NVIDIA AI Infrastructure

📣 New name, same mission. NVIDIA Data Center is now @NVIDIAAIInfra: the home for accelerated computing, AI factories, networking, hardware, software, and the technologies powering modern AI. Follow along for all things related to NVIDIA AI infrastructure.

👏 Congratulations to @cohere on Command A+ — a powerful new model optimized for NVIDIA Blackwell and trained using NVIDIA CUDA-X libraries. Proud to be a part of it! Learn more ⤵️

CUDA Live: CUDA-Q Academic Demo Day x.com/i/broadcasts/1…

#NVIDIAGTC is coming to Berlin, October 20–22 📣 Deep dives. Hands-on sessions. Real-world AI. It all starts here. Join the developers, researchers, and business leaders building the next generation of AI infrastructure. Registration opens soon. Follow for updates. 🔗nvda.ws/494uoTs

📣@LambdaAPI is the first to publish an audited STAC-AI LANG6 result on NVIDIA HGX B200 — delivering 91% latency improvement over previous systems for workloads in quantitative research and algorithmic trading. @stacresearch Read the full report ⤵️

ICYMI: AI will help build the energy it needs. At #AIExpoDC, our VP of Hyperscale and HPC Ian Buck joined U.S. Energy Secretary Chris Wright to speak to two decades of NVIDIA supercomputing with the national labs, the intersection between AI and energy, and the Genesis Mission — the @ENERGY's initiative to apply AI to scientific discovery. Read the recap ➡️ nvda.ws/4ulzKlA

NVIDIA’s Ian Buck hand-delivered the first-ever NVIDIA Vera CPUs to our partners @AnthropicAI, @OpenAI, @SpaceX, and @OracleCloud. 🎉 Vera is NVIDIA's first custom CPU, purpose-built for the age of agentic AI. This is just the beginning. The road to Vera-powered systems starts here. Thank you to our partners for being on this journey with us. The best is yet to come. 💚

@elonmusk 👏👏👏

Vera nice, Vera nice … 🤌

Vera CPUs are moving from announcement to production. Read our blog to learn more. blogs.nvidia.com/blog/vera-cpu-…

🎉 Excited for our partner @SpaceX to try out the NVIDIA Vera CPU. This is just the beginning for Vera, our CPU purpose-built for agentic AI. Thank you to @elonmusk and the SpaceX team 🚀

What a moment! 💚 Thank you for having us @MichaelDell and the @Dell team. 🤝

Signed by Jensen at Dell Technologies World. Built for the future of agentic AI. 🚀

4,032 NVIDIA Blackwell Ultra GPUs. 56 NVIDIA GB300 NVL72 systems. 1 AI data center. Hear from the @JaneStreetGroup team as they take @dwarkesh_sp through their AI data center, running LLM training and quantitative trading workloads at up to 140kW per rack. Watch the tour. nvda.ws/4dub1Ev

AI factory revenue is defined by performance per watt. ⚡ Our full-stack accelerating computing platform enables AI builders to efficiently turn electrons into more tokens, higher revenue, and expanded margin at production-scale. This delivers industry-leading tokenomics, including: ✅ Up to 50x higher throughput per MW on NVIDIA GB300 NVL72 systems, compared to NVIDIA Hopper ✅ Up to 35x higher throughput per MW and 10x more revenue on NVIDIA Vera Rubin with Groq 3 LPX, compared to NVIDIA Blackwell Read the technical deep dive ➡️ nvda.ws/3PMMHWB

The metric that matters for real-world AI TCO? Lowest token cost. Getting there requires extreme co-design across three dimensions: ✅ Models and algorithms for mixture-of-experts (MoE) architectures ✅ Compute systems such as NVIDIA GB300 NVL72 for scale-up ✅ Software libraries including NVIDIA Dynamo for disaggregated inference, using techniques such as wide expert parallelism and multi-token prediction 🔗 Learn more about scaling agentic inference: nvda.ws/4eJMi1k

What does it take to serve agentic workloads on trillion-parameter models at 400 tokens per second per user — without trading throughput for latency? The NVIDIA Vera Rubin platform pairs Vera Rubin NVL72 with NVIDIA Groq 3 LPX to deliver low latency on trillion-parameter MoE models with 400K-token context with a 35x higher throughput per megawatt. Learn how the deterministic LPU chip-to-chip (C2C) fabric and extreme co-design address agentic AI's scale-up challenges. ➡️ nvda.ws/3RGZvhJ

📣 Join us at #CiscoLive 2026 for a series of sessions on how we're collaborating with @Cisco to accelerate and simplify AI adoption in the enterprise. Sessions: ➡️ AI Is a 5-Layer Cake ➡️ Operationalizing AI at Scale: NVIDIA IT’s End-to-End Journey to an Enterprise AI Factory ➡️ From Legacy to AI Ready: Transforming the Datacenter for the Next Era ➡️ Evolution of Agentic AI for Enterprise 📅 May 31–June 4 📍 Las Vegas, NV 🔗 Learn more: nvda.ws/4dbbywk

💡 Why did @togethercompute choose NVIDIA Blackwell to serve DeepSeek-V4? Because NVIDIA Blackwell is built for the bottlenecks that matter most in long-context inference: → KV-cache pressure during decode → MoE weight bandwidth during prefill A single NVIDIA HGX B200 system can keep DeepSeek-V4’s compressed CSA/HCA/SWA cache layouts resident across many concurrent long-context requests, while native MXFP4 support enables efficient end-to-end quantized inference for V4’s MoE weights. The result? Higher throughput, lower overhead, and optimized serving efficiency at scale.

Don’t miss our CEO Jensen Huang and @Dell Technologies Chairman & CEO Michael Dell at the Unleash the Future Keynote at #DellTechWorld to hear how we are collaborating to harness the power of AI and accelerate enterprise solutions.