Alphabet, Google’s parent company, says about half of its code is now initially written by AI coding agents, with human engineers reviewing, approving or correcting the output. Chief financial officer Anat Ashkenazi disclosed the figure on the company’s fourth-quarter earnings call on February 4, 2026, the same day Alphabet reported annual revenue above $400 billion for the first time.
The disclosure marks a doubling of the share of AI-generated code at the company in a little over a year. In October 2024, chief executive Sundar Pichai said more than 25 percent of new code at Google was generated by AI. By April 2025 the company put the figure at roughly 30 percent. Ashkenazi’s update suggests the share reached about 50 percent by the end of 2025.
Framed as productivity, not replacement
Alphabet presented the shift as a productivity gain rather than a substitute for engineers. Management said AI-assisted coding has increased development speed and freed engineers to spend more time on complex problems, system design and quality control. The company also said it intends to continue hiring engineers through 2026, signaling that it still sees human judgment as central to the process.
The distinction the company draws is between generating code and owning it. Under the workflow Ashkenazi described, the agent produces a first draft, and a human engineer remains responsible for deciding whether it is correct and whether it ships.
Part of an industry-wide trend
Google is not alone. Microsoft has said that around 30 percent of code in some of its repositories is written by AI, and several AI labs have claimed that most of their internal code is now machine-generated, with humans limited largely to review. Those claims have prompted debate among developers about whether code quality is improving or whether the burden on senior engineers to review large volumes of generated code is simply growing.
Record spending on AI infrastructure
The coding figures arrived alongside the largest capital spending forecast in Alphabet’s history. The company guided 2026 capital expenditures to between $175 billion and $185 billion, with most of the money going to servers, data centers and networking equipment to support AI workloads. That is roughly double what the company spent in 2025.
Taken together, the two disclosures show a company that is both consuming AI heavily inside its own engineering organization and betting that demand from outside customers will justify a historic build-out. For the software industry more broadly, Alphabet’s numbers are among the most concrete data points so far on how far AI-assisted development has spread inside a large, established engineering culture, where speed and scale have changed even as decision-making and accountability remain with people.
