Google has set an internal target to increase AI capacity by 1,000 times by 2030. The company must double its AI serving capacity every six months to meet demand. This aggressive plan contrasts with CEO Sundar Pichai’s public warnings about irrationality in the AI market.
According to CNBC, infrastructure VP Amin Vahdat presented the mandate at a November 6 all-hands meeting. The directive targets the age of inference, where models like Gemini 3 Pro require massive compute power for reasoning tasks and code execution. Vahdat warned that the competition in AI infrastructure is the most critical and most expensive part of the AI race.
Custom Silicon Strategy
Google plans to achieve this growth without proportional cost increases. The company aims to deliver 1,000 times more capability for essentially the same cost and power levels. The strategy depends on the Ironwood TPU chip, which recently entered general availability.
Efficiency and Custom Chips
The Ironwood TPU claims a 10x peak performance improvement over the v5p. It offers 2x performance per watt compared to the previous Trillium generation. Google is also using new Arm-based Axion CPUs for general-purpose workloads. This frees up power and thermal capacity for AI tasks.
The company uses a co-design philosophy, integrating software directly with hardware architecture. Research from Google DeepMind informs chip design. Vahdat noted that achieving these goals will not be easy but collaboration will make it possible.
Market Tensions and Investment
The expansion continues despite growing skepticism about AI return on investment. In a BBC interview, Pichai conceded there are elements of irrationality in current market valuations. Yet Alphabet has raised its 2025 capital expenditure forecast to $93 billion, with a significant increase planned for 2026.
Employees questioned leadership about the tension between soaring spending and bubble fears. Pichai defended the strategy by citing the company’s strong balance sheet. He argued that the risk of underinvesting and becoming irrelevant is worse than overinvesting.
The Big Four tech companies are projected to spend over $380 billion on infrastructure this year. Google is betting it can outlast competitors in a capital-intensive war. The outcome depends on whether premium features like Deep Think reasoning can generate revenue faster than hardware depreciates.
