Author：Wall Street CN

The AI programming wave is reshaping software development infrastructure at an unprecedented pace. Microsoft's code hosting platform GitHub is experiencing a surge in traffic driven by AI agents, so large that it has overwhelmed its servers.

On April 7th, according to a report by tech media outlet The Information, GitHub COO Kyle Daigle revealed that the platform's code commit volume is climbing at an astonishing rate.Currently, weekly submissions have reached 275 million, and it is projected that the total number of submissions for the year will exceed 14 billion, representing a roughly 14-fold increase compared to a year ago.Meanwhile, the number of pull requests initiated by the AI Agent...The number of visits surged from approximately 4 million in September of last year to over 17 million in March of this year.

The report pointed out that,While the surge in traffic brought business prosperity, it also triggered a series of service interruptions.GitHub is accelerating the expansion of server capacity and restructuring of its backend architecture to cope with the pressure, but the changing competitive landscape cannot be ignored either—AI tool providers such as OpenAI and Anthropic are eroding the market share of GitHub's own products.

Traffic Surge: From 1 Billion to 14 Billion

GitHub's traffic growth curve has been almost vertical. Daigle said that last year the platform's annual code commits surpassed 1 billion for the first time, which excited employees; now, that number has been left far behind.

"Since January of this year, we've been breaking historical records almost every month and every week," Daigle said. He attributed this growth to "the combined efforts of AI agents and human developers," and specifically noted that the proliferation of AI programming tools is attracting a large number of users who previously lacked a strong programming background to the GitHub platform.

Public data confirms this trend:The weekly frequency of code submissions to GitHub by Anthropic's programming tool Claude Code has increased nearly 25 times in the past six months, from about 100,000 times to more than 2.5 million times last week.

This surge in traffic is driven by a batch of recently released AI programming agents.OpenAI launched Codex in February of this year. Although Anthropic's Claude Code was released earlier, its usage has seen a leap in recent years with the release of more powerful models.

at the same time,Open source tools have also played a crucial role.Tools such as OpenClaw can schedule multiple agents, including Codex and Claude Code, to automate programming tasks across applications on the user's computer, further amplifying the scale effect of AI code generation.

Tech companies like Meta have even held "tokenmaxxing" competitions, encouraging engineers to compete to see who can generate the most code with AI the fastest speed—this competitive culture reflects the growing popularity of AI-assisted programming in the industry.

Service under pressure: Frequent outages and API rate throttling spark complaints.

The dramatic increase in traffic has put substantial pressure on GitHub's infrastructure.

According to reports, the platform has experienced a significant increase in outages recently. GitHub attributes this to peak traffic surges and the impact of the transition period when migrating applications from its own servers to the Microsoft Azure cloud platform.

Developer discontent is also building. OpenClaw founder Peter Steinberger publicly complained last week that he frequently hit the call limit when using the GitHub API, and bluntly stated that "this system is not designed for agents at all."

GitHub is responding quickly: accelerating the deployment of new servers and refactoring backend software to improve stability, Daigle said.

The report also pointed out that,Behind GitHub's rapid business growth, competitive pressure is also intensifying.

The rise of Claude Code and Codex directly threatens the market share of GitHub's proprietary AI programming assistant, Copilot. Former GitHub CEO Thomas Dohmke, who left the company last year, recently founded a new company focused on creating code storage and testing environments more suited to AI workflows for developers.

More noteworthy is that, according to a previous report by The Information, OpenAI is considering building its own internal code hosting system similar to GitHub and may sell it to Codex users—meaning that one of GitHub's most important sources of traffic may become its most direct competitor.

Daigle remained relatively calm about this. "As long as GitHub usage continues to grow rapidly, intense competition isn't a problem," he said. "Ultimately, all these tools are ultimately pushing code to GitHub…that has always been our most important driver of growth."

It is worth noting that,Whether the explosive growth in traffic can be simultaneously converted into revenue remains an open question.

GitHub's current business model primarily relies on subscription fees per user, with additional charges based on the extent to which users utilize GitHub's proprietary AI features. However, the above pricing rules do not apply to interactions between third-party agents such as Claude Code and Codex and GitHub—meaning that the large influx of agent traffic does not directly generate corresponding revenue.

Daigle did not disclose specific revenue growth figures for GitHub recently, only stating that a recent update to the GitHub Copilot tool has driven sales growth for the product.