Alphabet Expands AI Chip Strategy by Allowing Customers to Install TPUs in Private Data Centers
Alphabet (GOOGL) is pivoting its artificial intelligence hardware strategy by allowing select customers to install its custom Tensor Processing Units (TPUs) directly into their own data centers. Previously, TPU access was restricted to Google's cloud environment. The new TPU 8tx model is specifically designed for regulated industries and government entities that require high-level data confidentiality and cannot migrate sensitive information to the public cloud. While this move expands Alphabet's addressable market—competing against giants like Nvidia (NVDA)—the financial impact is currently in a transition phase. Alphabet has seen an inventory surge to $13.3 billion as it builds these systems, and most revenue from these hardware sales is not expected to be recognized until 2027. However, Google Cloud's rapid growth, including a third-quarter revenue of $30.7 billion, suggests strong momentum in the AI infrastructure sector.
Topics: AI chips Data centers Google Cloud Hardware sales Tensor Processing Units (TPU)