Amazon.com Inc (NASDAQ:AMZN), Microsoft Corp (NASDAQ:MSFT), Alphabet Inc (NASDAQ:GOOG), Oracle Corp (NYSE:ORCL, XETRA:ORC), and CoreWeave (NASDAQ:CRWV) are expected to remain key beneficiaries of accelerating AI compute demand, according to a Jefferies analyst note following attendance at Google Cloud Next.

The analysts wrote in a note that the event reinforced a clear message that AI infrastructure demand continues to outstrip supply, supporting a constructive view on hyperscalers into upcoming earnings.

“Overall, we continue to be incrementally more positive on AI demand vs our checks in early April,” they wrote.

Ahead of first quarter financial results, investors are focused on cloud growth, backlog trends, and any changes to capital expenditure or capacity plans. Jefferies said it is incrementally more positive on hyperscalers, with the strongest cloud upside skew seen at Google Cloud, followed by AWS and Azure.

Jefferies highlighted elevated expectations for AWS, though the firm maintained a high conviction view on Amazon. Microsoft is increasingly being assessed on Microsoft 365 and Copilot traction rather than Azure beat size, while Alphabet remains positive on trajectory but may be constrained by valuation.

A major theme was continued AI compute scarcity, with tight supply across GPU generations including Hopper and Blackwell, alongside limited spot availability for older chips. Industry feedback suggested a growing shift toward committed capacity contracts as hyperscalers lock in supply amid persistent shortages.

Capital expenditure across cloud providers is also expected to remain elevated, with Jefferies estimating roughly 64% growth in 2025 and 58% in 2026, driven by ongoing AI infrastructure buildout.

The analysts also pointed to inefficient enterprise usage of AI infrastructure, with low GPU utilization despite rising adoption. At the same time, demand is being fueled by competitive pressure and “fear of missing out” on AI deployment cycles.

Further, they noted that enterprises are increasingly seeking model agnostic tooling to manage rising token costs and optimize workloads across models.

“A common focus that started to emerge is that enterprises wanted model agnostic harnesses, as they want to optimize token consumption and have control over what model is used, which is tough if they don’t own the harness,” the analysts wrote.

While frontier models from Anthropic, OpenAI, and Google remain widely used, firms are looking for flexibility in routing tasks to lower cost alternatives where possible.