AMD Is Reportedly Pricing Its Helios Rack 40% Above NVIDIA’s Second-Gen Rubin, Confident That Customers Will Pay The Premium - Wccftech

What the report says

Wccftech reported on July 21, 2026, that AMD is expected to price its Helios AI rack-scale system at a significant premium to NVIDIA’s second-generation Rubin rack. Citing estimates from Futurum, the article said Helios could be priced at about $5 million to $5.5 million per rack, compared with an expected $3.5 million to $4 million retail range for NVIDIA’s second-gen Rubin, implying roughly a 40% gap. Wccftech cautioned that such pricing estimates should be treated carefully because bill-of-materials assumptions and market competition may not reflect final commercial terms.

According to the report, Helios is AMD’s first full-stack rack-level platform for AI workloads. The system combines AMD Instinct MI455X GPUs, sixth-generation EPYC CPUs, Pensando networking components, Infinity Fabric interconnect technology and the ROCm software stack. Wccftech described the MI455X as using AMD’s CDNA 5 architecture and HBM4 memory, with stated performance of 40 PFLOPs for FP4 and 20 PFLOPs for FP8 compute. The report compared that with NVIDIA Rubin GPU figures of 50 PFLOPs FP4 and 17.5 PFLOPs FP8.

The article also said Microsoft has agreed to deploy Helios for Azure AI services, while naming OpenAI, Meta, Oracle, HPE, TCS, Celestica, Nutanix and the U.S. Department of Energy among organizations AMD is seeking or listing as customers. If confirmed, the pricing strategy would signal AMD’s attempt to compete in AI infrastructure on platform capability rather than simply lower cost. More broadly, the report reflects intensifying competition between AMD and NVIDIA as cloud providers and AI developers buy increasingly large, integrated compute systems.