Taiwanese manufacturers of servers, networking equipment, and power and cooling systems are showcasing their solutions for artificial intelligence (AI) data centers this week at the Open Compute Project (OCP) Global Summit 2026, held October 12-15 in San Jose, California. Among the standout offerings are Compal’s high-density rack infrastructure and WNC’s connectivity equipment, with particular focus on power delivery, liquid cooling, and next-generation Ethernet networks.
AI infrastructure key takeaways in 30 seconds
- OCP Global Summit 2026 runs October 12-15 in San Jose and brings together companies developing hardware and software for data centers.
- Compal is presenting infrastructure for NVIDIA accelerated systems, with 800-volt direct current power proposals and liquid cooling.
- WNC is showing 1.6 Tb/s Ethernet switches with switching capacity of up to 102.4 Tb/s, plus 800 Gb/s optical transceivers.
- The demonstrations show that scaling AI requires coordinating servers, networking, power, and cooling — not just adding more accelerators.
- This year’s event runs under the theme “Scaling Innovation for the AI Era,” focused on large-scale computing infrastructure.
This year’s edition arrives at a moment when data center operators need to connect ever more accelerators while keeping them running with adequate power delivery and thermal management. Generative AI and agentic workloads are raising compute and communication demands, while infrastructure has to adapt to higher power densities per rack.
OCP, an initiative that promotes open designs and technical collaboration for data center equipment, brings together hardware manufacturers, component suppliers, and operators. Its goal is to share advances and specifications that enable interoperable, efficient, and scalable infrastructure.
Compal Bets on High-Voltage Power and Liquid Cooling
Compal Electronics is centering its demonstration on the infrastructure needed to deploy NVIDIA accelerated computing systems at rack scale. The Taiwanese company is presenting a combination of high-density compute platforms, power delivery, cooling, and integrated design, aimed at meeting the demands of future AI facilities.
One of the highlighted elements is the development of 800-volt direct current (800 VDC) power architectures and megawatt-scale electrical systems. Raising the voltage can help reduce losses and the complexity of certain conversion and distribution stages, though the actual result depends on the facility’s overall design.
Compal is also showing liquid cooling technologies and solutions for managing high thermal loads. As the power of accelerators packed into the same space increases, air cooling can become insufficient or less efficient for certain configurations. Liquid systems move heat away from components to dissipation equipment, but they require specific planning for piping, coolant distribution, maintenance, and safety.
The company frames these elements as parts of the same infrastructure. Power delivery determines how much power each rack can receive; cooling determines how heat is removed; and mechanical and network integration affects installation and maintenance of the equipment.
The approach reflects a trend affecting AI infrastructure providers: moving from supplying individual servers to designing complete systems that integrate compute, power distribution, and thermal management. Compal has not announced a specific commercial deployment date in this presentation, nor independently verified energy savings figures for all the solutions it is showing.
WNC Prepares Networks for the New AI Clusters
WNC, a Taiwanese communications equipment manufacturer, is presenting a portfolio focused on increasing data center network capacity. The company is showing 1.6 Tb/s Ethernet switches, 800 Gb/s optical transceivers, and optical circuit switches — technologies designed to connect compute systems with high bandwidth and low latency requirements.
Among the announced equipment are three data center switches with a switching capacity of 102.4 Tb/s. One of the proposals uses optics placed near the integrated circuit package, known as near-packaged optics (NPO). This design places the optical engines close to the switching chip to shorten high-speed electrical paths, aiming to reduce signal loss and power consumption in the connections.
Optical connectivity becomes important when clusters bring together numerous accelerators that constantly exchange data. The speed of each processor doesn’t determine a system’s performance on its own — how long it takes equipment to share information and whether the network can keep pace with distributed operations also matters.
WNC explains that its designs integrate signal, power, thermal management, high-speed connector, and optical engine technologies. The company is also presenting post-quantum cryptography solutions developed with partners, aimed at addressing the potential risks quantum computing poses to current cryptographic systems. These solutions are part of its security offering, though they shouldn’t be confused with universal protection against any future threat.
The 800 Gb/s transceivers and 1.6 Tb/s connections represent different capacity options within high-speed networks. The choice between them depends on the data center’s architecture, connection distance, power requirements, cost, and compatibility with installed equipment.
The Challenge of Building AI-Ready Data Centers
Compal’s and WNC’s proposals show two complementary parts of the infrastructure. One focuses on powering and cooling compute systems; the other, on moving data between accelerators, servers, and racks. Both determine a facility’s real capacity to run large-scale AI workloads.
The need to coordinate these elements also explains the interest in the open designs OCP promotes. Sharing specifications and developing compatible components can make it easier to integrate equipment from different vendors, though effective interoperability depends on implementation and the testing carried out in each environment.
The 2026 summit runs under the theme “Scaling Innovation for the AI Era” and includes technical sessions on compute, networking, power, cooling, storage, firmware, and validation. The organization has confirmed that the San Jose event is sold out and will not offer additional in-person registration.
For Taiwanese manufacturers, the event is an opportunity to showcase capabilities across a supply chain that includes servers, boards, electrical systems, connectivity, and cooling. The presence of these solutions doesn’t by itself confirm new contracts or sales volumes, but it offers a glimpse of where next-generation data center design is heading.
Frequently Asked Questions
When is OCP Global Summit 2026 taking place?
The event runs October 12-15, 2026, in San Jose, California. The organization has announced that it is sold out and will not offer additional in-person registration.
What is Compal presenting at the OCP Summit?
Compal is showing AI data center infrastructure based on NVIDIA accelerated systems, with high-voltage power delivery, 800 VDC proposals, liquid cooling, and rack integration.
What networking technologies is WNC presenting?
WNC is showing 800 Gb/s optical transceivers, 1.6 Tb/s Ethernet switches, and equipment with 102.4 Tb/s switching capacity, along with optical connectivity and security solutions.
Why do AI data centers need better power and cooling systems?
AI accelerators concentrate heavy electricity demand and generate heat during operation. Power delivery and cooling determine the compute density that can be installed in each rack and the efficiency of the infrastructure.
