Alphabet reported second-quarter cloud revenue of $24.8 billion, an 82% year-over-year increase, as its Google Cloud unit continued to expand rapidly. While the headline growth underscored the unit’s momentum, a less-publicized development may prove more consequential: revenue from Tensor Processing Units (TPUs) for the first time.
The TPUs, custom AI accelerators developed in-house, had previously been used exclusively within Alphabet’s own infrastructure or leased to clients via Google Cloud. In the latest quarter, however, some units were sold to external data center operators, marking the company’s formal entry into chip manufacturing and monetization. Financial details for the TPU revenue stream were not disclosed, but the sales were included within the broader cloud segment.
The move reflects a broader industry shift, as major technology firms increasingly design their own AI processors to reduce reliance on third-party suppliers such as Nvidia. Alphabet joins peers like Microsoft and Amazon in vertically integrating chip production to support AI workloads. Industry forecasts suggest substantial growth ahead: Technavio projects the global AI chip market will expand at a compound annual rate of over 24% through 2030, while Global Market Insights estimates annual sales could reach $1.1 trillion by 2035.
Alphabet’s positioning in AI infrastructure is further strengthened by its cloud backlog, which grew to $514 billion at the end of June, up from $50 billion in the prior quarter. Despite this growth, the company’s core search and advertising businesses continue to dominate revenue and operating income. TPU sales remain a small fraction of Google Cloud’s total, with supply constrained by internal demand and limited external allocations.
The monetization of TPUs signals a strategic pivot for Alphabet, aligning with long-term bets on AI-driven infrastructure. While the financial impact is currently marginal, the move positions the company to capture a share of the expanding AI chip market as demand for specialized hardware accelerates.
