Sundar Pichai says Gemini 5 will surpass Argon, backing a bigger compute bet

Google just shipped Gemini 4 Argon a few weeks ago. Its CEO is already talking up the next model.

On October 21, 2026, Sundar Pichai said Google’s research indicates Gemini 5 will be significantly better than Argon. He also tied that confidence to a push for more computing infrastructure, which is the expensive part of this whole enterprise.

What Pichai actually said

Pichai, who runs both Google and parent company Alphabet, made the remarks in an interview. He framed the Gemini 5 claim as grounded in extensive internal research and benchmarks rather than a hunch.

The more consequential part of his comments concerned money. Pichai pointed to long-term revenue contracts as the justification for Google speeding up its spending on compute.

Pichai also described demand as strong across both consumer and enterprise customers. He paired the growth talk with an emphasis on responsible deployment, signaling Google intends to be careful about how it releases high-capability systems.

The bar Gemini 5 has to clear

Argon is not a low hurdle. Google launched Gemini 4 Argon on September 30, 2026, and made it generally available on October 13, 2026.

The model was pitched at complex, real-world work in fields like software engineering, cybersecurity, and finance.

In its first 72 hours, Argon processed approximately 25 trillion tokens, which was three times what Gemini 3.0 handled over the same opening stretch.

Argon also set a leading mark on DeepSWE v1.1, a benchmark for software engineering tasks, with a score of 77.9%.

Why the compute angle is the real story

Pichai’s answer is that Google’s investments rest on contracted revenue rather than speculative forecasts. Contracts give a company visibility into future income, which makes heavy upfront spending easier to defend to shareholders.

It also helps explain why Google is talking about Gemini 5 so soon. Confidence in the next model’s performance is part of the case for building the capacity it will need.

What this means for Google and the AI race

For investors watching Alphabet, the early Argon data offers a concrete read on demand. Approximately 25 trillion tokens in three days suggests customers are not just experimenting but running real workloads.

Gemini 5’s superiority over Argon is, for now, a projection based on Google’s own research. Until the model ships and faces outside testing, the claim stays a forecast.

Pichai’s emphasis on responsible deployment is worth watching. As models grow more capable in sensitive areas like cybersecurity, how Google gates access could shape both adoption speed and regulatory attention.