Supercomputers/Systems/CZI GPU Cluster
USA · operational · ai training
CZI GPU Cluster
Also known as Biohub AI compute cluster
Operated by Chan Zuckerberg Initiative .
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 | - Hosted on CoreWeave GPU cloud platform (bare-metal Kubernetes environment) | CoreWeave | - | Reported | chanzuckerberg.com |
| Accelerator | NVIDIA H100 SXM5 Accelerators CUDA (Hopper GH100) · 16,896C · 1.98 GHz · TSMC 4N · 700 W 1,024 NVIDIA H100 GPUs in a DGX SuperPOD configuration | NVIDIA | 1,024 reported | Reported | chanzuckerberg.com |
| Storage | VAST Data Platform Storage Disaggregated shared-everything VAST Data flash (QLC NVMe SSD) storage | VAST Data | - | Reported | chanzuckerberg.com |
The read
The Chan Zuckerberg Initiative's dedicated AI cluster for biology, announced 19 September 2023 (CZ Biohub Network HPC team to design and operate it) and described in CZI's GPU-allocation request for applications as 1,024 NVIDIA H100 GPUs in a DGX SuperPOD configuration with VAST Data flash storage, hosted on CoreWeave's GPU cloud platform. CZI states it is operational and used to develop cell models on its virtual-cells platform, with GPU time allocated to outside researchers on a rolling basis. Biohub announced in November 2025 that it plans to expand compute to about 10,000 GPUs by 2028, which is not counted here. The physical data centre is not disclosed, so no site is assigned; because the hardware is hosted by a cloud provider rather than in CZI-owned facilities, this is a borderline entry kept at reported confidence. CZI's 2023 announcement (via The Register) gave only 'more than 1,000' H100 GPUs; the exact 1,024 figure comes from the RFA page. No HPL result or power figure is published.
Timeline
- 2023-09 Announced source
Change history
- 2026-10-01 Added CZI GPU Cluster (Chan Zuckerberg Initiative)
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.
Every source cited on this page was readable, and none of them mentions this system. They are cited at page granularity, normally because no stable deep link could be found. That is a real weakness in this row and it is why the system carries a confidence tier below verified. See how this check works.
| Cited source | Result | Found on the page |
|---|---|---|
| chanzuckerberg.com | does not name this system | - |
| chanzuckerberg.com | does not name this system | NVIDIA H100 SXM5, VAST Data Platform · counts: 1024 |
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.
Operating a supercomputer
Acceptance testing, why "installed" and "in production" are six months apart, and the machine lifecycle.
Storage and I/O
Parallel file systems, why checkpointing dominates the write load, and the metadata failure mode nobody expects.