Google's parent company Alphabet is reserving its most powerful AI chips, tensor processing units (TPUs), primarily for its own artificial general intelligence (AGI) research rather than selling all available capacity to cloud customers, CEO Sundar Pichai revealed during the company's second-quarter earnings call. This matters because it signals Google's long-term strategic priority is competing at the frontier of AI development, even as it faces surging demand for cloud computing that it can only partially meet by buying in third-party capacity as a stopgap.

Pichai told analysts that TPU allocation is led by what is needed "to compete at the frontier in terms of AGI development," followed by core products such as Search, YouTube and Cloud services like Vertex and Gemini Enterprise. Google Cloud revenue rose 82 percent year-on-year to $24.75 billion, with profit up 214 percent to $8.8 billion, and the division now holds a $514 billion backlog of unfulfilled orders. Search revenue grew 17 percent and YouTube ads 13 percent, while CFO Anat Ashkenazi said planned AI infrastructure spending has risen to between $195 billion and $205 billion this year, partly because a "supply-constrained environment" is forcing Google to rent extra third-party compute to bridge the gap.