Alphabet said 75% of all new code at Google is now generated by AI and then approved by engineers, a sharp increase from 50% last fall and more than triple the level Google disclosed in October 2024, when it said more than a quarter of new code was AI-generated.
The new figure was disclosed by CEO Sundar Pichai in Google’s Cloud Next 2026 remarks, where he framed the company as moving beyond assisted coding into agentic workflows. In practical terms, Google is no longer presenting AI as a helper layered onto engineering work. It is presenting AI as the default engine behind most fresh code creation, with humans increasingly positioned as reviewers, orchestrators, and governors of output quality.
On Alphabet’s February 2026 earnings call, management said AI investments were already translating into business performance across the company, while also guiding for 2026 capital expenditures of $175 billion to $185 billion. The spending range follows $91.4 billion in 2025 capex, including $27.9 billion in the fourth quarter alone, with most of that spending aimed at technical infrastructure.
The financial backdrop is large enough to make the coding statistic economically relevant rather than cosmetic. Alphabet reported $113.8 billion in Q4 2025 revenue, up 18% year over year, with operating income of $35.9 billion, net income of $34.5 billion, operating cash flow of $52.4 billion, and free cash flow of $24.6 billion. Google Cloud revenue rose to $17.6 billion in the quarter, while Cloud operating income more than doubled to $5.3 billion and Cloud backlog reached $240 billion after increasing 55% sequentially and more than doubling from a year earlier.
Those numbers help explain why management is leaning so heavily into AI-generated software. Google said one internal software migration task completed using these more autonomous AI workflows was finished six times faster than before.
Bloomberg reported this week that Google has been trying to unify coding initiatives internally as concern grows that rivals, especially Anthropic, are gaining ground in AI coding tools. Separately, Business Insider reported that some Google DeepMind teams have been allowed to use Anthropic’s Claude Code, even as most of Google remains focused on Gemini-based tooling, creating internal friction over tool quality and access.
That tension makes the 75% disclosure read partly like a progress report and partly like a rebuttal. Publicly, Google is telling investors and developers that its internal AI adoption is already deep, measurable, and accelerating. Unofficially, the surrounding reporting suggests Google is still under pressure to prove that its own coding stack can match the strongest tools in the market, especially as agentic coding moves from novelty to frontline enterprise use case.
Information for this story was found via the sources and companies mentioned. The author has no securities or affiliations related to the organizations discussed. Not a recommendation to buy or sell. Always do additional research and consult a professional before purchasing a security. The author holds no licenses.
