DIGITIMES ASIA: Meta-backed OCP pushes open alternative to closed AI hardware ecosystem

Korea Herald

6 days ago

TAIPEI, July 24, 2025 /PRNewswire/ -- As surging demand threatens to fragment the supply chain into costly proprietary silos, the Open Compute Project, an nonprofit organization that has standardized data center hardware for tech giants like Meta and Microsoft, has positioned itself as the antidote to a looming crisis in AI infrastructure, with CEO George Tchaparian pointing out that open hardware standards will prevent market fragmentation as AI infrastructure demands explode.
Tchaparian warned that AI's unique requirements are pushing the industry toward dangerous specialization. Without coordinated standards, he argued, the race to build ever-larger AI clusters could splinter the market and drive up costs for everyone.
"AI workloads are different, much more so than other virtualized and cloud-native applications," Tchaparian said in a recent interview. "In the pursuit of greater performance, infrastructure specialization for some workload categories is burgeoning, but risks fracturing the supply chain into silos."
The stakes are enormous. Current projections suggest AI and high-performance computing buildouts between 2024 and 2028 will push data center power consumption to "dangerously high levels," according to Tchaparian. Annual carbon emissions from these facilities are expected to grow exponentially, creating what he calls a concerning impact on humanity.
The Chiplet Gambit
OCP's answer lies in what Tchaparian calls the "Open Chiplet Economy" - a standardized marketplace where semiconductor components can be mixed and matched like Lego blocks. The organization has launched a dedicated marketplace section featuring over 25 chiplet suppliers, aiming to create the industry's first truly interoperable silicon ecosystem.
"The next inflection point for the silicon supply chain is open," Tchaparian said. "Developing an open stand-alone chiplet silicon supply chain will require a rethink of the supply chain."
The approach mirrors OCP's successful standardization of server hardware over the past decade, which helped hyperscalers slash costs and accelerate innovation. Now, with AI clusters demanding unprecedented compute density - including proposed 1-megawatt racks - the organization is betting that open standards can prevent the kind of vendor lock-in that has historically plagued enterprise technology.
The strategy carries particular significance for the Asia-Pacific region, where Taiwan's semiconductor manufacturing dominance intersects with surging AI infrastructure demand. OCP counts over 130 APAC members and is preparing for its 2025 summit in Taipei, designed to address regional supply chain challenges.
"In APAC, a manufacturing hub, we're engaging through collaborations with local organizations such as ITRI and IOWN," Tchaparian said, referring to Taiwan's Industrial Technology Research Institute and Japan's Innovative Optical and Wireless Network initiative.
The timing appears strategic. As geopolitical tensions reshape global technology supply chains, OCP's open approach offers an alternative to the proprietary ecosystems that have dominated AI development. Major players, including Nvidia, Intel, and AMD, are participating in the organization's "Open Systems for AI" initiative, launched in January 2024.
Beyond cost and compatibility, Tchaparian frames open standards as essential for addressing AI's environmental impact. OCP has established formal partnerships with organizations like iMasons to develop standardized methods for reporting carbon emissions in IT equipment production - information intended to influence purchasing decisions.
"We must reduce the environmental impact of today's computational infrastructure," he said. The organization is even funding research into reducing concrete's carbon footprint, a major source of emissions in data center construction.
For companies navigating this transformation, Tchaparian offers straightforward advice: "Eighty percent of success is just showing up." Even without formal membership, organizations can participate in OCP's collaborative development process, gaining access to cutting-edge designs and influence over future standards.
The test will come as AI infrastructure demands continue their exponential growth. If OCP succeeds in creating truly interoperable standards, it could prevent the kind of market fragmentation that has historically driven up technology costs. If it fails, the AI boom may leave the industry more fractured - and expensive - than ever.
The 2025 OCP APAC Summit will be held in Taipei, featuring discussions on next-generation data center infrastructure and emerging technologies.