OpenAI announced on November 4 that it will acquire Pierre Computer Company, the company behind the Git infrastructure product Code Storage and the open-source libraries Trees and Diffs. Pierre confirmed the deal on its own website. The purchase price was not disclosed, closing is subject to the usual regulatory approvals, and the two companies will continue to operate independently until then.
OpenAI's stated rationale centers on Codex, which it wants to turn into "a place where teams can build software together with AI, from first idea to the version they ship."
The post-deal arrangements break down roughly into three points:
- Trees and Diffs will remain open source. Both libraries are already used in the Codex app: the former handles file browsing, the latter displays code changes.
- Existing Code Storage customers can keep using the product, and its features will be folded into Codex over time.
- The Pierre team will join OpenAI to work on new products and platforms.
For Codex users, nothing changed on the day of the announcement: no new features, no new plans, no price changes.
What Pierre sells
Pierre was founded in 2023 by Jacob Thornton, one of the authors of Bootstrap, and Ian Ownbey, a former Twitter engineer. The company closed a $22 million Series A in November 2025, led by CRV and 01A. On its Y Combinator directory page, it describes itself as "a new, opinionated, git platform."
Code Storage launched in October 2025, so it was acquired roughly 13 months after going live. It is a white-label Git layer for AI apps and agents: standard Git clients can clone, fetch and push as usual, customers drive it through an API, and it supports syncing with GitHub, GitLab and Bitbucket.
Publicly listed figures include a maximum of 32 TB per repository, more than 30 concurrent writers per repository, more than 500 Git requests per second per repository, and a 99.99% SLA. The customers named include Lovable, Bolt, Amp, Poke, Anything and Parahelp.
Running the numbers
An agent writing code puts very different pressure on Git than a human does. Lovable's August case study shows it best: more than 8 million new repositories a week, with a peak of 250 per second. A rough calculation gives an average of about 13 per second, with the peak around 19 times the average. If every generation opens a new repository, most of those repositories probably go cold quickly.
Code Storage's pricing fits that pattern. Hot data accessed in the past 7 days costs $0.005 per GB per hour, or about $3.65 a month; cold data untouched for 7 days costs about $0.15 a month, roughly 24 times less. Push traffic is $0.06 per GB and pull traffic is $0.15 per GB.
For a mid-sized AI site-building product with 200 GB of hot data, 2 TB of cold data, 300 GB of push traffic and 1.5 TB of pull traffic, that comes to about $1,270 a month. This is our own rough estimate from the public price list; large customers get custom pricing, so real bills will not follow it.
OpenAI already has its own system
There is an awkward detail in this deal. Two days earlier, on November 2, OpenAI published a post about Pando, its in-house Git-compatible backend. It said Pando already hosts the company's main code repository, has cut median push latency by nearly 80%, handles close to 1 PB of repository data a day, and is being used to study workloads where large numbers of agents share one repository.
Pando is an internal system and Code Storage is a product sold externally, and the two address overlapping problems. OpenAI has not said whether Pando will merge with Code Storage or be opened to outside users.
Lining up this year's acquisitions makes the direction fairly clear: Python tooling company Astral in March, cloud development environment company Ona in June, and now Git storage and the front-end components for code review. Codex is expanding outward from writing code, filling in environments, dependencies and repositories one piece at a time.
Customers still have no answers
Lovable, Bolt and Amp, all on the Code Storage customer list, build things that overlap with Codex. Their code repositories will now sit on infrastructure owned by OpenAI, and OpenAI has so far only promised that customers can keep using it.
Whether Code Storage will continue as a standalone service over the long term, how existing contracts and pricing will be handled, which Codex plan it will belong to and with what limits once merged, and whether Codex will offer hosted repositories directly are all undisclosed. The announcement also gave few details on licensing and governance for the open-source libraries after closing, and no closing date was given.
Sources: OpenAI official announcement, Pierre website, CocoLoop, Data Studios, RuntimeWire; the Code Storage public price list and the Lovable case study were used to verify unit prices and repository figures, and the monthly cost is an editorial back-of-the-envelope estimate.