NVIDIA Vera Rubin platform targets lowest token cost for AI factories

NVIDIA has launched its Vera Rubin platform, a new gigascale AI supercomputer designed for AI factories. This platform emphasizes performance per watt and the lowest token cost, featuring extreme co-design across seven chips and five rack trays. Vera Rubin NVL72 production is ramping up with partners like CoreWeave, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure, aiming to accelerate AI development and deployment. AI

IMPACT This new platform aims to significantly reduce the cost of AI computation, potentially accelerating the deployment of large-scale AI models and agentic systems.

RANK_REASON NVIDIA's announcement of a new hardware platform (Vera Rubin) with specific performance claims and partner integrations. [lever_c_demoted from frontier_release: ic=2 ai=1.0]

Read on NVIDIA Blog →

• Microsoft Azure
• ConnectX-9
• CoreWeave
• DeepSeek-R1
• Google.Cloud
• Grace Blackwell NVL72
• Groq 3 LPX
• Microsoft
• NVIDIA Vera CPU
• NVIDIA Vera Rubin
• NVLink
• Olympus core
• Oracle Cloud
• SpaceXAI
• Spectrum 6 (1025)
• Spectrum-6 SPX
• Spectrum-X Ethernet
• Vera BlueField-4 STX
• Vera Rubin NVL72

• infra
• product

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →