Supercomputers/Systems/Meta Research SuperCluster (RSC)
USA · operational · ai training
Meta Research SuperCluster (RSC)
Operated by Meta Platforms at Meta (undisclosed US site) .
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 | - 760 NVIDIA DGX A100 systems as its compute nodes | NVIDIA | 760 reported | Verified | ai.meta.comblogs.nvidia.com |
| Integrator | - Penguin Computing, architecture partner and managed services | Penguin Computing | - | Reported | ai.meta.comblogs.nvidia.com |
| Accelerator | NVIDIA A100 (in DGX A100) Accelerators 760 NVIDIA DGX A100 systems for a total of 6,080 GPUs | NVIDIA | 6,080 reported | Verified | ai.meta.comblogs.nvidia.com |
| Storage | Pure Storage FlashArray Storage 175 petabytes of Pure Storage FlashArray | Pure Storage | - | Reported | ai.meta.comblogs.nvidia.com |
| Storage | Pure Storage FlashBlade Storage 10 petabytes of Pure Storage FlashBlade | Pure Storage | - | Reported | ai.meta.com |
AI datacenter metrics
Figures beyond an accelerator count, each with who stated it. What can and cannot be compared across AI datacenters.
| Metric | Value | As of | Basis | Note | Source |
|---|---|---|---|---|---|
| Accelerators | 16,000 accelerators | 2022-01-24 | Design target | Meta: phase 2 raises RSC from 6,080 to 16,000 GPUs, full build mid-2022; target stated at launch, phase 1 was 760 DGX A100 | ai.meta.com |
| Accelerators | 6,080 accelerators | 2022-01 | Operator stated | Phase one: 760 DGX A100 systems, 6,080 A100 GPUs; Meta targeted 16,000 GPUs later in 2022, which no fetched source shows was reached | ai.meta.comblogs.nvidia.com |
The read
Meta's AI research cluster, announced in January 2022 as 760 NVIDIA DGX A100 systems and 6,080 GPUs, with a stated plan to reach 16,000 GPUs by mid-2022. Meta and NVIDIA independently give the 760 nodes and 6,080 GPUs, which is why the tier is verified; the phase two size is a target and no fetched source shows it was reached, so it is not recorded as a count. Meta lists 175 PB of Pure Storage FlashArray, 10 PB of FlashBlade and 46 PB of cache in Penguin Computing Altus systems. Meta and NVIDIA disagree on the InfiniBand link speed, so no speed is recorded (see open questions). Neither source says where RSC is hosted, so it sits at Meta's undisclosed site and the USA is only the operator's home country. Both sources quote AI-precision peaks, which are targets or vendor figures rather than HPL, so no FLOPS are recorded.
Timeline
- 2022-01 Announced source
Site & facility
Where sources disagree
We record conflicts instead of picking a winner quietly. All open questions.
- InfiniBand link speed: NVIDIA Quantum 1600 Gb/s InfiniBand two-level Clos fabric against NVIDIA Quantum 200Gb/s InfiniBand network. Meta and NVIDIA describe the same fabric at different speeds, so no speed is recorded and no fabric part edge is added.
Change history
- 2026-10-04 AI datacenter metrics: Meta Research SuperCluster (RSC)
- 2026-09-20 Added Meta Research SuperCluster, verified against Meta and NVIDIA
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 |
|---|---|---|
| ai.meta.com | could not be fetched | - |
| blogs.nvidia.com | names system, part and count | NVIDIA A100 (in DGX A100), Pure Storage FlashArray, Pure Storage FlashBlade · counts: 6080 |
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.