Alphabet to Sell AI Chips Externally as Cloud Backlog Hits $462 Billion

Alphabet is transforming its AI hardware strategy by selling its custom Tensor Processing Units (TPUs) directly to external customers for the first time. The move comes as the company faces a massive surge in AI demand, evidenced by a contracted revenue backlog for Google Cloud that has nearly doubled to approximately $462 billion.

In its Q1 2026 results, Alphabet reported total revenue of $109.9 billion, a 22% increase year-over-year. This growth was driven heavily by Google Cloud, which saw revenue jump 63% to approximately $20.03 billion. This expansion is mirrored by the adoption of its AI models; CEO Sundar Pichai noted that Gemini first-party models are now processing more than 16 billion tokens per minute via direct API use, representing a 60% increase over the previous quarter.

A Shift in Silicon Strategy

Historically, Google's TPUs were available exclusively as a rented service within the Google Cloud platform. Under the new strategic shift, Alphabet will sell these chips to a select group of external clients, including AI research labs, high-performance computing (HPC) entities, and capital markets firms. These customers will be permitted to run the hardware within their own private data centers rather than relying solely on Google's infrastructure.

This transition allows Alphabet to compete more directly with Nvidia's hardware sales model. By offering the silicon as a retail product, Google provides a path for large-scale users to avoid cloud vendor lock-in, which is often a primary concern for firms with strict regulatory compliance or specific operational requirements.

Market Implications

The scale of the $462 billion backlog—up from $240 billion in the previous quarter—signals a future revenue stream that requires aggressive infrastructure scaling. By diversifying its distribution model, Alphabet can capture a larger share of the AI infrastructure market without being entirely limited by the physical speed of its own data center construction.

Transforming from a pure cloud provider into a silicon vendor reduces the bottleneck of internal capacity. It allows Alphabet to monetize its hardware engineering expertise directly while ensuring that the massive demand for AI compute is met, regardless of where the chips are physically housed.

The Road Ahead

Industry observers will now watch how this shift affects the competitive landscape between cloud hyperscalers and independent hardware vendors. While the move opens a new revenue stream, it remains to be seen how Alphabet will manage the support and maintenance of TPUs operating outside its controlled environment. Additionally, the company continues to scale its capital expenditure to support the underlying growth of the Gemini ecosystem.