Google begins shipping custom AI processors directly to customer data centers

Google began recognizing revenue from its Tensor Processing Unit system sales directly to customer data centers in the second quarter of 2026, marking a shift in how it monetizes its custom silicon. Rather than selling access only through Google Cloud, the company can now install TPU systems inside customer facilities, allowing enterprises tighter control over sensitive workloads and data. This direct distribution creates a new revenue route and positions Google to compete for spending that would otherwise flow to Nvidia. The financial impact remains early-stage, with Alphabet expecting to recognize only a small portion of existing TPU-system sales agreements during 2026, with the bulk expected in 2027 as inventory investments support the ramp. Inventory climbed to $9.99 billion from $2.44 billion year-over-year.

1. Google began recognizing revenue from TPU-system sales during Q2 2026 after delivering systems directly to customer data centers for the first time.
2. Customers can now install Google's custom AI infrastructure inside their own facilities, providing tighter control over sensitive workloads and data.
3. Alphabet expects to recognize only a small portion of revenue from existing TPU-system sales agreements in 2026, with the majority expected in 2027.
4. Inventory climbed to $9.99 billion from $2.44 billion as the company builds inventory to support system sales.
5. Google Cloud product revenue comes primarily from TPU-system sales, contributing to the segment's 82% year-over-year growth.