At CES 2026, AMD unveiled its Helios turnkey solution, bundling MI455 GPUs, Milan Venice EPYC CPUs, 200-gigabit Ethernet networking and AMD’s ROCm software stack into a repeatable, rack-scale design. Hyperscale customers can now deploy 50-rack blocks with 100 petaflops of AI performance each, enabling straightforward roll-outs of 1,000 racks or more. AMD projects that every 10-rack increment adds $250 million in revenue, and the modular architecture has already drawn commitments for over 500 racks from three leading cloud providers. By tying system-level sales directly to customer CapEx budgets, AMD expects to grow its AI segment by 35% year-over-year in 2026 and reduce quarterly revenue swings tied to standalone CPU or GPU cycles.

Industry analysts at Morgan Stanley and Goldman Sachs now forecast AMD’s data center CPU revenue to outpace the overall server market by 20 percentage points in 2026. Following Q4 results—where EPYC server processor shipments rose 42% year-over-year and average selling prices climbed 15%—AMD secured design wins for its Genoa-X CPUs in 80% of new high-performance computing clusters announced this year. With an expanded partner program now delivering over 60 validated server configurations, AMD expects its data center unit to contribute $8.4 billion in revenue for fiscal 2026, up from $6.2 billion in 2025. At that run rate, data center sales would account for 45% of AMD’s total top line, underscoring the strategic shift toward enterprise compute workloads.