The OCP Global Summit returned to San Jose, Calif., this week with a clear message for the infrastructure industry: the next phase of AI computing depends as much on power, cooling, storage, and modular design as it does on faster processors and accelerators. The annual gathering of the Open Compute Project brings together server makers, chip designers, cloud operators, storage vendors, cooling specialists, and data center builders to share designs and specifications for more open, efficient, and scalable hardware. This year’s news cycle touched nearly every layer of the physical stack behind cloud and AI data centers.

The Open Compute Project began as an effort to rethink data center hardware through open collaboration. Its work has expanded from motherboards and racks to networking, storage, power distribution, liquid cooling, and complete data center modules. At a time when AI workloads are pushing rack densities and thermal limits higher, that broad scope matters. Operators are trying to deploy large clusters quickly while controlling energy use, serviceability, and cost. The summit announcements showed vendors responding with products and standards aimed at those pressures.

Key facts from the summit

• Nvidia contributed rack and compute tray designs for its Vera Rubin platform to the Open Compute Project through its modular MGX architecture.
• Arm released a white paper proposing a use-case-driven method for measuring CPU contributions to agentic AI workloads.
• Arm cited independent testing that compared its AGI CPU with AMD Epyc Turin processors across selected rack-level agentic AI workloads.
• Kioxia introduced LD4 Series enterprise SSDs using eighth-generation BiCS FLASH 3D QLC NAND, with 15.36TB and 30.72TB capacities and an architecture validated for up to 122.88TB.
• Cooler Master expanded its enterprise liquid-cooling portfolio with a 2,500 kW liquid-to-liquid coolant distribution unit.
• Flex expanded its prefabricated modular data center portfolio with AI Modules that combine power, computing, and liquid cooling.

Nvidia opens Vera Rubin rack and tray designs

Nvidia submitted key rack and compute tray designs for its Vera Rubin platform to the Open Compute Project through its modular MGX architecture. The company’s goal is to give server vendors and other infrastructure suppliers a shared design foundation for building AI systems. That kind of contribution can reduce the engineering work required to bring rack-scale AI products to market, especially for suppliers that want to differentiate at the integration, service, and deployment layers rather than redesigning basic mechanical and thermal elements.

One distinguishing feature of Nvidia’s rack and tray designs is that they contain no fans or cabling because the systems are liquid cooled. Removing fans and internal cabling can simplify deployment and service, reduce mechanical complexity, and make it easier to route heat away from dense compute components. The design also reflects a broader shift in AI infrastructure: high-power accelerators and tightly packed servers are moving toward liquid cooling as air cooling approaches its practical limits in many high-density racks.

Arm proposes a new way to measure CPU performance for agentic AI

Arm released a white paper, developed with partners, that proposes a use-case-driven methodology for measuring the CPU’s contribution to agentic AI systems. The framework defines five representative workload types: software engineering, enterprise operations, financial and scientific research, personal assistants, and robotics. It pairs component-level benchmarking with end-to-end task measurement built on re-playable agent trajectories. The approach is meant to move beyond simple synthetic benchmarks and capture how CPUs support the orchestration, tool use, planning, and execution steps that make up agentic AI workflows.

Arm also highlighted independent testing by Signal65 that found its AGI CPU could deliver up to twice the rack-level performance of AMD’s Epyc Turin generation server processors across select agentic AI workloads. The comparison emphasized completed work under rack-level constraints rather than conventional CPU benchmarks alone. In the testing, an Arm AGI CPU averaged 1.7x better performance for tooling, 2.0x for agentic workflows and orchestration, and 2.2x for reasoning and planning in modeled per-rack performance versus an AMD Epyc Turin processor. Arm said its AGI CPU reduces sandbox creation latency by up to 45% versus AMD Epyc 9965, helping agents spin up isolated execution environments faster.

The debate over how to benchmark AI infrastructure is becoming more important as buyers try to compare systems for agentic workloads. Agentic applications can spend significant time waiting on tool calls, coordinating multiple models, and managing isolated environments. A processor that accelerates those steps may deliver more useful throughput even if it does not lead on every traditional CPU test. Arm’s proposed methodology is an attempt to make those trade-offs more visible.

Kioxia targets AI storage density with LD4 Series SSDs

Kioxia unveiled its LD4 Series enterprise solid-state drives, designed to increase flash storage density in hyperscale data centers and high-density 1U servers. The drives target read-intensive workloads, including data storage applications that support AI infrastructure. As AI systems generate and consume larger data sets, storage density and power efficiency become more important at every layer of the data center.

The new SSDs use eighth-generation BiCS FLASH 3D quad-level cell NAND technology with capacities of 15.36TB and 30.72TB. Kioxia said the drives have a validated architecture designed to support capacities of up to 122.88TB. They support PCIe 5.0 and NVMe 2.0e, along with the Open Compute Project’s Datacenter NVMe SSD Specification 2.6. The drives feature a single-port design for hyperscale and data center-class servers. Kioxia said the LD4 Series is currently sampling to select customers.

Cooler Master brings gaming liquid-cooling experience to the data center

Cooler Master, a company best known for providing liquid cooling to gaming systems, is bringing that experience to the data center. It is expanding its end-to-end liquid-cooling portfolio for AI data centers and highlighted a 2,500 kW liquid-to-liquid coolant distribution unit designed for high-density AI and high-performance computing infrastructure. The product is aimed at operators that need to remove heat from racks carrying far more power than traditional enterprise workloads.

The system supports coolant flow rates of up to 990 gallons per minute and features dual-system architecture, intelligent load balancing, and centralized management for as many as 32 interconnected CDUs, according to the vendor. Mass production is targeted for the fourth quarter of 2026. Cooler Master’s expanded portfolio is aimed at data center operators building new AI facilities as well as those retrofitting existing sites to accommodate increasing rack densities and thermal demands. The move illustrates how cooling has become a strategic part of AI data center design rather than a secondary facility concern.

Flex expands prefabricated modular data centers for AI

Flex announced an expansion of its prefabricated modular data center portfolio with integrated systems designed to accelerate deployment of AI infrastructure by as much as 50% compared with traditional construction methods. The new AI Modules combine power delivery, high-density computing, and advanced liquid cooling in factory-built systems. This approach shifts much of the integration and testing work from the data center construction site to Flex, the company says.

The systems include an 800-volt direct-current power rack for power conversion and distribution, cooling distribution units to manage heat from power-intensive AI workloads, and high-density IT racks incorporating AI servers, networking equipment, and high-speed interconnects. The expanded portfolio is part of Flex’s cloud and power infrastructure business, which the company plans to spin off as an independent, publicly traded company called Axiom Solutions International in the first quarter of 2027. The announcement reflects a wider push to industrialize AI data center construction, moving complex integration into factories where systems can be assembled and tested before they reach the site.

The summit’s announcements, taken together, showed that AI infrastructure competition is no longer limited to chips. Vendors are competing on rack-scale integration, thermal management, storage density, power distribution, and the speed with which complete data center capacity can be deployed. Open designs and shared specifications are becoming a way for suppliers to participate in those markets while giving operators more choices across the physical stack.

Source:Network World News