Facebook needs powerful computers in order to store and process the millions of photos and videos it maintains for its users.
But instead of buying expensive servers from big-name brands like Hewlett Packard Enterprise and Dell Technologies, the social network designs its own servers so it can fine-tune the machines to its own specific computing needs and save some money.
As part of its continuing efforts to design its own servers, Facebook (fb) said Wednesday that it’s upgrading all of the servers used in its vast data centers. Additionally, Facebook said it would contribute the server blueprints to its Open Compute Project foundation, in which companies can exchange and access data center hardware designs for free so they can build their own custom gear.
Get Data Sheet, Fortune’s technology newsletter.
Facebook engineering manager Eran Tal said that one of the reasons the company designs its own computing gear is because “it is very hard to predict what the infrastructure needs may look like a year or two down the line.” By designing its own gear, Facebook can more quickly build its servers using contract manufactures like Taiwan-base Quanta, he explained.
The company’s new Bryce Canyon server will be used to store photos and videos while its Yosemite v2 and Tioga Pass servers will handle computing and processing needs. The new Big Basin server is Facebook’s next-generation machine for artificial intelligence tasks like deep learning that require heavy duty computing power so that Facebook can train its software to recognize images and understand text.
Intel (intc) is providing Facebook with its latest Skylake data center chips to power the new servers, while Nvidia (nvda) is supplying the social network with its graphics chips to help with AI-related tasks, Tal said.
“These are going to be the servers we use across the board for the next couple of years,” Tal said.
As Fortune’s Barb Darrow reported in February, Intel first debuted its new data center chips alongside its big customer Google, which was noteworthy because it chose to reveal its new chips through a big Internet company and not traditional server vendors like Hewlett Packard Enterprise (hpe) or Dell.
For more about tech and finance, watch:
Regarding Intel’s decision to showcase its new chips to big web companies, Tal said “it’s important for them to get up on these designs” that big “webscale folks” like Facebook and Google (goog) are developing. With cloud computing giants like Google and Amazon (amzn) voraciously building data centers worldwide, Intel stands to benefit if it can keep up with the needs of these companies.
Still, as Fortune’s Aaron Pressman reported Wednesday, Intel could potentially face competition from Qualcomm (qcom), as Microsoft said it would partner with the chip maker to use its own server chips for the Azure cloud computing service.