Google: 75% of All New Code Is Now AI-Generated — Up from 50% in Six Months

Three-quarters of all new code at Google is now written by AI. CEO Sundar Pichai revealed the figure at Google Cloud Next 2026, showing a dramatic acceleration from 50% just six months ago. At this pace, the majority of code at the world’s fourth-most-valuable company will be machine-generated before the year is out.

The Numbers

Google’s AI coding trajectory tells the story by itself:

October 2025: 25% of new code AI-generated (first disclosure)
November 2025: 50% of new code AI-generated
April 2026: 75% of new code AI-generated

That’s a tripling in less than twelve months. And Google isn’t some startup experimenting with copilots — it’s one of the largest and most sophisticated engineering organisations on Earth, with over 180,000 employees and some of the planet’s most complex software systems.

The Strike Team

Google recently formed an internal “strike team” specifically to improve its AI coding capabilities and close the gap with Anthropic, where 70–90% of code is reportedly written by Claude Code. The fact that Google is playing catch-up to a company with a fraction of its engineering headcount is telling. Anthropic’s advantage isn’t scale — it’s that they built their own AI coding tool and dogfooded it relentlessly.

The strike team’s existence confirms what many suspected: Google’s internal AI coding tools haven’t matched Claude Code’s capabilities for actual engineering work, despite Google having some of the world’s best AI research talent. Research leadership and product leadership are different skills.

Why This Matters Beyond Google

This isn’t a Google story. It’s a signal for every developer and every company that employs them.

If the world’s most resourced engineering organisation is at 75% AI-generated code and accelerating, the question for everyone else isn’t whether AI coding will arrive — it’s whether you’re already too far behind to catch up.

For New Zealand’s tech sector, where many companies are still evaluating AI coding tools, the gap between adopters and holdouts is widening fast. The organisations that integrated AI coding a year ago have compounding advantages: their engineers are fluent in AI-assisted workflows, their codebases are optimised for AI readability, and their velocity is pulling away from competitors every sprint.

The Inconvenient Truth

The same Pichai announcement contained a less comfortable datapoint. While 75% of new code is AI-generated, Google also acknowledged that AI code requires more review and maintenance than human-written code. The productivity gains are real, but so are the quality costs — at least for now.

This mirrors what’s happening industry-wide. AI coding dramatically increases output, but that output needs more oversight, more testing, and more experienced engineers doing the reviewing. The net effect is still strongly positive, but it’s not the free lunch that some vendors promise.

What Changes When Code Is Cheap

When 75% of new code costs essentially nothing to write, the economics of software development fundamentally shift:

Prototyping becomes nearly free. Teams can spin up and discard prototypes at a pace that was impossible even two years ago.
The bottleneck moves to review and architecture. Writing code is no longer the rate limiter — deciding what code to write and verifying it’s correct is.
Senior engineers become more valuable, not less. The people who can evaluate AI output, spot subtle bugs, and make architectural decisions become the scarce resource.
Small teams can compete with large ones. If output per engineer triples or more, a five-person startup can produce what previously required twenty.

The Pace Is the Story

The most striking element isn’t the 75% figure itself. It’s the rate of change. Going from 25% to 75% in under a year means the adoption curve is steeper than almost anyone predicted. If the trend continues — and there’s no sign it won’t — Google will likely cross 90% AI-generated code before 2027.

For developers watching from the outside, the window to get fluent with AI coding tools is closing faster than expected. Not because AI will replace you tomorrow, but because the gap between engineers who can effectively direct AI coding and those who can’t is becoming a career-defining difference.