What is CoreWeave?
CoreWeave is a cloud services provider specializing in GPU-accelerated workloads, particularly for AI and machine learning.
Company timeline
- Oct 2025 - Brad Gerstner said Nvidia has promised to buy any CoreWeave service availability that Nvidia cannot sell to other customers.
- May 2026 - Kyle Corbitt said CoreWeave can achieve latency down to about 30% of what one gets from using a frontier model, with similar or higher quality.
- May 2026 - Kyle Corbitt said CoreWeave delivers at least an order of magnitude improvement in cost per token, often more.
Where it appears in the record
Every line below is attributed to a named speaker.
Neoclouds like CoreWeave cannot finance gigawatt-scale data centers independently because they lack investment-grade credit, making hyperscaler balance-sheet backstops a structural necessity for large leases.
“For a company like coreweave if they wanted to build a gigawatt of data center capacity or most likely to lease it that would be more likely for them they're just not going to be able to do that. They don't have the credit profile to do that, right? Because if you seen a gigawatt of capacity, then you know the data center operator needs to get that financed. typically very high loan to cost. You know, it's typical real estate. you need a construction loan that's not going to go through if you don't have an investment grade signature.”Jeremie Beaufils · 9 Jul 2026
Custom fine-tuned models can achieve latency as low as 30% of frontier model latency, with similar or higher quality.
“We can typically get you know, latency down to about 30% of what you get from using a frontier model with again similar or usually higher quality than what you were getting from the frontier model.”Kyle Corbitt · 1 May 2026
Custom fine-tuned models deliver at least an order-of-magnitude improvement in cost per token versus frontier models, often more.
“We're talking you know, order of magnitude at least improvement in cost per token, often times more than that.”Kyle Corbitt · 1 May 2026
CoreWeave's fine-tuning platform: delivering 30% frontier latency and 10x+ cost-per-token reduction for enterprise deployments, making it a key infrastructure player for production RL and SFT workloads.
“We can typically get you know, latency down to about 30% of what you get from using a frontier model with again similar or usually higher quality than what you were getting from the frontier model.”Kyle Corbitt · 1 May 2026
CoreWeave has not secured additional contracted power since Q3 2025.
“Since Q325 they haven't really been able to secure more.”Jeremie Beaufils · 9 Jul 2026
Nvidia has reportedly agreed to absorb unsold compute capacity, acting as a backstop buyer for AI infrastructure it cannot place elsewhere.
“Nvidia has promised to buy any of service availability that they can't sell to anyone else.”Brad Gerstner · 14 Oct 2025