The primary constraint on building new AI data centers isn't acquiring land or power, but securing "powered shells"—fully energized buildings with cooling and components. Supply chains for transformers and a severe shortage of accredited electricians are the true limiting factors.
A futures market for GPU compute is not viable yet because the product isn't fungible. The performance of an identical H100 chip varies significantly between cloud providers based on their proprietary software stack and operational excellence, measured by metrics like "goodput" and "MFUs."
The recent explosion in enterprise AI spending was triggered by the release of effective, specialized tools like coding assistants that provided clear ROI to specific professionals like developers. This suggests future growth hinges on targeted, vertical-specific applications, not just general-purpose models.
CoreWeave reports a financial services client backlog approaching $10 billion. These firms, like Jane Street, are not just using AI labs' models but are building their own proprietary systems, contracting directly for massive GPU capacity and diversifying the customer base beyond hyperscalers and AI labs.
CoreWeave pioneered financing its GPU fleet through special purpose vehicles (SPVs) that isolate assets and contracts. This de-risked structure achieved an investment-grade rating and attracted a new class of conservative investors, like insurance companies, unlocking billions in previously inaccessible capital.
Contrary to the belief that AI hardware becomes obsolete quickly, older GPUs like A100s will have a long depreciable life. As companies optimize costs, they'll use model routing to send simple queries to older, cheaper hardware, extending its utility for six to eight years.
A major shift in behavior among top AI labs is their move from three-year to five-year take-or-pay contracts for GPU infrastructure. They are locking in capacity at massive scale for longer durations, signaling extreme confidence in sustained, long-term demand for compute.
Despite major tech companies developing their own AI chips, CoreWeave's clients exclusively demand Nvidia hardware. This is attributed to the mature CUDA software platform, which provides an efficient, scalable, and reliable ecosystem that competitors have been unable to replicate.