Jefferies analysts reported that AI infrastructure demand continues to outstrip supply, with tight availability of GPUs across Hopper and Blackwell generations. CoreWeave was cited alongside Amazon, Microsoft and Alphabet as a key beneficiary of these persistent shortages.

The note projects that capital expenditure across cloud providers will rise by roughly 64% in 2025 and 58% in 2026, driven by ongoing AI infrastructure buildout. Hyperscalers are increasingly entering committed capacity contracts to secure scarce compute resources.

Enterprises are shifting toward long-term capacity agreements and seeking model-agnostic harnesses to optimize token consumption and control workloads. This trend underscores growing competitive pressure and the desire to manage costs across diverse AI models.