Timely delivery is key to AI data-center operators' success. Microsoft said at least $5B of GPU-as-a-service capacity was late in Q1. CoreWeave had no known new Q1 delivery due, and its known delay was too small; Nebius and IREN reported delivering theirs, making them almost certainly not responsible. Nscale's Texas timing fits, but it said all Microsoft tranches were delivered; the supplier remains unknown. x.comÂ



