Supercomputers/Systems/Project Ceiba
USA · under construction · ai training
Project Ceiba
Operated by Amazon .
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 exclusively on AWS, using the NVIDIA DGX Cloud architecture | Amazon | - | Reported | aws.amazon.com |
| Accelerator | NVIDIA B200 SXM Accelerators CUDA (Blackwell) · TSMC 4NP · 1,000 W 20,736 NVIDIA Blackwell (B200) GPUs deployed as NVL72 racks | NVIDIA | 20,736 reported | Reported | nvidianews.nvidia.comaws.amazon.com |
| CPU | NVIDIA Grace CPUs Armv9 (Neoverse V2) · 72C · 3.1 GHz · TSMC 4N 10,368 NVIDIA Grace CPUs | NVIDIA | 10,368 reported | Reported | nvidianews.nvidia.com |
| Interconnect | AWS Elastic Fabric Adapter Interconnect Cloud fabric · Up to 3.2 Tb/s per instance · Low, but not deterministic Fourth-generation AWS Elastic Fabric Adapter, up to 800 Gbps per Superchip in the 2024 release | Amazon | - | Reported | nvidianews.nvidia.comaws.amazon.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 | 20,736 accelerators | 2024-03 | Design target | NVIDIA release: supercomputer being built with 20,736 B200 GPUs on 10,368 Grace CPUs in GB200 NVL72 racks; no delivery date given | nvidianews.nvidia.com |
The read
An AI supercomputer that AWS is building for NVIDIA's own research, hosted exclusively on AWS and run on the NVIDIA DGX Cloud architecture. It was announced at AWS re:Invent 2023 and re-specified on 18 March 2024 as 20,736 Blackwell GPUs attached to 10,368 Grace CPUs in GB200 NVL72 racks, on fourth-generation Elastic Fabric Adapter networking, with a claimed 414 exaflops of AI compute (a low-precision figure, so no FLOPS are recorded). The two sources disagree on units: NVIDIA's release says 20,736 B200 GPUs and 10,368 Grace CPUs, while AWS's page says 20,736 GB200 Superchips; the GPU-to-CPU ratio of a GB200 points to NVIDIA's reading, which is the one recorded here. They also disagree on the EFA rate (800 Gbps per Superchip in the 2024 release, 1,600 Gbps on the current AWS page). AWS's page still describes the machine in the present continuous, so status is under construction; neither source gives a site, a go-live date or a power figure, so USA records only the operator's home country. Both pages are the partners' own statements, so this stays reported.
Timeline
- 2023-11 Announced source
Change history
- 2026-10-04 AI datacenter metrics: Project Ceiba
- 2026-10-02 Project Ceiba added (AWS and NVIDIA, 20,736 Blackwell GPUs, under construction)
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 |
|---|---|---|
| aws.amazon.com | names system, part and count | NVIDIA B200 SXM, NVIDIA Grace · counts: 20736, 10368 |
| nvidianews.nvidia.com | names system, part and count | NVIDIA B200 SXM, NVIDIA Grace · counts: 20736, 10368 |
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.