Microsoft CEO Says AI Now Writes Up to 30% of Company Code

At Meta’s LlamaCon 2025 developer conference on Tuesday, Microsoft CEO Satya Nadella made a bold revelation: AI is now responsible for generating 20% to 30% of the code written inside Microsoft’s internal repositories. He shared the insight during a fireside chat with Meta CEO Mark Zuckerberg, where the two tech leaders discussed the evolving role of generative AI in software development.

According to Nadella, the impact of AI code generation varies depending on the programming language. “We’re seeing more effectiveness in Python,” he noted, while progress has been slower in more complex languages like C++. Still, Nadella’s statement signals just how deeply AI is integrating into Microsoft’s engineering workflows.

“We are definitely in the early innings of seeing what AI-assisted programming can do,” said Nadella.

Microsoft has been aggressively investing in AI, particularly through its multi-billion-dollar partnership with OpenAI, the creator of ChatGPT and the GPT series. The tech giant has integrated OpenAI models into tools like GitHub Copilot, which assists developers by suggesting code snippets and even writing full functions.

Back in 2023, Microsoft CTO Kevin Scott predicted that 95% of all code could be AI-generated by 2030 — a figure that once seemed ambitious but now feels more plausible as tools mature and adoption accelerates.

When asked how much of Meta’s code is being generated by AI, Zuckerberg admitted he didn’t have a clear figure, indicating that Meta may still be evaluating the full impact of AI tools like Code Llama, its own open-source LLM for developers.

This exchange between the CEOs underscores how tech giants are racing not only to build AI models but to reshape their engineering cultures around them.

On a recent earnings call, Google CEO Sundar Pichai said AI currently generates more than 30% of Google’s internal code. However, none of the companies have shared how they measure AI-written code versus human-written contributions, which raises questions around consistency and transparency.

Still, the numbers provide a glimpse into the industry’s future: one where AI is not just enhancing productivity — it’s actively writing software at scale.

AI-assisted coding tools like GitHub Copilot, CodeWhisperer (from Amazon), and Code Llama are changing how developers work, allowing engineers to focus more on architecture, problem-solving, and optimization rather than repetitive syntax.

As these tools become more reliable, companies may face broader cultural and ethical questions — from code ownership to accountability for bugs, and even potential job displacement for junior developers.

But one thing is certain: AI is becoming a co-pilot in the truest sense, and Nadella’s remarks at LlamaCon make it clear — we’re just getting started.