The first entry is always the pin's source. Overall confidence is a weighted average of how firmly each reference supports the start and end times used above; a reference counts half as much for every 180 days older than the newest.
Dated Google[2][3] announcement; commercialization milestones are recorded separately in the sourced timeline.
Something missing or wrong? Say it in your own words: a link that backs this pin up, a different start or end date and why, or a fact it lacks or gets wrong. The AI checks it against this pin's sources, searches for better ones, and adds any page that backs you up. The pin's own sources still count most. A picture that shows something else, or shows it badly, is looked at too, and moved down or replaced.
| Training performance / dollar |
|---|
| Up to 2x TPU v4 |
| 2023 maker claim; workload dependent[1][2][3] |
| Inference performance / dollar | Up to 2.5x TPU v4 | 2023 maker claim; workload dependent[1][2][3] |
|---|
| Product | Published context |
|---|---|
| H100 SXM | Hopper GPU with 80 GB HBM and 3.35 TB/s bandwidth; framework and precision choices affect comparisons.[4][5] |
| A100 | Ampere training and inference accelerator shipping from May 2020; supports multi-instance GPU partitioning.[4][5] |
Product context, not a benchmark ranking. GPU, TPU and cloud-instance peak rates and memory units may differ.
