Google Unveils 8th-Gen TPUs, Plans $80 Billion Data Center Buildout for Agentic AI
Updated
Updated · Network World · Jul 23
Google Unveils 8th-Gen TPUs, Plans $80 Billion Data Center Buildout for Agentic AI
3 articles · Updated · Network World · Jul 23
Summary
Google said its data centers now process seven times more AI tokens than a year ago, prompting an accelerated infrastructure upgrade centered on new TPU-8t training and TPU-8i inference systems.
Agentic AI is driving the buildout because inference workloads can jump 50x to 100x versus traditional chat use cases, pushing Google to cut per-transaction costs by nearly half on its latest platforms.
TPU-8i boosts on-chip SRAM to 384MB—triple the prior generation—and raises HBM by 50%, while Google says TPU-8t can scale to more than 1 million chips through its Virgo network.
Google is also reworking its software stack for agents, turning GKE into an agent-native orchestration layer and adding storage systems including Managed Lustre 10T, which it says is 10 times faster than last year.
The $80 billion expansion underscores how Google is tying custom chips, liquid-cooled data centers and compatible TPU-GPU software to rising enterprise demand for large-scale AI services.
As Google spends billions on AI, will this infrastructure race truly benefit users or just the tech giants?
Can our power grids handle the AI boom's massive energy needs, or are we heading for a crisis?
When autonomous AI agents make costly mistakes, who will ultimately be held responsible?
Google's $80 Billion AI Bet: Powering the Agentic AI Era with Custom TPUs and Unprecedented Infrastructure Investment (2026)
Overview
In June 2026, Alphabet made a bold move by raising $80 billion in equity, supporting its massive $180–$190 billion capital expenditure plan for the year, with a major focus on advancing AI infrastructure. This investment comes as tech giants like Microsoft, Meta, and Amazon are also pouring over $700 billion into similar efforts, with industry-wide AI spending expected to surpass $1 trillion by 2027. Berkshire Hathaway increased its stake in Alphabet, reflecting strong confidence in the company’s strategy. These actions highlight Alphabet’s commitment to leading the next wave of AI innovation and staying ahead in the rapidly evolving technology landscape.