Google’s internal AI code generation tools are approaching their capacity limits as demand grows rapidly within the company. Currently, 75% of new code produced at Google is generated by AI, a significant increase from 50% just six months ago, following a rise from 25% reported in late 2024.
The AI tool chiefly responsible for this surge is Gemini. The dramatic increase in AI-generated code has created a supply-and-demand discrepancy: while Google encouraged engineers to utilize AI tools, the infrastructure is now strained, causing bottlenecks that impede productivity gains these tools were intended to enhance.
Some Google engineering teams have begun seeking exceptions to use Anthropic’s Claude for their projects, despite the company’s policy favoring its own tools. Complaints regarding the performance and accessibility of Google’s internal AI offerings have surfaced, prompting these teams to explore alternatives.
This situation reflects a philosophical divide within Google. Some teams advocate for the rapid deployment of AI models across all workflows, while others caution against compromising code quality and reliability in the haste to adopt AI.
As engineers continue to compete for computing resources, the long-term implications for project timelines and overall productivity at Google remain uncertain.
This material is informational and should not be considered financial advice.
