Google’s new hirings in 2025 had 20 per cent ex-Google employees, reveals study

Noam Shazeer, who left in 2021 to co-found Character.AI, returned to Google DeepMind this year as part of a staggering $2.7 billion licensing deal.

John Casey, Google’s head of compensation, recently noted that AI researchers are increasingly drawn back by the company’s unmatched computational scale
John Casey, Google’s head of compensation, recently noted that AI researchers are increasingly drawn back by the company’s unmatched computational scale

In a year defined by a multi-billion dollar talent war, Google is finding that its best new recruits might actually be its old ones. According to a new report from CNBC, approximately 20% of the AI software engineers hired by Google in 2025 were ‘boomerang’ employees. i.e., former staffers who had previously left the company. The surge in returning talent marks a strategic turnaround for the tech giant, which spent much of 2023 and 2024 struggling with brain drain as top researchers fled to rivals like OpenAI, Meta, and Anthropic.

Ex-Google employees return to Google

The most high-profile example of this trend is the return of Noam Shazeer, a co-author of the seminal “Attention Is All You Need” paper that birthed the Transformer architecture. Shazeer, who left in 2021 to co-found Character.AI, returned to Google DeepMind this year as part of a staggering $2.7 billion licensing deal.

While the official reason for the payout was to license Character.AI’s technology, insiders suggest the primary motivation was to bring Shazeer back to co-lead the Gemini AI project alongside Jeff Dean. The deal reportedly netted Shazeer nearly $1 billion, signaling Google’s willingness to spend unprecedented sums to correct past mistakes in talent retention.

Why are they coming back

While rivals like Meta are reportedly dangling $100 million signing bonuses, Google is winning the war through “infrastructure envy.” John Casey, Google’s head of compensation, recently noted that AI researchers are increasingly drawn back by the company’s unmatched computational scale:

Custom Silicon: Access to the latest Tensor Processing Units (TPUs) allows researchers to train models with 2–3x better efficiency than general-purpose hardware.

Gemini 3 momentum: The successful launch of Gemini 3 has restored confidence that Google can once again lead in reasoning and multimodal performance.

Institutional scale: For many, the ability to deploy AI models across 15 products with over 500 million users (including Search and Gmail) provides a sandbox no startup can match.

Flattening the ‘Big Ship’ bureaucracy

To ensure these returning stars stay, Google has undergone a radical internal restructuring. CEO Sundar Pichai has eliminated more than one-third of middle-management roles, specifically targeting supervisors of small teams to accelerate decision-making.

Co-founder Sergey Brin has come out of retirement too, emerging as a key recruiter, reportedly spending nearly every day at the office and personally calling former employees to persuade them that the “bureaucratic bloat” of the past has been replaced by a “startup-like” urgency.

As Alphabet’s stock surges 60% this year, the “Boomerang” strategy has effectively reframed Google not as a legacy giant, but as the premier destination for the engineers who built the AI revolution in the first place.

This article was first uploaded on December twenty-one, twenty twenty-five, at seven minutes past two in the afternoon.