To put 3.2 quadrillion in perspective, that’s 3,200,000,000,000,000 tokens. Google is now processing that volume every single month across its AI products, a figure CEO Sundar Pichai dropped during the Google I/O 2026 keynote on May 20.

That number represents a 7x increase from last year.

The numbers behind the number

The 3.2 quadrillion figure spans token inference across Google’s entire AI-powered surface area: Gemini, Search with AI Overviews, YouTube, Workspace, Cloud APIs, and multimodal data processing across images, video, and audio.

Here’s the growth trajectory in plain terms. In April 2024, Google was processing approximately 9.7 trillion tokens per month. By May 2025, that had jumped to 480 trillion. By October 2025, the company was approaching 1.3 quadrillion. Now, just seven months later, it’s sitting at 3.2 quadrillion.

The user metrics paint a similar picture. Gemini, Google’s flagship AI application, now has 900 million monthly active users. AI Overviews, the feature that puts AI-generated summaries at the top of search results, serves more than 2.5 billion users globally. And 8.5 million developers are building with Google’s models each month.

Pichai also noted that Google now has 13 products each boasting over a billion users.

How Google keeps the lights on at this scale

Processing 3.2 quadrillion tokens monthly requires hardware that doesn’t exist on the open market. Google’s answer is its custom-built Tensor Processing Units, or TPUs, purpose-engineered silicon designed specifically for the matrix math that powers AI inference and training.

Google has been building TPUs since 2016.

What this means for investors and the broader AI market

Google’s stock has more than doubled since last year’s I/O conference, according to Pichai’s presentation.

When 2.5 billion people are using AI Overviews and 900 million are engaging with Gemini monthly, the demand side of the equation is very real.

Going from 480 trillion to 3.2 quadrillion in one year means Google roughly 6.7x’d its inference volume.

With 8.5 million developers on Google’s platform, the network effects are compounding.