Comment by gpm

> This goes so far as to have companies or business units share wafers for prototyping runs so as to minimize cost per device

Can this be done in production? Is there a chance that the portion of the wafer cerebras.ai can't fit their giant square in is being used for production of some other companies chips?