Big Technology Podcast · Oct 20, 2026 · 47m · 40m saved
With Sundar Pichai, CEO of Google and Alphabet
Photo via Wikipedia
Google's new Gemini 4 Argon model logged roughly 25 trillion tokens of usage in its first 72 hours — about three times the pace Gemini 3.0 saw at launch. Google is still "capacity constrained" serving it.
Some had read Google's recent moves as quietly ceding frontier-model development to lean on faster, "good enough" models, a framing floated via Meta's Muse. Sundar Pichai says the capacity crunch is the opposite signal: demand for the frontier model outran what Google built to serve it.
"We are at early stages of the most consequential technology ever. This is not like fiber, it's not dark — we are using it." — Sundar Pichai
Sundar Pichai, CEO of Google and Alphabet, told Alex Kantrowitz on the Big Technology Podcast that frontier work compounds: it is what lets Google see emerging capabilities first, and in an "era of recursive self-improvement," frontier models will increasingly help build the next generation of models.
The full interview is covered here so you can skip it. 47 minutes of audio, 7 minutes of reading.
Here are the 6 insights that matter.
Key Takeaways
Gemini 4 Argon logged about 25 trillion tokens of usage in its first 72 hours, roughly 3x Gemini 3.0's pace at launch, and Google is still "capacity constrained" serving it
A customer asking Google for 2027 compute today is "incredibly difficult" to satisfy, because today's capacity is the output of commitments made back in 2024
Sundar says essentially every company in the industry is short relative to what it planned for
Google's cloud business is deliberately forgoing short-term revenue to subsidize frontier AI research
Google is already running early recursive-self-improvement loops, where AI systems help researchers build parts of the next models
Sundar wants independent, industry-wide AI safety auditors, similar to financial accounting standards, with regulation eventually playing a role
Space-based data centers are 2 to 3 years out for Google, against roughly a decade for quantum computing
1. Never Left the Frontier
Sundar pushed back hard on the idea that Google had quietly stepped back from frontier-model development. He pointed to Gemini 4 Argon's benchmark results and its early usage numbers as the evidence.
Sundar says Gemini 4 Argon is beating OpenAI's and Anthropic's public models on coding, math and science benchmarks
"In the first 72 hours since we launched, it's almost up to 25 trillion tokens used, which is about 3x what Gemini 3.0 had in the first 72 hours." — Sundar Pichai
Demand outran supply immediately, Sundar said, leaving Google "capacity constrained" serving its own new model. He argues frontier work compounds rather than standing apart from Google's product business: it is what lets Google see emerging capabilities first, before they show up anywhere else.
"I think the frontier models will be better at developing the next set of models in an era of recursive self-improvement." — Sundar Pichai
