USA · operational · ai training

IBM Vela

Operated by IBM .

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	- IBM Research design, deployed as virtual machines inside IBM Cloud VPC	IBM IBMNYSE	-	Reported	research.ibm.com
Accelerator	- NVIDIA A100 80GB, eight per node, 360 nodes at launch (2,880 GPUs); expanded about 2x in 2023	NVIDIA NVDANASDAQ	2,880 estimated	Reported	research.ibm.com nextplatform.com
CPU	- Two 2nd Gen Intel Xeon Scalable (Cascade Lake) per node; 3rd Gen (Ice Lake) in the 2023 expansion nodes	Intel INTCNASDAQ	-	Reported	research.ibm.com arxiv.org

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
Rack power	40 kW	2024-07	Operator stated	Vela racks have two redundant 20 kW PDUs (40 kW available) and hold six 8-GPU A100 servers drawing up to 6 kW each, throttled if a PSU fails	arxiv.org

The read

IBM Research's cloud-native AI training supercomputer, housed inside IBM Cloud and online since May 2022 for the IBM Research community. Each node has eight 80GB A100 GPUs on NVLink and NVSwitch, two second-generation Xeon Scalable CPUs, 1.5TB of DRAM and four 3.2TB NVMe drives, on 100G Ethernet in a two-level Clos with no oversubscription; IBM later added RoCE and GPU-direct RDMA. The Next Platform's February 2023 write-up counts 360 nodes at six per rack (2,880 A100s, about 27.9 PFLOPS theoretical peak by its arithmetic, which is not an operator figure and is not recorded), and IBM's December 2023 update and its 2024 infrastructure paper say the system was expanded roughly two-fold in 2023 with Ice Lake nodes. IBM publishes no current total, so the 2,880 figure describes the launch configuration only. IBM does not name a data center or region, so USA records only the operator's home country. It has never submitted an HPL result.

Timeline

2022-05 First operational source

Change history

2026-10-04 AI datacenter metrics: IBM Vela
2026-10-02 IBM Vela added (IBM Cloud A100 training supercomputer, 360 nodes at launch)

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
nextplatform.com	does not name this system	-
research.ibm.com	does not name this system	-

Checked 2026-10-09 by pnpm verify. Re-run it and the table changes with the web.