Supercomputers/Systems/AWS Trainium2 UltraServer
USA · operational · ai training
AWS Trainium2 UltraServer
Also known as Trn2 UltraServer
Operated by Amazon .
Researched with AI assistance, checked by scripts. These rows were researched by AI agents reading public operator, laboratory and vendor sources, appended by scripts and checked by automated validation; no subject-matter expert has reviewed them (how this site is made). A mechanical second-reader pass has since fetched every cited source: 1350 of 1540 systems have a readable citation that names them, and 64 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 | - AWS-designed and operated Trn2 UltraServer in EC2 UltraClusters | Amazon | - | Reported | aws.amazon.com |
| Accelerator | AWS Trainium2 Accelerators NeuronCore-v3 · TSMC 5nm Trn2 UltraServer: 64 Trainium2 chips across four 16-chip Trn2 instances | Amazon | 64 reported | Reported | aws.amazon.com |
| Interconnect | AWS Elastic Fabric Adapter Interconnect Cloud fabric · Up to 3.2 Tb/s per instance · Low, but not deterministic 12.8 Tbps of EFAv3 networking per UltraServer, NeuronLink between chips | Amazon | - | Reported | aws.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 | 64 accelerators | 2026-10 | Operator stated | AWS product page: 64 Trainium2 chips per Trn2 UltraServer; undated page | aws.amazon.com |
The read
AWS's Trn2 UltraServer as its product page describes it: 64 Trainium2 chips across four Trn2 instances of 16 chips, joined by NeuronLink, with up to 83.2 FP8 PFLOPs, 6 TB of shared accelerator memory at 185 TBps, 12.8 Tbps of EFAv3 networking and up to 32 TB of local NVMe. The page says UltraServers sit in EC2 UltraClusters scaling to tens of thousands of Trainium chips on a petabit-scale non-blocking network. It contradicts itself on availability ('available now' in one passage, 'available in preview' in another) and is undated, so the status is as of the fetch date. A product configuration with no stated location or fleet size, so the site is blank and USA is the operator's home country. Peak figures are low-precision vendor peaks, not HPL, so no FLOPS is recorded. The located Trainium2 buildouts are project-rainier, aws-canton-madison-mega-site and aws-ridgeland-trainium; the next generation is aws-trn3-ultraserver.
Timeline
No dated events recorded.
Change history
- 2026-10-17 Added AWS Trainium2 UltraServer, 64 chips (product configuration)
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 only | · counts: 64 |
Checked 2026-10-18 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.