Microsoft CEO Satya Nadella has revealed that 20% to 30% of the company’s internal codebase is now being written by AI-powered tools, a striking indicator of how artificial intelligence is reshaping software development at the enterprise level. Nadella made the remark during a fireside chat with Meta CEO Mark Zuckerberg at Meta’s LlamaCon 2025 conference on Tuesday.

Responding to Zuckerberg’s question about the current share of AI-generated code within Microsoft, Nadella said the percentage varies depending on the programming language. “We are seeing much better results in languages like Python,” he noted, while admitting that AI still struggles to achieve similar efficiency in lower-level languages like C++.

The insight comes at a time when leading tech firms are racing to embed AI deeper into their development pipelines, using tools like GitHub Copilot (backed by Microsoft) and other generative code models that can autocomplete, suggest, and even autonomously write functional code.

While Nadella offered the current estimate of AI contribution to Microsoft’s codebase, Microsoft CTO Kevin Scott had earlier predicted that by 2030, as much as 95% of code could be generated by AI. This projection aligns with the company’s ongoing efforts to build advanced developer copilots and automate routine coding tasks, freeing engineers to focus on design, logic, and innovation.

The conversation with Zuckerberg at LlamaCon also highlighted the uncertainties in tracking AI’s exact impact. When Nadella flipped the question, Zuckerberg admitted that Meta has not yet quantified how much of its code is being written by AI tools. This suggests that while AI adoption is increasing, measurement standards across companies remain inconsistent.

Meanwhile, Google CEO Sundar Pichai disclosed during the company’s recent earnings call that over 30% of Google’s code is now AI-generated. These figures point to a larger trend but also underline a critical issue: the absence of clear metrics for distinguishing AI-written versus human-written code.

Although the numbers shared by Nadella and Pichai are impressive, industry experts caution against taking them at face value. The methodology used to define what counts as “AI-generated code” is not uniform. For instance, does code merely suggested by AI tools count, or only code fully generated and committed with minimal human intervention?

Moreover, there are broader concerns around the quality, security, and maintainability of AI-generated code, especially in large-scale enterprise environments. While AI excels in routine, repetitive coding, it still falls short in understanding complex business logic, architectural patterns, or nuanced security constraints.

Nonetheless, Nadella’s comments reflect Microsoft’s deep investment in AI as a co-pilot for developers, reinforcing its strategy to lead in enterprise-grade AI tooling through Azure AI, GitHub Copilot, and OpenAI partnerships.

As the industry continues to embrace AI for coding, companies may need to develop standards for attribution, auditability, and ethical accountability in machine-generated software, especially as reliance grows.