GitHub Plans to Achieve 180x Throughput Improvement by Early Next Year to Support AI Agents

Google Ad Server Settings

Simple Analytics Placeholder event function

×

End of Header

End of superbanner /3002452/Pbk_Rect1_300x300

/3002452/Pbk_Rect2_300x250

GitHub to achieve a 180x throughput improvement over the conventional level, early next year. Adapting to the AI agent era. GitHub Universe 2026, November 4, 2026

At its annual event "GitHub Universe 2026," held in San Francisco, USA, over two days on October 28 and 29, GitHub revealed that it will achieve a major performance improvement to adapt to the AI agent era.

The performance improvement was explained as being achieved by migrating GitHub's infrastructure to Microsoft Azure, adopting a new architecture that separates compute and storage, and improving locking by changing it from the repository level to the branch level.

From the keynote, here is a digest of the explanation about GitHub's performance improvement.

Jay Parikh, Executive Vice President of CoreAI at Microsoft

About improving GitHub's performance and reliability. This year alone, we have expanded by 4 million CPU cores, 160PB of storage, and 50Tb/s of network bandwidth.

Furthermore, we are migrating to Microsoft Azure, and multi-region deployment is also underway. This large-scale engineering effort will be completed early next year.

However, the load on GitHub is increasing rapidly, and in the AI agent era we will review all previous assumptions and further improve performance and strengthen reliability.

One of these efforts is the evolution of the Git architecture.

The new evolution enables better responsiveness and reliability.

We separate compute and storage in GitHub's infrastructure, minimizing and simplifying the coordination between the two.

As a result, in testing, even when compressing 24 hours' worth of traffic from GitHub's most active repository into one minute and ingesting it, it was able to process it with ease.

Let me explain in detail.

We have changed locking on new writes to Git so that instead of locking at the repository level, it locks at the target branch level.

This allows AI agents working on different branches to operate in parallel, improving throughput by 3x to 4x.

Major performance improvement through separation of storage and compute

In the aforementioned infrastructure architecture change, we worked with the Microsoft Azure storage team to remove the performance ceiling.

Code is stored in Microsoft Azure Blob Storage, and the servers that respond to Git requests merely cache data temporarily for speed.

This allows storage capacity and servers to be flexibly scaled on demand, with coordination between the two kept to a minimum.

The only critical path that remains as a performance constraint is a small number of reference transactions, which are handled by an optimized dedicated log service.

As a result, whereas the old architecture was limited to dozens of pushes per second to a single repository, the new architecture was designed to handle thousands of pushes per second.

When 5,000 AI agents were made to work on a single repository, it was able to process 7,200 writes per second. This is 180x the conventional throughput.

All users will be on the new architecture by early next year

The content introduced here is one example of efforts to scale GitHub to adapt to the AI agent era.

By early next year, all users are scheduled to be migrated to this new architecture.

Related articles on GitHub Universe 2026

GitHub officially releases "Copilot Fusion." High performance and low cost by selecting the optimal AI model according to the task. GitHub Universe 2026

GitHub announces "Copilot Code Review Max" for more advanced code review. GitHub Universe 2026

GitHub to achieve a 180x throughput improvement over the conventional level, early next year. Adapting to the AI agent era. GitHub Universe 2026

Read more

Software testing/quality Development tools GitHub Microsoft

≪ Previous article: GitHub announces "Copilot Code Review Max" for more advanced code review. GitHub Universe 2026

After working as an editor at an IT magazine and a publisher of an online media outlet, I became independent. I started Publickey in 2009.

(Detailed profile)

Would you like to check the latest news from Publickey?

GitHub to achieve a 180x throughput improvement over the conventional level, early next year. Adapting to the AI agent era. GitHub Universe 2026

GitHub announces "Copilot Code Review Max" for more advanced code review. GitHub Universe 2026

GitHub officially releases "Copilot Fusion." High performance and low cost by selecting the optimal AI model according to the task. GitHub Universe 2026

Oracle announces "Oracle App Composer," which lets employees automatically create business apps with AI and share them on Oracle Cloud. Oracle AI World 2026

Oracle Cloud announces adoption of Google TPU, to begin offering in 2027. Oracle AI World 2026

JavaScript runtime "Bun" becomes a Prettier-compatible formatter more than 300x faster. Prettier configuration files can also be used as-is

AWS and Google Cloud open-source versions of "PyTorch" for AWS Trainium and Google Cloud TPU. Taking on the challenge of migrating from NVIDIA

"DuckDB 2.0 LTS" officially released. Client/server functionality via the Quack protocol and speedups through asynchronous processing, etc.

Cloudflare announces "tsrs," a port of TypeScript to Rust. Up to 70% less memory and 8x faster than TypeScript 7. Plans to support all TypeScript features including LSP in the future

Managed service provision for "Cloudflare OS." Connection with GitHub and strengthened integration with Gmail, etc.

Our efforts to ensure readers can read articles with peace of mind

Simple Analytics 100% privacy-first analytics

- script for simple analytics events --

Global site tag (gtag.js) - Google Analytics