Should an AI vendor sell multi-year minimum commitments to match its compute leases?

THE SHORT ANSWER

Yes, and track what the match does to the buyer. Tomasz Tunguz's November 3, 2026 reading of Anthropic's S-1/A shows $5.0 billion of burst compute at $7 to $13 per GPU-hour, blended cost up 82% in two quarters, and Q2's $974 million operating profit becoming a $567 million Q3 loss; his fix is to sell a multi-year minimum commitment at a locked-in rate that matches the GPU lease underneath it. That closes the vendor's duration mismatch by moving it to the customer. In the per-seat sunset I ran, 12 accounts paid their committed minimums while usage ran under them, the dashboard read on plan, and two of them left when the legacy surface went away. Against a cost per task that just fell about 40%, a three-year minimum is that gap with a term on it. Keep utilization against commitment per account per month, write the pass-through decision before the minimum is sold, and run Jason Lemkin's top-1% check in tasks rather than dollars.

Yes, because the lease is the vendor's problem and the match solves it. Then watch the minimum, because the match moves the mismatch to the buyer, and a buyer paying a minimum with usage underneath is churn with a renewal date.

The filing figures as Tunguz and Lemkin report them, and my 12 accounts, are in Anthropic Bought $5B of Burst Compute and Flipped to a Loss. Match the Lease, Then Watch the Minimum.. The pass-through decision is from Anthropic's S-1 Has No Gross Margin Line. Your AI Product Needs One..

SOURCES

THE LONG VERSION

RELATED ANSWERS

Last reviewed 2026-11-05 · 1 min read