AMD Q2 2026: Record $11.5B Revenue as Data Center More Than Doubles — Helios, MI400, and the ROCm Software Pivot

AMD has reported its strongest quarter in company history: $11.5 billion in revenue, up 50% year-over-year, with Data Center revenue more than doubling to $6.7 billion. The results confirm that AMD's multi-year bet on AI accelerators and server CPUs is compounding — and that the competitive landscape in AI silicon is no longer a two-horse race between NVIDIA and the hyperscalers' custom chips.

The Numbers That Matter

Q2 2026 revenue of $11.536 billion was up 50% year-over-year and 13% quarter-over-quarter. On a non-GAAP basis, operating income hit $3.1 billion (27% operating margin, up 15 points year-over-year), and diluted EPS came in at $1.66 — a 246% improvement versus Q2 2025, though that comparison is inflated by the $800 million inventory charge AMD took last year following US export controls on its Instinct MI308 GPUs.

Stripping out that one-time headwind, the underlying trajectory is still striking. Gross margin expanded from 43% to 56% non-GAAP year-over-year. Net income on a non-GAAP basis reached $2.76 billion. The company's full Q2 2026 earnings release and slide deck are available on AMD's investor relations site.

Data Center: The Engine

At $6.7 billion, AMD's Data Center segment now represents 58% of total company revenue — up from roughly 40% a year ago. The growth is being driven by two concurrent cycles: EPYC server CPUs continuing to take share from Intel in cloud and enterprise deployments, and Instinct GPUs ramping into AI training and inference infrastructure at scale.

The headline product announcement this quarter was AMD Helios — a rackscale AI server solution the company is positioning as the world's most powerful AI server rack by inference tokens per dollar. Helios is already being deployed by a roster that reads like the AI industry's who's who: Anthropic, Meta, Microsoft, OpenAI, Oracle, and others. That customer list matters as much as the product specs — it signals AMD has cleared the qualification hurdles that historically kept new entrants out of hyperscaler AI infrastructure.

Alongside Helios, AMD launched the Instinct MI400 Series, comprising the MI455X (large-scale training and inference) and the MI430X (HPC and sovereign AI). The MI455X in particular is designed to compete directly with NVIDIA's Blackwell generation on memory capacity and bandwidth — the two metrics that most constrain large model training runs.

EPYC, ROCm, and the Software Play

The Q2 results also mark the launch of 6th Generation EPYC server CPUs, optimised across agentic AI, general-purpose, and enterprise workloads. EPYC has quietly been one of AMD's most durable competitive advantages — x86 server CPU share gains against Intel have been consistent for six consecutive years, and the 6th Gen launch extends that trajectory into the next server refresh cycle.

Equally significant was the release of ROCm.ai — AMD's rebranded, AI-native developer platform. The historical knock on AMD's GPU stack has always been software: CUDA's ecosystem depth made NVIDIA the default even when AMD's hardware was price-competitive. ROCm.ai represents AMD's most serious attempt to close that gap, with a redesigned developer experience focused on deployment speed and cross-platform optimisation. A strategic partnership with Anthropic — which will use Claude to help optimise AMD's Instinct GPUs and ROCm software — adds a credibility dimension that pure developer tooling announcements rarely carry.

The Rest of the Business

Client revenue (Ryzen CPUs) came in at $3.1 billion, up 23% year-over-year on strong consumer and commercial PC demand. Gaming revenue fell 31% to $779 million as semi-custom revenue (primarily console chips for PlayStation and Xbox) declined — a well-flagged cycle that is approaching its floor as next-generation console programmes move toward design-in. Embedded revenue of $977 million was up 19% as industrial and communications end markets continued their recovery from the 2024–25 inventory correction.

Forward Look

AMD guided Q3 2026 revenue of approximately $13.0 billion (±$500 million), implying sequential growth of roughly 13% and year-over-year growth of around 50%. Management specifically called out Data Center as the accelerant for H2 2026 growth, with Helios ramping and Instinct deployments scaling. For the competitive context, AMD's results should be read alongside Etched's inference-only challenger approach and NVIDIA's Alpamayo 2 Super open model push. On the infrastructure side, see Meta and BlackRock's $14B data center deal — AMD's Helios racks are among the infrastructure being procured for deployments of this scale. The policy framework shaping all of these investments is covered in our analysis of Anthropic's new Chief Global Affairs Officer hire and OpenAI's enterprise education push.

References

• AMD — Q2 2026 Earnings Press Release (Aug 4, 2026)
• AMD Q2 2026 Earnings Slides (PDF)
• AMD — Instinct MI400 Series Launch
• AMD — 6th Gen EPYC Server CPU Launch
• AMD × Anthropic Strategic Partnership
• AMD × Microsoft Azure AI Infrastructure Expansion
• ROCm.ai — AMD's AI-Native Developer Platform
• AMD EPYC Server Processors
• AMD — Full-Stack Compute & Agentic AI Announcement
• AMD Q2 2026 Earnings Webcast