Google’s parent company Alphabet is managing its fleet of AI accelerators with a blunt priority: artificial general intelligence research comes first, cloud customers come second. The disclosure came during the company’s Q2 2026 earnings call, where CEO Sundar Pichai described TPU allocation as a deliberate hierarchy, not a free market.
“In terms of allocating our TPUs … our first priority is making sure we are allocating what we need to compete at the frontier in terms of AGI development,” Pichai told analysts. Search, YouTube, and Google Cloud get what remains.
The strategy reveals a fundamental tension inside Alphabet. The company is simultaneously the world’s largest consumer of its own custom TPU chips, devouring compute for internal AI research, and a public cloud provider that rents those same accelerators to enterprise customers. When supply is tight, internal AGI wins.
Alphabet reported $119.8 billion in quarterly revenue, up 24% year-over-year, and operating income of $40.8 billion, a 34% gain. Google Cloud revenue surged 82% to $24.75 billion, with profit jumping 214% to $8.8 billion. But the cloud business is growing so fast that Alphabet cannot manufacture enough hardware to satisfy both its own research agenda and its customers’ demand.
The company raised its full-year capital expenditure forecast to between $195 billion and $205 billion, up from an earlier $180 billion to $190 billion range. Even at that unprecedented spending level, roughly equivalent to the GDP of a small country, Alphabet is still supply-constrained. CFO Anat Ashkenazi described the company’s plan to “expand the use of third-party capacity in Q3 as a bridging strategy while we build out more internal capacity.”
Translation: Google, the company that designs its own AI chips, is renting compute from other providers to keep cloud customers happy.
The financial consequence is stark. Alphabet posted negative free cash flow of $5.9 billion for the quarter, the first time that has happened since 2004, when Google was still a private company building out its original server infrastructure. Investors sent shares down roughly 4% in after-hours trading.
Within Google Cloud, Pichai said compute is allocated first to Vertex AI, Gemini Enterprise, data analytics, and cybersecurity workloads. The company has started selling TPU access to some external customers, but allocation is calibrated by a lifetime-value calculation: Pichai noted that even short-term losses on third-party compute make sense when the multi-year deal ROI is “very, very attractive.”
The AI Mode feature in Google Search, which generates AI-powered answers rather than link lists, is driving an overall increase in search query volume. Pichai said engineering and hardware optimizations have brought the cost of each AI Mode response to its lowest level since launch, even as the system handles more complex queries.
The cloud backlog, contracts signed but not yet fulfilled, reached $514 billion, a figure that reflects both massive enterprise demand and the supply constraints that prevent Alphabet from delivering faster.
For enterprise customers evaluating Google Cloud against AWS and Azure, the message is clear: during periods of hardware shortage, Google will prioritize its own race toward AGI over their compute needs. Whether third-party bridging capacity will be enough to prevent customer defection is an open question.
Sources: Google is hoarding TPUs to develop Artificial General Intelligence (The Register, July 23, 2026); Alphabet Q2 2026 earnings call: Remarks from our CEO (Google Blog, July 22, 2026)