Google to Expand Third-Party AI Compute Use in Q3

Demand for compute to train and run generative AI models has outpaced the rate at which Google can add servers and data centers, constraining both Google Cloud customers and the company’s own products. Specialized “neocloud” providers such as Nebius and CoreWeave rent out GPU-heavy infrastructure, giving hyperscalers a faster, if more expensive, way to bridge shortages caused by limited chips, power and ready-to-use data-center capacity.

On July 22, Alphabet Chief Financial Officer Anat Ashkenazi said Google would expand its use of third-party capacity in the third quarter while continuing to build internal infrastructure. The stopgap should help the company serve more customers and capture demand, but will put modest near-term pressure on margins. Alphabet also lifted its 2026 capital-spending forecast to $195 billion-$205 billion from $180 billion-$190 billion. The comments sent Nebius and CoreWeave shares up about 5% and 4%, respectively, in after-hours trading.