otherScore 55/100Watch
AMD Helios outperforms NVIDIA H100 equivalent (Vera Rubin) on peak performance and HBM capacity; strong demand expected to drive Q3 shipments and Q4 volume scaling
Lisa Su· AMD· AI· 2026-07-24· about NVIDIA (NVDA)
“Helios reportedly is more advanced than Nvidia's Vera Rubin when it comes to peak performance and high bandwidth memory capacity... demand was strong and that shipments would begin in Q3 and in higher volumes in Q4.”
Why it matters
Direct performance claim vs. NVIDIA's latest inference accelerator, coupled with near-term production ramp, signals AMD is gaining traction in data center inference—the fastest-growing AI accelerator segment. Shipping volume in Q3/Q4 suggests enterprise customers are actively deploying alternatives to NVIDIA.
Investment implication
NVIDIA inference ASP and market share at risk in inference workloads. HBM (high-bandwidth memory) suppliers (SK Hynix, Samsung) will see demand diversification across AMD and NVIDIA. System integrators (Dell, HPE, Lenovo) may accelerate dual-vendor qualification strategies.