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 profitable
CoreWeave said on its Q2 earnings call that it has signed contracts for Nvidia A100 GPUs extending to 2029, about nine years after the chip’s 2020 debut. The company reported $2.58B quarterly revenue (+112% YoY) and a $104B revenue backlog. CoreWeave cited power and cooling limits that keep older air-cooled systems in use.
How this was made
The 30-second read
Why it matters
CoreWeave’s disclosed 2029 A100 contract and commentary on prior-generation pricing provide a concrete datapoint that older Nvidia GPUs can remain profitable due to datacenter power and cooling constraints, supporting backlog durability narratives.
Market read
A long-dated A100 contract into 2029, paired with reported backlog and revenue growth, is a tangible update to AI infrastructure revenue durability and GPU obsolescence assumptions.
What to watch
The article cites facility power/cooling limits and utilization claims, but does not address whether A100 performance-per-watt or customer workload mix could force earlier-than-2029 retirements, nor does it quantify the contract’s dollar value.
Background
The piece frames a debate about GPU useful life and depreciation assumptions, citing Michael Burry’s concerns that hyperscalers may be understating depreciation by stretching GPU life.
Ticker impact
CoreWeave disclosed on its Q2 earnings call a contract for Nvidia A100 GPUs running into 2029, extending contracted revenue on 2020 silicon.
Near-term sentiment likely positive for CRWV on improved backlog durability, though magnitude depends on market focus on AI capex intensity versus utilization economics.
The article provides specific, time-extended contract duration (into 2029) plus CoreWeave’s reported backlog and revenue growth, which are direct inputs to revenue durability expectations.
CoreWeave signed a contract for Nvidia A100 GPUs that runs into 2029, with CEO remarks that prior-generation A100 pricing is at or above prior levels.
NVDA could see modest positive read-through from evidence of sustained A100 utilization and pricing, but the impact may be capped if it signals customers delaying upgrades to newer racks.
While the contract is with CoreWeave, the article does not quantify NVDA revenue impact from the specific A100 contract, and it frames constraints that may limit Blackwell deployment rather than expand total GPU demand.
Market effects
Reinforces a datacenter infrastructure constraint story (power and cooling) that can extend the economic life of older AI accelerators, affecting upgrade cadence assumptions across the AI hardware supply chain.
No specific regional demand signal; implications are tied to facility design constraints rather than geography.
Could influence global AI capex planning by hyperscalers and GPU service providers if legacy GPU utilization remains profitable longer than expected.
Counterpoint
Long-dated A100 contracting may reflect constrained ability to deploy newer liquid-cooled systems, which could ultimately cap growth in higher-margin next-gen GPU rack deployments.
Key entities
- companyCoreWeave
GPU cloud provider disclosing a contract for Nvidia A100 GPUs running into 2029 and reporting Q2 revenue, backlog, and power capacity metrics.
- companyNvidia
GPU supplier whose A100 architecture is referenced in CoreWeave’s long-dated contract and pricing/utilization commentary.
- personMike Intrator
CoreWeave CEO who disclosed the 2029 A100 contract and discussed pricing for prior-generation SKUs.
- personColette Kress
Nvidia CFO who previously defended A100 utilization and pricing durability.
- personNitin Agrawal
CoreWeave CFO who added commentary on fleet renewals and capacity coming up for renewal.
