It's completed and been brought into production. It's online. Dodger's transcript source had a typo:
"We recently completed a 10,000 GPU cluster of h100s. We think, probably
bringing it into operation faster than anyone's ever bought that much compute per unit time into production"
Timestamped audio: