NVIDIA Networking

Still here?

The conversation has moved.

NVIDIA Networking is now part of NVIDIA AI Infrastructure.

Join us there 💚

We’ve rerouted 🔀

NVIDIA Networking has joined NVIDIA AI Infrastructure.

One home for the fabric, the compute, and everything in between.

Follow along as we build the future of AI infrastructure.

We’ve rerouted 🔀

NVIDIA Networking has joined NVIDIA AI Infrastructure.

One home for the fabric, the compute, and everything in between.

Follow along as we build the future of AI infrastructure.

📍 Next stop: Taipei

Taipei Live keynote pregame is coming to .

Join us on Monday, June 1 at 9 a.m. Taipei time to kick off the show.

More details coming soon. https://nvda.ws/4fb8Sjx

Taipei, are you ready? 🌏

NVIDIA CEO Jensen Huang is taking the stage at Taipei Music Center to unveil the breakthroughs driving the next generation of AI.

📍 Monday, June 1 | 11 a.m. Taipei Time
🎥 Streaming worldwide: https://nvda.ws/49lf4Si

Taiwan doesn’t just host the future. It builds it. 💚

NVIDIA Spectrum-X Ethernet, the open, AI-native Ethernet fabric, now supports Multipath Reliable Connection (MRC).

Proven first in production on NVIDIA Spectrum-X Ethernet platform and now released as an open specification through the Open Compute Project. MRC is used by OpenAI running on infrastructure from Microsoft and Oracle on top of Spectrum-X Ethernet to help keep massive GPU clusters running in lockstep with high bandwidth utilization and resilience.

With NVIDIA Spectrum-X Ethernet, enterprises can build AI factories on an intelligent, open and resilient networking platform engineered for the next era of frontier AI.

Learn more: https://nvda.ws/4cSm8bk

Agentic AI is changing the rules for inference. With DeepSeek V4, NVIDIA Blackwell delivered 20x lower cost per token out of the box, running a 1.6T parameter MoE model with a 1M token context on day one.

But the real story is how:

NVIDIA is the only platform co-designed end-to-end across five rack-scale systems—engineered to operate as a unified AI factory rather than a collection of discrete components.

That’s what enables:
→ Higher throughput for agentic workloads
→ Lower latency across multi-step reasoning loops
→ Sustained improvements in token economics over time

As AI factories scale, cost per token becomes the metric that matters and extreme co-design is the advantage that compounds.

📗 https://nvda.ws/4f1JRHl

Q: What's the backbone of every AI factory? 🏭

A: The network!

NVIDIA Vera Rubin brings NVLink 6, ConnectX-9, BlueField-4, and Spectrum-6 together into one tightly integrated fabric.

✅ 3.6 TB/s of GPU-to-GPU bandwidth
✅ Proactive congestion control
✅ A software-defined infrastructure layer built to eliminate idle time at scale

Step inside the NVIDIA Vera Rubin platform 👉 https://nvda.ws/4tdQzNT

Still thinking about . 💚

From our CEO Jensen Huang unveiling NVIDIA Spectrum-X CPO on the keynote stage, to packed sessions diving into networking innovation and DOCA development, to the incredible energy on the show floor.

Summer can't come soon enough. See you in Taipei. 👀

Copper and fiber aren't competing. They're dividing the job.

Gilad Shainer, SVP of Networking at NVIDIA, breaks down how copper and fiber each play a distinct role in AI data center design, and why co-packaged optics is changing the power equation.

Read the Network World article to learn more ➡️ https://nvda.ws/3ON8rRC

🌐 Is your network built to scale with your AI factory?

In this webinar, Gilad Shainer, SVP of Networking at NVIDIA, and Alan Weckel, Founder and Technology Analyst at 650 Group, break down what it takes to build a networking foundation for enterprise AI, from on-prem deployments to gigascale systems.

Learn how NVIDIA Spectrum-X Ethernet gives enterprises the performance, flexibility, and ecosystem support needed to deploy AI factories at any stage of growth.

🎥 Watch now: https://nvda.ws/4d4FeLm