Google-parent Alphabet's $190 billion capex plan has a 'warning' for Nvidia, as CEO Sundar Pichai says: Key part of this investment is ...

Sundar Pichai, at Google I/O 2026, reiterated the company’s aggressive plan to spend close to $200 billion in AI infrastructure, signaling the company’s push in the semiconductor segment.

This massive budget represents a six-fold increase from the $31 billion the company spent in 2022. While a significant portion of this capital will fund AI data centers and model training, Pichai issued what appears to be a direct warning to dominant chip suppliers like Nvidia by emphasising Google's self-reliance.“It’s incredible to see the pace of innovation rolling out across our products. Supporting all of this scale for our users, while also serving enterprises and developers around the world, requires massive investments in infrastructure,” Pichai said at the developer conference, announcing that Alphabet expects its infrastructure spending to reach approximately $190 billion this year.“We’ve been investing for now and for the future. In 2022, we were spending $31 billion annually in capex. This year, we expect that number to be about six times that, approximately $190 billion. A key part of this investment is our custom silicon,” he said.

Alphabet may reduce reliance on third-party chips

For the past several years, tech giants have engaged in a fierce bidding war for Nvidia's graphics processing units (GPUs) to power the generative AI boom. However, Alphabet's latest financial roadmap highlights a concerted push to route around this dependency using its own line of Tensor Processing Units (TPUs).

Google recently unveiled its 8th generation of custom silicon, introducing a dual-chip architecture split by specific workloads: TPU 8t (Training) and TPU 8i (Inference).TPU 8t (Training): Pichai explained that this processor is optimised specifically for large-scale model pretraining. It delivers nearly three times the raw computing power of Google's previous generation. “With JAX and Pathways, our training is no longer constrained by the limits of a single, massive data center.

Instead, we can now seamlessly distribute training across multiple sites, scaling training across more than 1 million TPUs globally. This gives us the ability to create the largest training cluster in the world. For model builders, this means training larger, more capable models in weeks rather than months,” Pichai said.TPU 8i (Inference): This chip is built exclusively to handle live user queries and run active AI applications.

Designed with a strict focus on reducing latency, it ensures that AI responses return to users instantly.“In addition to speed, we’re also thinking about scaling sustainably. Both chips are more energy efficient, delivering up to two times better performance-per-watt,” Pichai added.

Other key numbers that Sundar Pichai shared

The $190 billion infrastructure injection is designed to support a massive surge in software adoption across Google’s developer and consumer ecosystems.Pichai revealed that more than 8.5 million developers now build applications using Google’s AI models each month. The company's model application programming interfaces (APIs) currently process roughly 19 billion tokens per minute. On the enterprise side, over 375 major Google Cloud customers processed more than one trillion tokens each over the past year.Moreover, According to Pichai, the core Gemini app has crossed 900 million monthly active users, more than doubling its user base since last year. Furthermore, the company's integrated search tools are seeing widespread deployment; AI Overviews now reach over 2.5 billion monthly users, while the dedicated AI Mode has surpassed one billion monthly active users.