Thanks. I've updated the article with clarifications and correct TPU numbering. The text already mentions TPU v1 is inference only, and I think it's useful to include as context.
In the table, the figures are for a single die in the wafer. This is to make a meaningful comparison with the other chips listed (there is a table footnote for this). The 15 KW is the power consumption of the whole wafer (a detail I think was mentioned in the Hot Chips presentation). Why are they clearly working with FP16? Are there any public details on this?