At Google IO 2026’s opening keynote, Sundar Pichai’s calm and cool demenaour was in direct contrast to all the gargantuan numbers he was dropping to underscore just how fast the whole world was adopting AI.
One after another, the Google CEO belted out a sequence of stats that paints a vivid picture of AI adoption moving at a velocity that’s hard to grasp in any single frame. How AI is being used by average users, developers, and all the cool products being made.
Pichai also spoke about how AI is eating into Google’s infrastructure capex, and how the company’s pouring tens of billions into keeping the AI party going. Here are some broad strokes of some of the most humbling stats and figures Pichai spoke about during his Google IO 2026 keynote.
The cleanest illustration of the scale of AI being used came from Pichai’s point on token volume – the fundamental unit of data that AI models process. “Two years, we were processing 9.7 trillion tokens a month across the surface, that’s a huge number,” Pichai said. “Last year at IO, that grew to about 480 trillion tokens. And fast forward to today, that number has jumped seven times to 3.2 quadrillion tokens per month.”
Also read: Google I/O 2026: What is Gemini Spark, the AI assistant that works even when your laptop is off
He acknowledged the absurdity of this moment, absolute bonkers in terms of the unit itself. “Never imagined I’d say quadrillion in an IO keynote, but here we are,” Pichai said with his charismatic smile.
That growth isn’t only consumer-facing. On the developer side, the numbers are similarly outsized. “Over 8.5 million of you are now building new apps and experiences with our models monthly. And our model APIs are now processing around 19 billion tokens per minute,” Pichai said. “Over the past 12 months, over 375 customers each processed more than 1 trillion tokens, representing incredible demand for AI across the industry.”
“We now have 13 products with over a billion users each. Five of those have more than 3 billion users,” Pichai noted, framing Gemini as the main reason this adoption curve hasn’t flattened.
Google Search, the company’s foundational product, has seen its impact as well. “AI Overviews now has over 2.5 billion monthly users. And AI mode has been a revelation, our biggest upgrade to search ever. People love it,” he said. “In just a year, it’s already surpassed 1 billion monthly users. When people use our AI-powered features in search, they use search more.”
Also read: Google I/O 2026: Gemini 3.5 Flash, AI video tool and major Gemini app upgrades announced
The Gemini app itself has more than doubled in a year. “Last year at IO, the Gemini app had 400 million monthly active users. Today, we have surpassed 900 million, more than doubling in a year. In that same time, daily requests have grown over seven times.”
One feature category – image generation, specifically – will make you sit up straight in your seat, if it doesn’t make your jaw drop. “More than 50 billion images have been generated with our Nano Banana models,” Pichai said. “It was a breakout star this past year.”
Numbers like these are impossible without capital to allow for such crazy growth, and Pichai underlined the figure in his opening keynote. “In 2022, we were spending $31 billion annually in capex. This year, we expect that number to be about six times that at approximately $180-$190 billion.”
A majority of that spend is going into custom silicon, according to Pichai. The proprietary distributed training and inference Google has built around its eighth-generation TPUs – TPU-80 for training and TPU-80i for inference.
Pichai said, “we can now seamlessly distribute training across multiple sites, scaling across more than one 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.”
Perhaps the most telling indicator of speed in Pichai’s speech was how fast Google itself is consuming the models it builds. Working with its agent-first development platform Antigravity and the newly announced Gemini 3.5 Flash, Pichai said, “in March, we were processing half a trillion tokens a day internally for our developers. We’ve been doubling every few weeks, and now we are processing more than 3 trillion tokens a day.”
A single demo demonstrated the ROI curve. Inside Antigravity, a team of 93 sub-agents running on Gemini 3.5 Flash built a working operating system from scratch in about 12 hours – making over 15,000 model requests and processing 2.6 billion tokens, with a total compute cost of under $1,000 in API credits. “Multi-day engineering efforts are collapsing into hours, if not minutes,” Antigravity lead Varun said on stage.
For anyone tracking numbers around AI adoption, the Google IO 2026 keynote offered a useful set of reference points. The curve is steep, the AI spend is unprecedented, and the numbers Google is now flexing openly shows just how much of AI is being used all over by all kinds of users.
Also read: Google I/O 2026: 5 ways Gemini is getting a new agentic makeover