CoreWeave's chief executive, Michael Intrator, said the company is still booking contracts for Nvidia GPUs first released in 2020. Some of those agreements run through 2029, according to comments reported by Yahoo Finance. Intrator described the pricing on this older hardware as "full freight," meaning customers are paying full rates rather than discounted terms.
CoreWeave built its business by renting out Nvidia graphics processing units to companies training and running artificial intelligence models. The firm has grown from a niche cloud provider into one of the most closely watched infrastructure companies in the AI boom. Its stock listing and expansion have been tied directly to demand for Nvidia's chips.
Nvidia has released several generations of GPUs since 2020, moving from its Ampere architecture to Hopper and then Blackwell chips. Each new generation typically offers faster performance and better efficiency for AI workloads. Under normal market conditions, older chips would be expected to command lower prices as newer options become available.
Intrator's comments suggest that dynamic has not played out as expected. Customers appear willing to commit to years-long contracts for older-generation hardware at unreduced prices. That points to computing capacity remaining scarce relative to demand, even as newer chips reach the market.
Long-duration contracts also give CoreWeave visibility into future revenue. Booking capacity through 2029 signals that some customers are planning AI infrastructure needs many years in advance. It also suggests they see value in locking in supply now rather than waiting for newer chips or lower prices later.
The broader AI infrastructure market has been defined by tight supply since the technology's rapid growth began. Data center operators, cloud providers, and chipmakers have struggled to keep pace with demand for training and inference capacity. CoreWeave's position as a major GPU lessor puts it close to the center of that supply crunch.
Intrator's remarks were reported by Yahoo Finance. No additional financial figures, contract values, or customer names were disclosed alongside the comments. The statement centers on booking activity and pricing behavior rather than specific deal terms.
Market Impact
If accurate, sustained full-price demand for older Nvidia GPUs would reinforce the view that AI computing capacity remains undersupplied industry-wide. That dynamic has supported elevated valuations for infrastructure providers like CoreWeave and has kept Nvidia's chip sales strong across multiple product generations.
Long-term contracts extending to 2029 could also signal how customers are budgeting for AI workloads well beyond the current hardware cycle. Investors watching CoreWeave and similar cloud providers may treat this kind of forward booking activity as an indicator of demand durability, though it does not by itself guarantee future revenue realization or contract fulfillment.
The comments from CoreWeave's chief executive add to a broader picture of tight AI infrastructure supply, with demand apparently strong enough to sustain full pricing on years-old hardware.
Frequently Asked Questions
What did CoreWeave's CEO say about older Nvidia GPUs?
Michael Intrator said CoreWeave is still signing contracts for 2020-era Nvidia GPUs, with some agreements running through 2029, and that customers pay full price for that older hardware.
Why does demand for older GPUs matter?
Typically, older chips see falling demand and prices once newer generations arrive. Continued full-price bookings suggest AI computing capacity remains scarce relative to overall demand.
What does 'full freight' mean in this context?
It refers to customers paying the full standard price for GPU capacity, without discounts usually associated with older or previous-generation hardware.
What is CoreWeave's role in the AI infrastructure market?
CoreWeave rents out Nvidia GPUs to companies that train and run AI models, positioning it as a key supplier of computing capacity in the AI boom.