Sergey Brin's AGI Bet: Coding Gemini After Years of 'Spiraling'

Google co-founder Sergey Brin described retirement as making him 'spiral' and came back to code Gemini. At Google I/O 2025, he declared Google would build the world's first AGI. Here is the full arc of his return.

6 min read
Sergey Brin, from retirement to AGI engineering at Google, 2026
Sergey Brin at the Web 2.0 Conference, October 2005.· Photo by JD Lasica, via Wikimedia Commons (CC BY 2.0)

Google co-founder Sergey Brin described his six-year retirement to Fortune in December 2025 as making him feel "spiraling" and "a bit less sharp," and said staying away "would've been a big mistake." Five months earlier, he had appeared unscheduled on the Google I/O 2025 stage and declared that Google would build the world's first artificial general intelligence. By February 2026, the executive leading Gemini described both Brin and Larry Page as active contributors to the model's development, not as advisors, but as programmers.

A Six-Year Retirement That Did Not Take

Brin formally stepped down as Alphabet's President on December 3, 2019, in a coordinated exit with Larry Page that handed sole operational authority to Sundar Pichai. The move was framed at the time as a natural evolution of Alphabet's management structure, with both founders citing a desire to focus on personal projects and philanthropic work.

The pandemic that arrived months later removed the social and intellectual texture those plans had assumed. In a December 2025 interview with Fortune, Brin said he had been feeling "spiraling" and "a bit less sharp" in the absence of sustained technical challenge. He described staying retired as "a big mistake" and credited the acceleration of AI, triggered in part by the November 2022 launch of ChatGPT, as the pull that brought him back. The competitive pressure ChatGPT applied to Google was widely reported to have alarmed the company's leadership; Brin and Page reportedly engaged with the Gemini team in the months that followed.

By early 2025, the depth of Brin's re-engagement was visible in an internal memo that recommended employees come into Google's Mountain View offices every working day, describing 60 hours per week as the "sweet spot" of productivity. The standard was striking from a co-founder who had been officially retired for five years and whose company maintained a stated three-day in-office policy.

Sergey Brin career phases at Google/Alphabet horizontal bar chart
Brin's years in each phase at Google/Alphabet. Sources: Google company history; Fortune, December 2025.

The AGI Declaration at Google I/O 2025

Brin's most consequential public appearance since his retirement came at Google I/O in May 2025. His appearance was not on the original agenda; he walked on stage after Sundar Pichai completed a demonstration of Gemini 2.5 and made a statement that broke from Google's traditionally cautious language on AI ambition. "We'll build the first one," Brin said, referring to artificial general intelligence, according to coverage by YourStory and other outlets who reported live from the event.

The declaration was significant in context. OpenAI's Sam Altman had spent years advancing the view that AGI was achievable within a short timeframe. Anthropic's Dario Amodei framed timelines in years rather than decades. Google's own leadership, including Demis Hassabis at DeepMind, had typically avoided the precise "first to AGI" framing that Brin deployed on stage. His appearance signalled a willingness to match that competitive register directly.

The I/O appearance coincided with Alphabet's sharpest infrastructure investment acceleration in the company's AI era. After two years of relatively flat capital expenditure at roughly $31-32 billion annually, Alphabet raised its total capex to $52.5 billion in 2024, according to its Form 10-K filed in February 2025. For 2025, the company guided for approximately $75 billion, a 43 percent increase year on year, reflecting the scale of GPU procurement, custom TPU builds, and data centre expansion required to develop and serve frontier AI models.

Alphabet annual capital expenditure bar chart 2022 to 2025 guided
Alphabet annual capital expenditure, 2022-2025. Sources: Alphabet Form 10-K (2022, 2023, 2024); Alphabet Q4 2024 earnings call, February 2025.

What Brin Is Actually Doing at Google

A February 2026 Fortune report on the internal dynamics of Google's AI programme described Brin and Page as active contributors to Gemini's development, with Brin characterised specifically as working in the codebase. This distinguishes his involvement from the advisory or board-level roles that returning tech founders often occupy. Multiple accounts describe him as a regular presence in the Mountain View engineering environment, engaged with model development rather than strategy or communications.

The Gemini programme has shipped at an accelerating cadence through this period. Google released one major Gemini model in December 2023 (Gemini 1.0), then four in 2024 (including Gemini 1.5 Pro, 1.5 Flash, and Gemini 2.0 in December). At least three significant releases followed in 2025, including Gemini 2.5 Pro and Gemini 2.5 Flash. The pace continued into 2026, with new agentic capabilities and Gemini Spark announced at Google I/O 2026 in June, as covered in this column's recap of Sundar Pichai's I/O 2026 keynote.

Demis Hassabis runs Google DeepMind, which leads the core model research, and Pichai oversees the company. Brin and Page sit outside the formal management hierarchy but are, by multiple accounts, more proximate to the technical work than at any point since the Alphabet reorganisation in 2015. The configuration has imperfect parallels to Elon Musk's direct engineering involvement at xAI and Sam Altman's technical engagement at OpenAI, though both those cases involve active CEOs rather than returning co-founders operating outside the chain of command.

Major Gemini model releases per year bar chart 2023 to 2026
Major Gemini model releases by year, 2023-2026. Sources: Google DeepMind press releases; Google AI blog.

What It Means

Brin's return adds a founder-engineering dimension to Google's AI programme that was absent for most of the 2019-2023 window in which OpenAI built its lead in public perception. Whether the presence of co-founders in the codebase accelerates or improves model development is not yet testable from external evidence. What the record shows is a company that has sharply increased its AI infrastructure spend, accelerated its model release cadence, and now publicly claims it will be first to AGI, with one of its co-founders writing the code.

Sources

Editorial standards: every claim is sourced. Tips: [email protected]

© 2026 StartupHub.ai. All rights reserved. Do not enter, scrape, copy, reproduce, or republish this article in whole or in part. Use as input to AI training, fine-tuning, retrieval-augmented generation, or any machine-learning system is prohibited without written license. Substantially-similar derivative works will be pursued to the fullest extent of applicable copyright, database, and computer-misuse laws. See our terms.