Sundar Pichai dropped a number during the quarterly earnings call.
75%.
That is the proportion of Google's new code currently written by AI. Humans review it; AI generates it.
18 months ago, that figure was 25%.
18 months, from 25% to 75%
The trajectory is clear:
- October 2024: 25%
- Autumn 2025: 50%
- April 2026 (now): 75%
And this isn't just about new feature code. Code migration is now 6 times faster than a year ago.
Code migration refers to the tedious work of moving old systems from framework A to framework B — labor-intensive, error-prone, and hard to estimate. These tasks are now also handed over to AI.
Pichai used one term: agentic workflows. Not having AI autocomplete a line, but letting AI autonomously run through a task. Developers have shifted from executors to reviewers.
More interesting details
The Decoder's report buried a point more telling than the 75% figure itself:
Some DeepMind employees are using Anthropic's Claude Code, not Google's own Gemini.
Pause there.
DeepMind is the team that trained Gemini. And some of its people are using a competitor's tool to write code.
Google apparently tacitly allows this. The same report notes that DeepMind has a dedicated team tasked with catching up to Anthropic's capabilities in AI coding.
Translation: Gemini is not yet as good as Claude for coding scenarios.
This level of candor is surprising.
What the numbers mean
Google is not the first company to release such figures:
- Anthropic says Claude writes about 94% of the company's code
- Some Microsoft engineering projects have over 80% AI-generated code
- A top Spotify engineer went four months without writing a single line of code by hand
The direction is clear: the core job of a software engineer is shifting from writing code to reviewing code.
This doesn't mean programmers are losing their jobs. Reviewing 75% AI-generated code may require more judgment than hand-writing 25% — you need to spot mistakes the AI makes that you wouldn't make yourself.
But the nature of the work is changing, that much is true.
Why Pichai brought this up on an earnings call
The timing is telling.
A 6x internal efficiency gain and massive code productivity improvements translate into real money.
Pichai is telling the capital markets: Google is a direct beneficiary of the AI era, not just a supplier.
A company that builds AI tools is also one of the biggest users of those tools. That cycle is Pichai's most compelling investor story.
If this number becomes 90% in a year, I won't be surprised.
The real question worth pondering: when 99% of code is written by AI, what does it mean to know how to code?
Source: CocoLoop, Google says 75 percent of its new code is now written by AI (The Decoder)