, Google Cloud, and Azure are all broadening their Epyc-based instance lineups in step
Supply Tightness and Visibility Into Power
Faced with supply chain doubts, Su’s remarks are worth recording word for word: AMD’s view of 2027 deployments now extends “down to which data centers the GPUs are going to be installed in” — a requirement, in her framing, under the constraints everyone is working against. She granted that the supply chain and data center construction are both tight, but said AMD is confident it can ship at or beyond the growth levels it forecasts, adding that “there is much more power that’s coming online in 2027.”
How Developers and Procurement Should Read It
Three consequences for practitioners. One, GPUs finally have a genuine second source: Meta and OpenAI have handed gigawatt-scale orders to AMD, so the architecture risk of leaning on a single chip vendor now has an answer — and ROCm maturity is the next milestone to track. Two, CPUs are first-class citizens of AI clusters once more: orchestration and real-time inference lift Epyc demand in tandem with GPUs, which leaves accelerator-only cost models obsolete. Three, 120–140 kW racks with liquid cooling are now mainstream deployments, and that retires a stack of assumptions baked into facility design and capacity planning. Our 2026 opening outlook made the case that infrastructure constraints would dictate this year’s pace — this quarter reads as the newest footnote to it.