CoreWeave proves Nvidia's aging AI GPUs from 2020 can generate profit nine years after deployment, signs A100 contracts into 2029 — power constraints and legacy infrastructure keep old GPUs prof
CoreWeave signed a contract for Nvidia A100 GPUs that runs into 2029, nine years after the Ampere-based SKU debuted, CEO Mike Intrator disclosed on the company's second-quarter earnings call on Monday. The company reported $2.58 billion in quarterly revenue, up 112% year over year, and a $104 billion revenue backlog that excludes more than $25 billion in new commitments booked since July. In remarks made during the earnings call, Intrator said, pricing for prior generation SKUs is at or above where it was years ago."
Go deeper with TH Premium: AI and data centers
- Photonics and high-speed data movement is the next big AI bottleneck
- The data center cooling state of play
- Massive AI data center buildouts are squeezing energy supplies
- Ultra Ethernet: The data center interconnection of tomorrow
Michael Burry accused hyperscalers in November of understating depreciation by $176 billion between 2026 and 2028 by stretching GPU useful-life assumptions to five or six years against a chip cadence that now delivers a new Nvidia architecture roughly every year. Concerns that accelerating GPU obsolescence could unravel AI financing predate his estimate, and an unnamed Google architect put datacenter GPU service life at one to three years back in 2024.
Intrator told CNBC at the time that a batch of H100s coming off an expired contract was immediately rebooked at 95% of the original price, and Nvidia CFO Colette Kress countered Burry directly, saying A100s sold six years ago still run at full utilization. The 2029 deal stretches that record further, putting contracted revenue on 2020 silicon beyond even the six-year depreciation schedules defended by hyperscalers.
An air-cooled Nvidia DGX A100 system draws 6.5kW at maximum load and fits comfortably in legacy data center halls designed for roughly 20kW per rack. Nvidia's current GB200 and GB300 NVL72 racks draw 120kW to 140kW and require direct-to-chip liquid cooling, roughly six times what those older facilities can feed and cool. Blackwell therefore can't move into the halls where Ampere is currently deployed without a rebuild of power delivery and cooling infrastructure.
That mismatch is helping to keep A100 fleets in service regardless of how gracefully or not silicon itself ages. The energized, air-cooled capacity holding them has no higher-value use, and renting nine-year-old GPUs is a better alternative to leaving it in the dark.
CoreWeave's contracted power grew to 3.7 GW in Q2 and stood at 4.2 GW as of Monday, against just 1.5 GW online, so customer commitments already cover nearly triple the capacity the company can deliver. CFO Nitin Agrawal added on the call that capacity coming up for renewal represents a very limited share of CoreWeave's fleet, with older-generation ASPs at or above levels from a year ago.