NuAxiSemiconductor trade intelligence

Another Giant Leap: The Rubin CPX Specialized Accelerator & Rack

What happened

NVIDIA announced the Rubin CPX, a specialized GPU accelerator for the prefill phase of AI inference, featuring 20 PFLOPS FP4 compute and 128GB of GDDR7 memory. This move further extends NVIDIA's architectural lead in rack-scale AI systems, forcing competitors like AMD to re-evaluate their roadmaps. Watch: This could shift future demand away from high-bandwidth HBM towards higher-volume GDDR7 for inference workloads, but immediate supply chain impact is limited.

Impact and signal

Low impactImpact score: 25/100

Affected components and suppliers

GDDR7HBM

NVIDIA Corporation AMDCustom Silicon Providers

product-launchtechnology-shiftinference-acceleration

Source and time

Source: SemiAnalysis · 9/10/2025 · View original source

Open the NuAxi live feed →

Real-time pricing, supply signals and AI analysis for semiconductor traders

NVIDIA Could Account for 37% of 2027 HBM Capacity, Worth $279 BillionTechPowerUp · 10/1/2026
Approvals for Nvidia and AMD AI chip exports to China stall under government bottleneck — 20% staff turnover hobbles Bureau of Industry and SecurityTom's Hardware · 4/13/2026
Top cloud providers to outspend Ireland's GDP on AI in 2026The Register · 2/26/2026
Alphabet's purchase commitments surge past US$800 billion amid AI infrastructure expansionDigiTimes Asia · 7/25/2026
AMD's Lisa Su says memory is becoming another pressure point for AI chipsDigiTimes Asia · 5/22/2026
Datacentre performance now sold by the rack not the chipElectronics Weekly · 9/13/2026

Another Giant Leap: The Rubin CPX Specialized Accelerator & Rack

What happened

Impact and signal

Affected components and suppliers

Source and time

Related briefings