Nvidia A100 GPUs Defy Depreciation as CoreWeave Signs Contracts Through 2029

CoreWeave Signs A100 Contracts Into 2029

CoreWeave has signed contracts for Nvidia A100 GPUs running through 2029, nine years after the Ampere generation debuted in 20201

1

. CEO Mike Intrator disclosed the agreement during the company's second-quarter earnings call, where CoreWeave reported $2.58 billion in quarterly revenue, up 112% year over year, and a $104 billion revenue backlog1

1

. Intrator noted that pricing for prior-generation GPUs remains at or above levels from years ago, demonstrating sustained demand for older AI hardware1

1

.

Power Constraints Keep Legacy Infrastructure Viable

The extended hardware utility of Nvidia AI GPUs stems partly from infrastructure limitations at existing data centers. An air-cooled Nvidia DGX A100 system draws 6.5kW at maximum load and fits comfortably in legacy infrastructure designed for roughly 20kW per rack1

1

. In contrast, Nvidia's current GB200 and GB300 NVL72 racks draw 120kW to 140kW and require direct-to-chip liquid cooling, roughly six times what older facilities can feed and cool1

1

. Blackwell systems cannot move into halls where Ampere is currently deployed without a complete rebuild of power delivery and cooling infrastructure, making the long-term value of older GPUs more attractive than leaving air-cooled capacity idle1

1

.

CUDA Platform Extends GPU Lifespan

Jensen Huang recently emphasized that CUDA gives developers and Nvidia engineers a common platform to continually upgrade Ampere, Hopper, and Blackwell throughout their useful lives2

2

. He stated that CUDA makes Nvidia computing versatile, and that versatility makes it fungible, driving utilization and extending durability to make Nvidia compute a productive asset that is rentable, durable, and financeable2

2

. This software-driven approach allows AI infrastructure to become more valuable over time rather than steadily depreciating2

2

. Continued efforts to optimize new AI models and upgrades within the CUDA ecosystem have made it possible to run AI workloads on the same amount of compute rather than upgrading to newer hardware, which requires much higher investments3

3

.

Hardware Deprecation Debate Intensifies

The 2029 deal stretches contracted revenue on 2020 silicon beyond even the six-year depreciation schedules defended by hyperscalers1

1

. Michael Burry accused hyperscalers in November of understating hardware deprecation 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 year1

1

. An unnamed Google architect put datacenter GPU service life at one to three years back in 20241

1

. However, Nvidia CFO Colette Kress countered directly, saying Nvidia A100 GPUs sold six years ago still run at full utilization1

1

.

Dual Strategy for AI Chips Market

Nvidia is advancing a nuanced message that customers should embrace its newest AI systems while recognizing that older Nvidia hardware can remain productive and economically valuable for years2

2

. The company is trying to accomplish two seemingly conflicting goals: persuade customers to buy its latest AI chips while assuring them that older hardware will continue holding value for years2

2

. As AI adoption expands beyond hyperscalers and frontier model developers, more cost-conscious enterprises may find older AI GPUs sufficient for many inference and production workloads2

2

. This approach could expand AI adoption without undermining the premium pricing of newest chips2

2

.

Market Dynamics Support Extended Utility

Demand for AI computing infrastructure continues to outstrip supply, meaning customers often value access to GPUs whether they're the latest Blackwell systems or older Ampere-based hardware2

2

. 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 deliver1

1

. CFO Nitin Agrawal added 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 ago1

1

. Using older GPUs comes at a time when GPU shortages and price hikes have affected almost every segment of the tech world, with even much older GPUs such as the Volta-based V100 series seeing a bump in prices3

3

.