News
Google expands TPU server production to Inventec with manufacturing planned in Texas
Thursday, January 1, 2026 at 04:44 AM
Google is reportedly expanding its TPU-based server production to Inventec, shifting away from its current reliance on in-house assembly and Celestica. Inventec is expected to produce these AI servers at its new facility in Texas, supported by a $1 billion capital expenditure plan intended to double its AI server manufacturing capacity by 2026.
Context
Google is diversifying its AI infrastructure by expanding Tensor Processing Unit (TPU) server assembly to Inventec, challenging Celestica’s historical dominance in L10 production. While Google previously relied on in-house assembly or Celestica, this shift aims to scale capacity for its proprietary chips as internal model training demand surges. This move transitions Inventec from a motherboard supplier to a full-system integrator for Google’s high-performance hardware.
To support this ramp, Inventec has unveiled a $1 billion capex plan for 2026, doubling its budget to expand AI server output. Production will be centered at a new manufacturing facility in Texas, backed by an initial $85 million investment to localize the supply chain for North American data centers and mitigate tariff risks.
This expansion reflects a strategic effort by hyperscalers to de-risk supply chains and meet a projected 42% surge in TPU shipments through 2026. For investors, it positions Inventec as a primary beneficiary of the AI hardware cycle while signaling increased competition for incumbents like Celestica.
Related Companies
Inventec
2356