The Open Compute Project Foundation (OCP) has used its 2026 EMEA Summit in Barcelona to advance its Open Data Center Ecosystem for AI vision, announcing newly approved community contributions, fresh alliances, and technical milestones spanning power architecture, cooling, and management infrastructure. The summit runs April 29 to 30 at the Barcelona International Convention Centre. Keynote speakers include leaders from AMD, Arm, Google, Intel, Meta, Microsoft, Nvidia, and Schneider Electric.
George Tchaparian, CEO of the OCP Foundation, said the community’s chip-to-grid work directly responds to an open letter from Google, Meta, and Microsoft last October. That letter, specifically, called for greater collaboration on data center physical infrastructure to reduce the fragmentation slowing large-scale AI cluster deployment. Tchaparian added that with no end in sight to the OCP AI data center buildout, community collaboration and open standards are more important than ever.
New Alliances Expanding the OCP AI Data Center Ecosystem
Two new strategic alliances were formalised at the summit. The first pairs OCP with the Electric Power Research Institute (EPRI). Together they aim to accelerate digital innovation and develop data centers as flexible resources for the power system. The partnership, moreover, reflects growing interest in demand response as a tool for grid stability. AI training loads create demand curves that conventional generation dispatch struggles to anticipate.
The second alliance connects OCP with the IOWN Global Forum, led by NTT. IOWN develops high-bandwidth, low-latency compute and network infrastructure spanning centralised and edge deployments. Together, they will build a roadmap for multi-site infrastructure supporting the next wave of AI inference at the edge. Additionally, a third alliance with Current/OS advances the shift from AC-based power to hybrid AC/DC and DC-native data center architectures. A joint keynote at the summit presented progress on low-voltage direct current power distribution standards.
Meta Contributes Catalina Compute Shelf, Supermicro Expands ORv3 Portfolio
Among the technical contributions at the summit, Meta contributed its Catalina AI Compute Shelf to the OCP community. Catalina supports Nvidia’s GB200 platform and is built on the OCP ORv3 rack standard. It supports up to 140 kilowatts per shelf. It incorporates Meta’s Wedge fabric switches to enable the Nvidia NVL72 architecture. That complements a prior Nvidia donation of its MGX-based GB200-NVL72 platform.
Supermicro furthermore expanded its OCP ORv3-compliant portfolio, announcing new rack configurations and servers for high-density AI deployments. A new 2U GPU system features dual Intel Xeon 6 processors and a 1400A busbar with power shelves. It also integrates with the Nvidia HGX B300 8-GPU platform. As we have covered in our analysis of the gigawatt campus problem, lack of standardised physical infrastructure is consequently one of the key factors slowing large-scale AI campus deployment. OCP’s chip-to-grid collaboration model is increasingly seen as the industry’s best path toward reducing that fragmentation.
