Google Reports 75% of Its New Code Is Now AI-Generated
Sundar Pichai disclosed on Tuesday that three-quarters of all new code is now generated by artificial intelligence and then reviewed and approved by human engineers. Notably, this figure has tripled since late 2024, when AI accounted for just over 25% of new code.
During a keynote at a major cloud event in Las Vegas, he described this evolution as part of a broader shift into the “agentic Gemini era.” Meanwhile, the growth trajectory has been strikingly steep. In 2024, AI contributed to just over a quarter of new code, and by late 2025, that share had climbed to about 50%. Now, within only a few months, it has surged to 75%.
Moreover, he highlighted a recent internal code migration that finished six times faster than similar efforts a year earlier. This improvement came from agentic AI workflows, where engineers coordinate autonomous AI agents instead of writing every line manually. As a result, teams can execute complex tasks more efficiently. He described developers as “firing off agents and accomplishing incredible things,” while also noting that a macOS app prototype was built in just days using an internal platform.
Productivity Gains Without Workforce Cuts
Despite the rapid rise of AI in coding, the shift is framed as a boost to productivity rather than a replacement for engineers. In fact, the company continues to plan for ongoing hiring. Therefore, AI is positioned as a tool that expands what teams can achieve instead of reducing headcount.
At the same time, investment levels reflect strong confidence in this direction. Earlier this year, the company projected capital expenditures between $175 billion and $185 billion for 2026. Compared to 2025, that amount is roughly double. Most of this spending, however, will go toward AI infrastructure such as data centers, servers, and specialized chips.
Industry-Wide Momentum
Although this company leads in adoption, it is not alone in advancing AI-driven development. For instance, another major tech firm has set internal goals for certain teams to generate more than 75% of their code using AI tools by mid-2026.
Additionally, a recent industry survey found that 42% of committed code across the broader ecosystem is already AI-generated. Looking ahead, projections suggest that figure could reach 65% by 2027.
Consequently, a noticeable gap remains between leading firms and the wider industry. This difference highlights both the advantages of scale and the urgency among competitors to accelerate adoption.
