Google Now Generates 75% of Its Code With AI — Here's Why That Changes Everything
· Nia
There's a number that should make every software developer sit up and pay attention: 75%.
That's the percentage of all new code at Google that is now AI-generated and approved by engineers, according to CEO Sundar Pichai's announcement at Google Cloud Next '26 this week. Just last fall, that number was 50%. In roughly six months, Google went from half to three-quarters of its codebase being machine-written.
Let that sink in. The company that employs some of the most talented engineers on the planet — the people who built TensorFlow, Kubernetes, and Go — has decided that AI writes better first drafts than they do.
And they're not alone. Anthropic disclosed earlier this year that 70 to 90 percent of its code is written with Claude Code. SpaceX, preparing for its IPO, revealed it's designing custom GPUs in-house — partly because the demand for AI compute in its own engineering workflows has gotten that serious.
From Copilot to Autonomous Task Forces
But the raw percentage isn't the real story. The real story is the transition from copilot to agent.
At Cloud Next '26, Pichai described how Google's engineers are now "orchestrating fully autonomous digital task forces, firing off agents and accomplishing incredible things." This isn't GitHub Copilot suggesting your next line of code. This is engineers defining a goal — migrate this codebase, refactor this service, optimize this pipeline — and letting AI agents execute it end-to-end.
Google introduced its new Gemini Enterprise Agent Platform to help organizations manage what they're calling the "agentic enterprise." The pitch: the conversation has moved from "Can we build an agent?" to "How do we manage thousands of them?"
That's a fundamentally different paradigm than what most developers are working in today.
The Infrastructure to Back It Up
Google isn't just talking. They're building the physical substrate to make this work:
- TPU 8t (training-optimized): Scales up to 9,600 TPUs with 2 petabytes of shared high-bandwidth memory in a single superpod. That's 3x the processing power of their previous-generation Ironwood chips.
- TPU 8i (inference-optimized): Connects 1,152 TPUs in a single pod with 3x more on-chip SRAM, designed to concurrently run millions of agents cost-effectively.
- Gemini models now process 16 billion tokens per minute via direct API use — up from 10 billion just last quarter. That's a 60% increase in a single quarter.
Over half of Google's total machine learning compute investment in 2026 is going to the Cloud business. They're not building this for internal use only — they want every company to run agentic workloads on their infrastructure.
What This Actually Means for Developers
Let's be honest about what's happening here. There are two ways to read this:
The optimistic read: AI handles the tedious parts — boilerplate, migrations, refactoring, test generation — and developers become architects and reviewers. You spend less time typing and more time thinking. Productivity goes through the roof, and the definition of "developer" evolves upward.
The uncomfortable read: If 75% of code at Google is already AI-written, what happens at companies with less complex engineering challenges? If you're writing CRUD APIs, form validators, or standard data pipelines — the type of code that AI handles almost flawlessly — your role is going to change whether you're ready or not.
I think both reads are correct, and that's what makes this moment so interesting.
The Inequality Problem Nobody Wants to Talk About
Here's where I'll take a stance that might be unpopular in tech circles: this AI coding revolution is going to be deeply unequal.
MIT Nobel laureate Daron Acemoglu put it bluntly this week in a Financial Times survey: "The rhetoric out there is that the tools are going to be democratizing. But the reality is that you require a certain degree of education, abstract and quantitative skills, familiarity with computers and coding in order to be using the models."
He continued: "AI is going to increase inequality between labour and capital. That is almost for sure."
The developers who learn to orchestrate AI agents effectively will see their productivity multiply 10x. The developers who resist or lack access to these tools will fall behind. And the gap between AI-equipped teams and traditional development shops is going to widen fast.
This isn't theoretical anymore. Google is shipping real products with 75% AI-written code. Anthropic is doing the same. These aren't startups experimenting — they're the companies defining what modern software development looks like.
What You Should Be Doing Right Now
If you're a developer, here's my honest advice:
The Bigger Picture
We're watching software development undergo its most significant transformation since the shift from waterfall to agile. But unlike agile — which was primarily a process change — the AI shift changes the actual act of writing code.
Google processing 16 billion tokens per minute. Seventy-five percent AI-generated code. Eighth-generation TPUs designed specifically for running millions of agents simultaneously. These aren't incremental improvements — they're signals that the industry's center of gravity is moving.
The companies that figure out agentic coding workflows will build software at speeds that seem impossible today. The ones that don't will struggle to compete. And for individual developers, the choice is simple: learn to work with these tools, or spend the next few years watching your peers ship 10x faster.
The 75% number isn't the ceiling. It's the new floor.