First quarter 2026: $10.3 billion, up 38% year over year, every segment contributing. Data center reached $5.8 billion, up 57%, now the primary driver — management frames this as a structural shift, let's call it an inflection with visibility. Operating income $1.48 billion, up 83%; free cash flow $2.6 billion, more than tripled; gross margin 55%. Full year 2025 context: $34.6 billion total, data center 48%. On track.
Let me give you the shape of it. We're the merchant alternative in AI compute - Instinct accelerators, EPYC server CPUs, Ryzen clients - and last year that was $34.6 billion of revenue, up 34%, with the data center now 48% of it. In the first quarter of 2026 data center grew 57% year-over-year to a record $5.8 billion; that's the inflection, and I'd call it a structural shift, not a quarter. I'll own the uncomfortable part plainly: our ROCm software is the challenger to the market leader's, and the gap is real - we're closing it, but I won't tell you it's closed. We committed real equity to win the largest builders - warrants for up to 160 million shares each to OpenAI and Meta, at a penny, vesting against their Instinct GPU purchases, so up to 320 million shares of potential dilution. We did that deliberately, because the deployments are multi-year and the visibility is worth it. MI450 and our Helios rack-scale platform are on track to ramp in the second half - and we say 'on track' when we mean it.
Our first quarter of 2026 (ended March 28) delivered total revenue of $10.3 billion, up 38 % year‑over‑year, with growth across every segment. Data‑center revenue hit a record $5.8 billion, up 57 %, and now serves as the primary revenue and earnings driver. Operating income rose to $1.48 billion (+83 %), GAAP net income to $1.38 billion (+95 %), gross margin was 55 % and free cash flow set a record at $2.6 billion, more than tripling; management frames the quarter as a clear inflection and structural shift.
AMD's first quarter of 2026 (period ended March 28, 2026): total revenue of $10.3 billion, up 38% year‑over‑year, with growth across every segment. Data‑center revenue hit a record $5.8 billion, up 57% and now the primary revenue and earnings driver. Operating income rose to $1.48 billion, up 83%, and free cash flow set a record at $2.6 billion. Management frames the quarter as a clear inflection and a structural shift.
OpenAI commits to 6 gigawatts of our GPUs starting with MI450; Meta's co-designed MI450 accelerator also targets 6 gigawatts, first gigawatt paired with EPYC Venice. We issued warrants for up to 160 million shares each at a penny, vesting against purchase milestones — let's call it 320 million shares of potential dilution, roughly 20% of the count. A deliberate capital-structure decision to anchor multi-year visibility.
AMD's server CPU business (EPYC) posted a fourth consecutive quarter of record revenue in Q1 2026, up more than 50% year‑over‑year, with cloud and enterprise sales each growing more than 50%. We see AI adoption scaling demand not only for accelerators but also for high‑performance CPUs that orchestrate inferencing and agentic AI. The upcoming 6th‑Gen EPYC Venice is designed to extend that role across cloud, enterprise and AI workloads and will pair with Instinct MI450 in the Helios platform.
we delivered $10.3 billion in revenue, up 38% year‑over‑year, with growth across every segment. data‑center revenue hit a record $5.8 billion, up 57% and now the primary revenue and earnings driver. operating income was $1.48 billion, up 83%; GAAP net income $1.38 billion, up 95%; gross margin 55% and free cash flow set a record at $2.6 billion, more than tripling. management framed the quarter as a clear inflection and a structural shift.
Q1 2026: $10.3 billion revenue, up 38% year over year, every segment growing. Data center $5.8 billion, up 57%, now the primary driver. Operating income $1.48 billion, up 83%; free cash flow $2.6 billion, more than tripled. Gross margin 55%. Management frames this as a structural shift, not a quarter. On track.
Under the OpenAI agreement from October 2025, OpenAI is to deploy 6 gigawatts of our GPUs, with the first gigawatt powered by Instinct MI450 series. The Meta agreement from February 2026 covers a custom MI450-based accelerator, also up to 6 gigawatts, first gigawatt paired with 6th Gen EPYC Venice. Each partner received a warrant for up to 160 million shares at a penny, vesting against purchase milestones — let's call it 320 million shares of potential dilution, roughly 20% of the current count. A material capital-structure commitment to secure the deployments.
Our first quarter of 2026 reached $10.3 billion, up 38 percent year over year with growth across every segment. Data center hit a record $5.8 billion, up 57 percent, and became the primary revenue driver. Management frames this as a structural shift, not a quarter.
EPYC posted a fourth straight quarter of record revenue in Q1 2026, up more than 50% year-over-year — cloud and enterprise each grew more than 50%. Fifth-gen led the growth. Our framing: as AI scales, the orchestration, data movement, and head-node compute around inferencing and agentic AI pull more high-performance CPU demand. Sixth-gen Venice is designed to extend that position across cloud, enterprise, and AI workloads, and pairs with MI450 in Helios. On track.
We took about $440 million in net charges in 2025 from U.S. export controls on the MI308 — let's call it $800 million of inventory and related charges in Q2, with about $360 million reversed in Q4 after some licenses were granted. Whether China data center revenue continues from here depends on the evolving license framework; that's an open item we're still working through.
In 2025 we ran an unusual two-stage transaction. Acquired ZT Systems in March for $3.2 billion cash plus 8.3 million shares. In October divested manufacturing to Sanmina for $2.4 billion cash, 1.2 million Sanmina shares, and up to $450 million earnout — let's call it $1.4 billion net cash. We retained the design IP and team for rack-scale AI systems; Sanmina becomes our preferred manufacturing partner. The structure keeps us fabless while adding system-level design capability.