Coinbase is increasing its reliance on artificial intelligence in software development, with almost 40% of the daily code generated by AI systems currently being generated.
Chief Executive Brian Armstrong said the cryptocurrency exchange plans to increase that figure to more than 50% by October, reflecting the company's determination to embed AI in every layer of its engineering process.
Armstrong's hard work regarding AI adoption
sponsor
sponsor
Armstrong disclosed his target in X's post, calling AI adoption a strategic order. This push follows a few months of internal programme to standardize the use of coding assistants, integrate new developer tools, and measure productivity with quantitative and qualitative metrics.
“~40% of the daily code written in Coinbase is generated by AI. I want to be >50% by October,” Armstrong wrote.
Coinbase's commitment to AI goes beyond spontaneous experimentation. In the report, Armstrong explained that he required engineers to be installed in AI tools such as Github Copilot and Cursor. Employees who failed to comply were asked to justify their stance, some of which were rejected.
“AI is important. We need you to learn it, and at least it's onboard,” Armstrong said, explaining his decision to “get fraudulent” by bypassing the long rollout timeline.
Coinbase also hosts monthly “AI Speedruns.” In this host, employees showcase practical applications of AI within their workflows. Armstrong says that about a third of the company's code already comes from AI systems, with the goal being 50% until the end of the year.
Developer Productivity Framework
sponsor
sponsor
Coinbase detailed its strategy on its company blog on August 6th. Written by Kyle Cesmat and Chitra Venkatramani, this post outlined a framework built on Dora Metrics and developer research. Emphasizing responsible AI integration, note that not all systems are suitable for code generation due to sensitivity and security considerations.
The company reported that over 1,500 engineers use AI-powered tools every day, including IDE integrations and custom model routers.
Alternatives such as Cody and Jetbrains continue to be used, but cursors have become the preferred integrated development environment for many engineers.
Coinbase tracks AI adoption through metrics such as lead times, changes, deployment frequency, and shares of AI-generated codelines. According to the blog, the AI output will be “on track to prey on human-generated code” by the end of the year.
Leaders in major technology highlight similar shifts
sponsor
sponsor
Coinbase's trajectory reflects broader trends among major technology companies. Google CEO Sundar Pichai says that over 25% of the company's new code is written by AI, describing it as a fundamental change in software development.
Microsoft CEO Satya Nadella is heading further, predicting that 30% of the company's code is generated by AI, and that the percentage could increase to 50% within the next two years.
These numbers show how AI is no longer a fringe experiment, but a key productivity driver at the heart of enterprise technology strategies.
Research has confirmed global adoption
Academic research supports disclosures of these companies. A 2025 survey on ARXIV surveyed millions of Github repositories and found that AI-generated code already accounted for 30.1% of US features, 24.3% in Germany and 23.2% in France.
The study also estimates that productivity gains from AI-generated code could translate into an additional annual economic value of between $10 billion and $14 billion in the US alone.
sponsor
sponsor
The analysis highlights important points. AI-driven development is not evenly distributed. Recruitment is high in the hub of advanced economies and technology, while other regions are heading backwards.
Developers report daily use
sponsor
sponsor
Research shows that shifts are felt strongly at the developer level. According to the Stack Overflow 2025 Developer Survey, 84% of respondents say they already use AI coding tools or plans, with over half reporting daily use. In particular, over 40% of developers using AI say that at least half of their codebase is generated by AI.
Coinbase's goal of exceeding 50% AI-generated codes by October reflects the wider industry changes. From Silicon Valley to Europe, companies are embedding AI into their development, reshaping the balance between human and machine contributions.
The future challenge is to increase productivity without compromising quality or security. As the company's blog stated, “We're building a culture of innovation that makes Coinbase the perfect place to build.”
