Anthropic GPU cost per hour jumps to $5.65 as rental prices hit $13

The cost pressure stems from a structural gap between what AI labs want and what the market can supply. Labs like Anthropic are trying to rent as many accelerators as possible on multi-year contracts that fix prices around $2 to $4 per hour. Only when that long-term capacity is exhausted must they turn to the short-term burst market, where prices have surged to $13 per hour on short-term contracts.

This is a unit-economics problem with limited levers: a doubling of GPU cost must be met by a doubling of model efficiency, a doubling of price to the customer, or a combination of the two. The price lever is constrained by a price war across frontier and mid-tier models, which makes raising token prices perilous because any lab that hikes rates cedes share. That leaves efficiency and contract structure as the main tools.

On the supply side, 40% to 50% of data center projects are delayed, billions of dollars of GPUs sit idle in warehouses waiting for power, and transformer and turbine lead times are measured in years. The forward-looking argument is that the 2026 race to find GPUs at any price gives way in 2027 to a contest over duration matching — aligning the length of customer commitments with the length of compute leases.