Supercomputers/Systems/Azure NDv6 GB300 Cluster
USA · operational · mixed
Azure NDv6 GB300 Cluster
Operated by Microsoft at Microsoft Azure (undisclosed US region) .
Phase 1 seed dataset, compiled by hand. These rows were built from public operator, laboratory and vendor sources. A mechanical second-reader pass has since fetched every cited source: 1092 of 1153 systems have a readable citation that names them, and 40 are genuinely weakly sourced. Every claim carries its source and a confidence tier. Treat anything below verified as a lead, not a citation.
Measured performance
- Rmax
- -
- Rpeak
- -
- Rmax ÷ Rpeak
- -
- Cores
- -
- Power
- -
- Per watt
- - GF/W
Figures are as last publicly reported for the configuration described below, not a live measurement. Where a system was upgraded in place, the post-upgrade configuration is shown and the earlier one appears in the timeline.
What this machine is made of
One row per supplier relationship. “Supplier at build” is the company that shipped the part at the time; where that company has since been acquired, the parent it rolls up to today is shown beside it. That distinction is what makes ticker-level aggregation possible across a thirty-year dataset.
| Role | Part | Supplier at build | Quantity | Confidence | Source |
|---|---|---|---|---|---|
| Integrator | - Microsoft Azure NDv6 GB300 VM series, 18 VMs per rack | Microsoft | - | Reported | azure.microsoft.com |
| Accelerator | NVIDIA GB300 NVL72 Accelerators More than 4,600 NVIDIA Blackwell Ultra GPUs in GB300 NVL72 racks (72 GPUs and 36 Grace CPUs per rack) | NVIDIA | - | Reported | blogs.nvidia.comazure.microsoft.com |
| CPU | NVIDIA Grace CPUs Armv9 (Neoverse V2) · 72C · 3.1 GHz · TSMC 4N 36 NVIDIA Grace CPUs per GB300 NVL72 rack | NVIDIA | - | Reported | blogs.nvidia.comazure.microsoft.com |
| Interconnect | NVIDIA InfiniBand XDR (Quantum-X800) Interconnect Fat tree or dragonfly+ · 800 Gb/s per port NVIDIA Quantum-X800 InfiniBand, 800 Gb/s per GPU across racks; fifth-generation NVLink Switch inside racks | NVIDIA | - | Reported | blogs.nvidia.comazure.microsoft.com |
The read
Microsoft's first at-scale production cluster of NVIDIA GB300 NVL72 racks, announced on 9 October 2025 for OpenAI workloads. NVIDIA describes it as more than 4,600 Blackwell Ultra GPUs; Microsoft's own post says 'more than 4,600 NVIDIA GB300 NVL72', which reads as a slip for GPUs since each rack holds 72 GPUs and 36 Grace CPUs (about 64 racks). Racks are liquid cooled, with 130 TB/s of NVLink bandwidth inside a rack and 800 Gb/s per GPU across racks on Quantum-X800 InfiniBand. Neither post names a datacenter, so this sits on the same undisclosed Azure site used for Eagle and only the operator's home country is recorded. Microsoft calls it the first of many, scaling to hundreds of thousands of Blackwell Ultra GPUs; only this first cluster is recorded. The two posts are joint partner statements, so this stays reported. No power or FLOPS figure is published.
Site & facility
Change history
- 2026-10-02 Azure NDv6 GB300 Cluster added (more than 4,600 Blackwell Ultra GPUs)
Source check
We fetched this system's own citations and recorded whether each page actually mentions it. This is a corroboration signal, not a fact check, and it is published so you can see how well the sourcing holds up rather than take it on trust.
| Cited source | Result | Found on the page |
|---|---|---|
| azure.microsoft.com | names system and part | NVIDIA GB300 NVL72, NVIDIA Grace, NVIDIA InfiniBand XDR (Quantum-X800) |
| blogs.nvidia.com | names system and part | NVIDIA GB300 NVL72, NVIDIA Grace, NVIDIA InfiniBand XDR (Quantum-X800) |
Checked 2026-10-09 by pnpm verify. Re-run it and the table changes with the web.
Further reading
Inside a node
Why accelerators exist, what the memory hierarchy costs, and how the CPU and accelerator merged onto one package.
The interconnect
Topologies, why latency and tail behaviour matter more than bandwidth, and how the fabric market consolidated.
Operating a supercomputer
Acceptance testing, why "installed" and "in production" are six months apart, and the machine lifecycle.