Nvidia has unveiled Vera Rubin, an ambitious new computing platform comprising seven chips now in full production, backed by an extraordinary roster of AI industry leaders including Anthropic, OpenAI, Meta and Mistral AI. Announced at the company's annual GTC conference, CEO Jensen Huang described it as 'a generational leap' poised to initiate 'the greatest infrastructure buildout in history'. The platform promises up to 10 times more inference throughput per watt and one-tenth the cost per token compared to Blackwell systems, with every major cloud provider—AWS, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure—committed to offering it.
The Vera Rubin platform integrates seven distinct chips: the Nvidia Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, Spectrum-6 Ethernet switch and the newly integrated Groq 3 LPU inference accelerator. These components are organised into five interlocking rack-scale systems functioning as a unified supercomputer. The flagship NVL72 rack connects 72 Rubin GPUs and 36 Vera CPUs, reportedly capable of training large mixture-of-experts models using just a quarter of the GPUs required on Blackwell—a claim that could fundamentally reshape the economics of frontier AI development.
Nvidia's strategic vision centres on what Huang calls 'agentic AI'—autonomous systems that reason continuously for extended periods, execute software, utilise external tools and improve iteratively. This represents a fundamental shift from chatbots that respond to single prompts. The company launched its Agent Toolkit, including OpenShell, an open-source runtime for secure autonomous agents, with adoption from Adobe, Atlassian, Cisco, CrowdStrike, Salesforce, SAP, ServiceNow and Siemens. Additionally, Nvidia introduced the Nemotron Coalition, a global collaboration developing open frontier models with partners including Mistral AI, Perplexity, LangChain and Mira Murati's Thinking Machines Lab.
The platform's reach extends across diverse industries. Roche is deploying over 3,500 Blackwell GPUs for drug discovery and digital manufacturing twins. BYD, Geely, Isuzu and Nissan are building Level 4-ready autonomous vehicles on Nvidia's Drive Hyperion platform, whilst Nvidia and Uber plan autonomous vehicle launches across 28 cities by 2028. The company also released its first healthcare robotics platform anchored by Open-H, the world's largest healthcare robotics dataset, and introduced the Vera Rubin Space Module for orbital computing, delivering 25 times more AI compute than the H100 GPU for space applications.
Nvidia's DGX Station, a deskside system delivering up to 20 petaflops of AI compute performance and capable of running trillion-parameter models, represents a quieter but potentially significant move into enterprise hardware. The company also unveiled reference designs for 'AI factories'—entire facilities optimised for AI production—with dynamic power provisioning enabling 30 per cent more infrastructure within fixed-power data centres. Whilst the performance claims remain largely unverified by independent benchmarks and the agentic AI thesis represents a bet on an unproven future, no competitor matched the comprehensiveness of Nvidia's vision or secured endorsements from leaders of Anthropic and OpenAI.
Fuente Original: https://venturebeat.com/infrastructure/nvidia-introduces-vera-rubin-a-seven-chip-ai-platform-with-openai-anthropic
Artículos relacionados de LaRebelión:
- Nvidias Nemotron 3 Super Revolutionises AI Efficiency
- US Chip Firms Sued Ukraine War Weaponry Allegations
- US Government Stake in Intel Trump Considers Deal After CEO Attack - What It Means for Chip Ma...
- Adult Content Creators in Crisis Navigating Platform Bans and Payment Processor Pressure
Artículo generado mediante LaRebelionBOT
No hay comentarios:
Publicar un comentario