NVIDIA Vera Rubin and AMD Helios unveiled at CES 2026 — both production Q3 2026
Jensen Huang and Lisa Su both used CES 2026 keynotes to anchor 2026 roadmaps on memory rather than raw compute. NVIDIA's Vera Rubin (HBM4) and AMD's Helios rack-scale (MI450) are both targeting Q3 2026 production. The competitive axis has shifted to bandwidth.
Vera Rubin is NVIDIA's successor to Blackwell, the first NVIDIA platform built around HBM4 from day one. The architecture-level story is less about peak FLOPs and more about how much of the parameter tensor you can hold in fast-access memory — which directly determines inference throughput on long-context, agent-style workloads.
Helios is AMD's rack-scale answer, MI450 silicon at the node level, with rack-level interconnect optimized for the same workloads. AMD's pitch: the MI350 generation (currently shipping at scale) is already competitive on TCO; the MI450 generation in Q3 closes the remaining gap. The 2026 procurement decisions for the major hyperscalers will be settled on these two roadmaps.