Supercomputers/Systems/Google TPU v3 pod
USA · operational · ai training
Google TPU v3 pod
Also known as TPU v3-2048
Operated by Alphabet at Google Cloud (undisclosed location) .
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 |
|---|---|---|---|---|---|
| Accelerator | - Google TPU v3: 1,024 chips (two TensorCores each) in a full v3-2048 pod | Alphabet | 1,024 reported | Reported | docs.cloud.google.com |
| Interconnect | - TPU inter-chip interconnect, 2D torus, 340 TB/s all-reduce bandwidth per pod | Alphabet | - | Reported | docs.cloud.google.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 | 1,024 accelerators | 2026-10 | Operator stated | Google Cloud docs: full v3 pod has 1024 chips; one documented pod configuration, not a site | docs.cloud.google.com |
The read
The full Cloud TPU v3 pod (accelerator type v3-2048) per Google's documentation: 1,024 chips with two TensorCores each (2,048 cores) on a 2D torus, each chip 123 TFLOPs bf16 with 32 GiB of HBM2 at 900 GBps and a measured power of 123, 220 and 262 W (minimum, mean, maximum). Google gives 126 PFLOPs bf16, 340 TB/s of all-reduce bandwidth and 6.4 TB/s of bisection bandwidth per pod. The page says the Cloud TPU API now receives only bug fixes and security updates, with support end dates listed as 'not yet set', so the status is operational as of the fetch date and could change. A product configuration with no stated location, so the site is Google's undisclosed one and USA is the operator's home country. Chip power is not multiplied into a system draw because host, network and cooling power are not given. Predecessor of the v4 supercomputer (google-tpu-v4-supercomputer); the v2 pod (512 chips per the v2 page) is not recorded because that page gives no per-chip figures.
Timeline
No dated events recorded.
Site & facility
Change history
- 2026-10-17 Added Google TPU v3 1,024-chip pod (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 |
|---|---|---|
| docs.cloud.google.com | names system only | - |
Checked 2026-10-18 by pnpm verify. Re-run it and the table changes with the web.