AMD's Data-Center Revenue Doubles as It Unveils the Instinct MI455X

AMD's Q2 2026 results, reported August 4, showed data-center revenue more than doubling year over year to $6.7 billion — 58% of the company's $11.54 billion total revenue — driven by intensifying AI infrastructure demand. At Advancing AI 2026, AMD announced the Instinct MI455X, its next AI-focused GPU, built on HBM4 memory and a 2nm process.

A Real Second Source, Not Just a Challenger Narrative

NVIDIA has effectively been the default AI chip vendor for most of the current buildout cycle. AMD's data-center segment crossing more than half of total company revenue, with growth this steep, is the clearest evidence yet that hyperscalers and AI labs are actively diversifying their compute supply — not just evaluating AMD as a hedge, but genuinely deploying at scale.

Why the MI455X Matters Specifically

HBM4 and a 2nm process put the MI455X's underlying specs in direct competition with NVIDIA's current top-tier offerings, at a moment when the industry has fully allocated DRAM and HBM capacity through 2027. Chip-level competition matters less than supply-chain reality here: with memory and advanced packaging capacity this constrained, having a credible second supplier is now a genuine risk-management consideration for any large AI infrastructure buyer.

What This Means for AI Infrastructure Planning

If your organisation's AI roadmap depends on GPU-backed cloud capacity, AMD's growth is good news regardless of which vendor's silicon you ultimately land on — more credible supply competing for the same demand curve tends to ease pricing and lead-time pressure over time, even if it takes several more quarters to show up in what you actually pay for compute.