TLDR
- Nvidia unveiled Spectrum-6, a 102.4-terabit-per-second Ethernet switch platform for gigawatt-scale AI data centers.
- CoreWeave, Microsoft, Nebius, SpaceXAI, and Tesla are among the first customers set to deploy Spectrum-6.
- The platform delivers twice the capacity of previous-generation systems and is built for the Vera Rubin AI supercomputer architecture.
- Nvidia stock rose around 2% on Tuesday following the announcement.
- AMD also announced Microsoft as a customer for its rival Helios Rackscale Solution on Monday, with Meta, OpenAI, and Oracle also on board.
Nvidia stock climbed roughly 2% on Tuesday after the company announced its next-generation Ethernet platform, Spectrum-6, targeting gigawatt-scale AI data centers.
The new platform is a 102.4-terabit-per-second Ethernet switch system. That’s double the capacity of the previous generation, and it’s built as part of the Vera Rubin platform — Nvidia’s multi-rack AI supercomputer aimed at agentic AI workloads.
NVIDIA Spectrum-6, a 102.4-terabit-per-second Ethernet switch built for the Vera Rubin platform, is now arriving across the world’s gigascale AI factories.@CoreWeave, @Microsoft, @Nebius, @SpaceXAI and @Tesla are among the first to bring in Spectrum-6 to accelerate their AI… pic.twitter.com/zhaUttCIb5
— NVIDIA (@nvidia) July 21, 2026
Some big names are getting first dibs. CoreWeave, Microsoft, Nebius, SpaceXAI, and Tesla are all listed as early adopters. CoreWeave, Microsoft, and Nebius will also be among the first to deploy Vera Rubin-based infrastructure with Spectrum-6 included.
Min Jun, director of product for networking at CoreWeave, said the new infrastructure will help the company “deliver the bandwidth, resilience and efficiency customers need to train frontier models and deploy inference faster.”
For cloud providers, Nvidia says Spectrum-6 allows more compute to run as a single unified resource — making it faster to train models and run inference at scale.
What Is Inference, and Why Does It Matter Here?
AI inference is when a trained model takes new data and produces an output — a prediction, a response, a decision. It’s one of the most compute-intensive workloads in production AI, and speed here directly affects cost and customer experience.
That’s a key part of what Spectrum-6 is designed to accelerate. Faster networking means less waiting between GPUs, which means models run more efficiently.
Nvidia also highlighted that the Spectrum-X Ethernet Photonics switch systems improve power efficiency and uptime by 5x — a number that matters at the scale these customers are operating.
The platform gives developers, startups, and enterprises broader access through these cloud providers. It’s not just the hyperscalers who benefit; Nvidia is positioning Spectrum-6 as infrastructure that cascades down to a wide developer community.
AMD Also Made Moves With Microsoft This Week
The announcement comes a day after rival Advanced Micro Devices said Microsoft will deploy its AMD Helios Rackscale Solution for frontier model AI inference. Meta, OpenAI, and Oracle are also listed as early Helios customers.
That’s a notable overlap in the customer list — Microsoft appears on both Nvidia’s and AMD’s early deployment rosters, which reflects how large cloud operators are hedging across multiple suppliers rather than committing exclusively to one.
Nvidia has not said when Spectrum-6 deployments will go live, but the early customer lineup suggests rollout timelines are being coordinated now.
Nvidia stock was up approximately 2% on Tuesday following the Spectrum-6 news.
Stop guessing and start investing with confidence. KnockoutStocks gives you the AI insights, market intelligence, and stock research you need to spot opportunities, cut through the noise, and make smarter investment decisions — all in one powerful platform.
Sign up today and get 50% OFF full access to our premium stock picks.
Simply use coupon code SPECIAL50 at checkout to claim your exclusive discount.
Be the first to comment