Microsoft CEO Satya Nadella revealed this week that nearly a third of the company’s codebase is now being written by AI.
Speaking at Meta’s inaugural LlamaCon developer event alongside Meta CEO Mark Zuckerberg, Nadella estimated that somewhere between 20 to 30 percent of Microsoft’s code in active projects is now produced by AI systems. He noted that this figure continues to rise steadily, highlighting just how deeply machine learning tools have embedded themselves into the day-to-day workflows of one of the world’s most influential software companies.
When Nadella turned the same question toward Zuckerberg, the Meta CEO admitted he didn’t know the exact number for his company. However, he shared Meta’s ambitious direction: the company is developing an AI model capable of designing future iterations of its own Llama AI family. Zuckerberg predicted that, shortly, as much as half of all development at Meta could be handled by artificial intelligence rather than human engineers, with the proportion only increasing from there.
These remarks come at a time when both companies employ vast teams of software engineers, but they’re clearly signaling that AI is no longer a support tool—it’s fast becoming a core engine of production. Since OpenAI’s launch of ChatGPT in late 2022, companies around the world have turned to AI for everything from customer service and sales scripts to, increasingly, the writing of software itself.
The shift isn’t limited to Microsoft and Meta. In October, Google CEO Sundar Pichai announced that over 25 percent of new code at Google was being written by AI. Shopify CEO Tobi Lutke recently told employees that before requesting new hires, they must demonstrate that AI cannot do the job. Meanwhile, Duolingo CEO Luis von Ahn declared in an internal memo that the company would begin gradually replacing its human contractors with AI-based systems.
Adding fuel to the trend, recent reports revealed that OpenAI is in talks to acquire a startup called Windsurf, which specializes in what’s being described as “vibe coding”—AI that can generate complete software programs from just a few words of human input. This marks a vision for the future in which machines not only assist developers but build entire applications autonomously.
For many, the promise of AI-assisted development is simple: write more code, better, and faster. The idea is that with AI handling repetitive or foundational coding tasks, organizations can improve efficiency, lower costs, and potentially even increase software quality. Yet, the implications go far beyond productivity.
With artificial intelligence now capable of writing entire blocks of code, the very nature of software development is evolving—and so is the role of the human developer.
