A Facebook-led organization is about to change the $469 billion semiconductor industry forever

• The $469 billion computer chip industry is about to be radically altered.
• Engineers at the Open Compute Project, a powerful industry organization founded by Facebook, are working on a new way to design, build and buy chips.
• The process puts the power in the hands of the people making the chips, not in the vendors who manufacturer them.
• It could reduce the costs of designing new chips by 35%, and make custom chips so affordable that more startups can do them.
• The largest internet companies in the world are all part of this organization and it has already taken on the serves, storage, network and telecom industries.
• Visit Business Insider's homepage for more stories.

The $469 billion computer chip industry is about to be radically altered.

The Open Compute Project, a Facebook-founded industry group, is now the home to a new organization that wants to do for chips what the OCP itself has done for the rest of the data center market: completely overhaul how they are designed, built and sold.

The new project will see single chips turned into "chiplets," where each part of the chip can be bought from different vendors and assembled together by the buyer. The chiplets will be built for a standardized interface. And the design for this interface will be freely given away to anyone.

Read: Microsoft, Google, and others 'are making little or no impact' on Amazon's cloud dominance, according to an analysis of their quarterly earnings reports

That project is known as Open Domain-Specific Architecture (ODSA). It got its start late last year, but was snatched up mere months later to join the OCP - which is the blanket organization that managed many of these kinds of open hardware projects.

OCP is the spot where the world's largest internet companies come together to whip up designs for their own hardware, custom-suited to the ever-growing needs. An ecosystem of contract manufacturers and small startups have sprung up, ready to take those designs and actually build it.

Six years after its founding, OCP now supplies $2.5 billion worth of hardware to its member companies, which include founder Facebook, Baidu, Facebook, Google, Microsoft, and Apple, as well as big non-tech companies like Goldman Sachs.

The idea for ODSA and this new way to build chips came about when "a bunch of engineers at a very large company were complaining," recounts project leader Bapi Vinnakota.

"One of them said, 'it hurts my head every time I have to implement same interface over again for a different product.' It was triggered by engineering boredom," Vinnakota recalls. For his day job, Vinnakota is the director of silicon architecture program management at semiconductor company Netronome Systems.

The engineers began to wish they could stop redesigning every single new graphics or CPU chip from scratch, and just focus on the real point of their work - to create a new part of the chip that would be optimized for specific tasks, like processing AI requests or analyzing video.

They wanted to buy the regular parts as off-the-shelf chiplets from a choice of vendors and snap them together into the chip with the new part they made, saving money and time.

In October the idea became an organization, ODSA

ODSA launched in October with support from big players in the semiconductor industry, including Netronome, Achronix, Kandou, NXP, Sarcina SiFive, and Globalfoundries - a major contract foundry player that would do well if this new type of chip-making took off. Netronome, Vinnakota's employer, is a maker of network interface cards that hopes the ODSA will save it effort and money in designing the cards, while also creating a new market for them among the OCP's membership, the largest internet companies in the world.

Word quickly spread.

By December, when the group first released a paper on the idea, there were already 10 companies involved. It held a workshop, and more showed up. It joined OCP in mid-March, and two weeks later, 52 companies were involved.

Today, the project includes engineers from processor giants like Samsung and NXP who have a lot to lose if this new way of making chips takes off and they don't get a piece of the action.

And while OCP is a completely separate entity from Facebook, the social network is still a powerhouse sponsor putting its weight behind the project.

Noticeably absent from the ODSA lineup are the processor companies most at risk, if a new way of building processors hits the mainstream in the industry. Those missing companies includes chip giants like, AMD, Broadcom, Intel, and Xilinx. All of them are watching this group closely, though, Vinnakota said.

They don't need to seem as if they approve of the idea quite yet.

The project expects to have its first designs by the end of the year, and some proof-of-concept chips next year. Even then, designing chips is hard. It could be a while before they have sturdy designs and manufacturers tooled up to build them.

Read: A male Microsoft programmer who came from the military, not a university, describes how his career got tanked by the 'brilliant jerks' culture

If it works, success is almost guaranteed

Should this group succeed - and there's every reason to believe they will - they will have an eager group of buyers who spend billions on this tech waiting.

OCP began as an idea inside Facebook when engineers there decided they wanted to do for data center hardware what the Linux operating system did for software: make it freely available for all engineers to work on, without worrying about the intellectual property of the big IT vendors.

Since OCP was launched in 2013, it has steadily altered the way data center hardware is designed, bought and sold for the biggest internet players in the world. They now design their own servers, computer storage, computer networking equipment. And an offshoot group led by Facebook is now gunning for the telecom equipment industry.

Chips were one of the remaining few hardware markets that hadn't been touched until now.

Vinnakota estimates that if their chiplet works as planned, it could save tech companies 25-30% of the cost of what they spend to design new chips.

On top of that, it will make chip design more affordable for startups.

And the big internet companies love the idea. They need to invent new kinds of chips to handle a growing crop of AI applications.

"This is a huge way to lower the costs," Vinnakota said.