The Open Compute Project is a collaborative community that open-sources data-center hardware designs — servers, racks, storage, networking, cooling and power — contributed by hyperscalers and vendors so that others can manufacture and deploy them without proprietary lock-in.
Open Compute Project
The Open Compute Project began when Facebook published the designs for its Prineville data-center hardware in 2011, and grew into the industry body where hyperscalers and vendors contribute open hardware specifications — rack standards, server designs, open networking, storage, cooling, power distribution, and more recently open accelerator modules for AI infrastructure.
- Designs, openly licensed - Mechanical, electrical and thermal specifications published for anyone to manufacture against.
- Open Rack and OAM - Rack-level and accelerator-module standards that let multiple vendors build interchangeable parts.
- Contributed by operators - The specifications come from the organizations running the largest fleets, which is why they are practical rather than aspirational.
- Supply-chain security work - OCP hosts efforts on firmware integrity and device attestation, which is the part closest to a machine-readable interface.
OCP is a genuine open-standards success story and, like the rest of this market, it ships specifications rather than endpoints. That is the right output for a rack design. It is worth noting alongside Redfish, which is the one place the same community produced a network-facing contract — and the one place The State of Compute & Hardware APIs found hardware companies scoring well.
Referenced in API Evangelist papers
This standard shows up in my published research. These reports read the machine-readable evidence provider by provider — and put this standard in the context of a real sector.
The State of Compute & Hardware APIs
Open hardware designs published as specifications rather than endpoints — characteristic of a market whose standards describe parts, not interfaces.
