NVIDIA announced Spectrum-6, described in the company's blog post as a 102.4-terabit-per-second Ethernet switch system delivering 2x the capacity of previous-generation systems. NVIDIA said the switch is "arriving across the world's gigascale AI factories" and is built as part of the NVIDIA Vera Rubin platform. According to the same post, the Spectrum-6 switch chip combines with the NVIDIA ConnectX-9 SuperNIC to form the next generation of Spectrum-X Ethernet — tying the standalone switch specification directly to the broader platform NVIDIA is shipping it under.
NVIDIA's blog post lays out the full architecture: Spectrum-6 is engineered as part of Vera Rubin, which brings together the NVIDIA Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU and the Spectrum-6 Ethernet switch itself. In other words, the 102.4 Tb/s switch NVIDIA detailed is not a standalone product — it is one of six named components in a single integrated platform, paired specifically with the ConnectX-9 SuperNIC to form the next-generation Spectrum-X Ethernet fabric.
NVIDIA said Spectrum-X Ethernet delivers up to 1.6x higher AI networking performance than off-the-shelf Ethernet and sustains up to 95% network efficiency across deployments exceeding 100,000 GPUs. NVIDIA framed additional gains as layered on top of that baseline: according to the blog post, hardware-accelerated Spectrum-X multiplane topologies reduce the number of switches required for data centers by 1.7x, which NVIDIA says further accelerates network efficiency beyond the chip-level specification alone.
NVIDIA quoted infrastructure partners to frame why coordination matters at scale. Min Jun, director of product for networking at CoreWeave, said: "CoreWeave is built for the most demanding AI workloads, and networking is central to delivering that performance at scale. Bringing NVIDIA Spectrum-6 and liquid-cooled Spectrum-X Ethernet infrastructure into our AI factories will help us deliver the bandwidth, resilience and efficiency customers need to train frontier models and deploy inference faster." Laurelle Roseman, vice president of global partnerships at Nebius, said: "At gigascale, performance comes down to coordination: keeping every GPU in lockstep so one slow link doesn't stall an entire job. That's what NVIDIA Spectrum-6 goes after, and why we brought it in early — a fabric that stays fast and resilient as our customers' most demanding workloads scale." NVIDIA's post further stated that layering on the benefits of Spectrum-X Ethernet Photonics adds up to 5x higher power efficiency and 10x improved mean time between incidents, which NVIDIA says allows network fabric enhancements to keep delivering workload acceleration beyond component-level optimization alone.
NVIDIA's blog post named CoreWeave, Microsoft, Nebius, SpaceX, xAI and Tesla as being "among the first to bring in Spectrum-6 to accelerate their AI factories." A narrower subset was named specifically for the Vera Rubin platform: NVIDIA said CoreWeave, Microsoft and Nebius "will be among the first providers to deploy NVIDIA Vera Rubin-based infrastructure with Spectrum-6, extending access to the platform across a broad community of developers, startups and enterprises."
| Metric | Value | Evidence |
|---|---|---|
| Spectrum-6 switch capacity | 102.4 Tb/s (2x prior generation) | E1 |
| Spectrum-X performance vs. off-the-shelf Ethernet | up to 1.6x higher | E7 |
| Sustained network efficiency at 100,000+ GPU scale | up to 95% | E7 |
| Switch count reduction via multiplane topology | 1.7x fewer switches | E8 |
| Power efficiency gain via Spectrum-X Ethernet Photonics | up to 5x higher | E8 |
| Mean time between incidents improvement | up to 10x | E8 |
NVIDIA's own figures show gains compounding at different layers of the same fabric: the Spectrum-6 chip itself doubles switching capacity to 102.4 Tb/s, the surrounding Spectrum-X fabric sustains up to 95% efficiency at 100,000-plus-GPU scale while cutting switch counts by 1.7x, and the Photonics layer adds up to 5x power efficiency and 10x MTBI improvement on top — a stack of claims NVIDIA presents as additive rather than isolated. The adopter list also splits along the same lines NVIDIA drew for the platform itself: CoreWeave, Microsoft and Nebius are named as first movers on Vera Rubin-based infrastructure with Spectrum-6, while SpaceX, xAI and Tesla are named only as early Spectrum-6 adopters generally, without the same Vera Rubin infrastructure claim attached in NVIDIA's post. That distinction lines up with the quoted commentary: it is CoreWeave's Min Jun and Nebius's Laurelle Roseman — both drawn from the narrower Vera Rubin cohort — whose statements NVIDIA used to explain the bandwidth, resilience and coordination goals the fabric is meant to serve.
According to NVIDIA's blog post, Spectrum-6 is a 102.4-terabit-per-second Ethernet switch system, delivering 2x the capacity of previous-generation systems.
NVIDIA named CoreWeave, Microsoft, Nebius, SpaceX, xAI and Tesla as being among the first to bring in Spectrum-6 to accelerate their AI factories.
NVIDIA said CoreWeave, Microsoft and Nebius will be among the first providers to deploy NVIDIA Vera Rubin-based infrastructure with Spectrum-6.
NVIDIA said Spectrum-X Ethernet delivers up to 1.6x higher AI networking performance than off-the-shelf Ethernet and sustains up to 95% network efficiency across deployments exceeding 100,000 GPUs.
Per NVIDIA's blog post, Vera Rubin brings together the NVIDIA Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU and the Spectrum-6 Ethernet switch.
According to a NVIDIA newsroom announcement, NVIDIA and KAIST (Korea Advanced Institute of Science and Technology) are launching a joint AI research lab at the KAIST Kim Jaechul Graduate School of AI in Seoul to advance agentic AI for South Korea, backed by a $300 million, five-year collaboration that includes $50 million a year in compute contributions.
According to an Intel Foundry newsroom announcement, Intel Foundry has completed the Rapid Assured Microelectronics Prototypes – Commercial (RAMP-C) program, a Department of War-backed effort that Intel says supported development of domestic, leading-edge CMOS manufacturing and now feeds into its Secure Enclave initiative.
According to TrendForce data cited by CNA and UDN, nine major cloud service providers' (CSP) 2026 capital expenditure is projected to rise about 90% to over $886.7 billion, while TrendForce raised its 2026 AI server shipment growth forecast from 28% to near 31%, citing stronger demand for NVIDIA rack-scale AI systems.