At Computex 2026 in Taipei, Wiwynn (cloud IT infrastructure provider) is presenting what the company describes as a full-stack, rack-scale technology blueprint for next-generation data centers, spanning AI compute platforms, liquid cooling innovations, optical interconnects, and high-voltage DC power delivery.

The company is exhibiting at Booth J0106 in TaiNEX1, with demonstrations developed alongside ecosystem partners including Wistron, Delta, TE Connectivity, Ayar Labs, and GUC.

AI Compute at Rack Scale

Among the centerpiece announcements is Wiwynn's readiness for the NVIDIA Vera Rubin NVL72, a fully liquid-cooled, rack-scale platform developed jointly with Wistron.

The platform unifies 72 NVIDIA Rubin GPUs with 36 NVIDIA Vera CPUs and is positioned for frontier AI model training, inference, and reasoning workloads. According to Wiwynn, the platform delivers up to ten times higher performance per watt through what the company calls extreme co-design.

Alongside the NVIDIA platform, Wiwynn is showing an AMD Helios rack-scale solution built on the OCP ORv3 Open Rack Wide specification. The system is powered by AMD Instinct MI400 Series GPUs, sixth-generation AMD EPYC CPUs, AMD Pensando networking technologies, and the AMD ROCm software stack.

Wiwynn describes its role as bringing the AMD Helios platform from reference design into production, with the stated goal of accelerating open, interoperable AI infrastructure for hyperscale deployments.

A storage server prototype is also on display, featuring 96 liquid-cooled E3.S solid-state drives delivering high aggregate input/output operations per second. The system is accessed using NVIDIA SCADA and employs NVIDIA GPUs capable of saturating PCIe connections to the drives, targeting applications in graph neural networks and vector databases.

Rethinking Data Center Power with 800V DC

One of the more architecturally significant demonstrations at the Wiwynn booth addresses power delivery. As rack power density continues to climb, the company is showcasing an 800 VDC rack-level solution developed with Delta and TE Connectivity.

The demonstration covers the full chain from the power rack through a liquid-cooled busbar to an 800-volt-to-50-volt power distribution board, which Wiwynn describes as an end-to-end high-voltage DC system integration designed for high-density AI deployments.

William Lin, President and CEO of Wiwynn, framed the broader challenge at the company's Computex announcement. "AI is no longer just a computing challenge. It's a full-system engineering challenge. Compute, thermal management, data transmission, and power delivery must be co-designed at rack scale," Lin said.

Moving Beyond Copper with Co-Packaged Optics

Bandwidth constraints within AI infrastructure are the focus of a separate demonstration involving co-packaged optics. Working with Ayar Labs and GUC, Wiwynn is showing a CPO-based optical scale-up rack design that integrates an AI application-specific integrated circuit with a 2.5D advanced package, TeraPHY optical engines, ELSFP SuperNova remote light sources, advanced fiber routing, and optical interconnects throughout the rack.

Wiwynn is also highlighting an in-house liquid cooling solution developed specifically for the ELSFP remote light source components, stating it ensures reliable in-chassis optical connections.

The company characterizes the demonstration as a critical milestone in moving co-packaged optics from component-level innovation toward hyperscale-ready deployment.

Cooling Technologies Designed for Extreme Power Envelopes

Two new cooling technologies are being introduced at the show to address what Wiwynn describes as chip power envelopes reaching new extremes.

The first is a 6kW double-sided cold plate using liquid metal thermal interface material. Designed to support AI ASICs drawing up to 6 kilowatts, the solution uses an ultra-thin construction integrating cavity PCB architecture to support vertical power delivery.

Wiwynn says this design shortens power delivery paths and improves power delivery efficiency by more than 80 percent within a compact form factor. The use of liquid metal thermal interface material is stated to reduce silicon-to-cold plate temperature differentials, resulting in over 30 percent gains in thermal efficiency.

The second cooling innovation involves diamond composite material. Wiwynn describes diamond composite as a next-generation thermal material offering superior thermal conductivity compared to copper, combined with a weight advantage.

The material is demonstrated in a microchannel cold plate design, and the company points to potential future applications at the IC package level as an area of ongoing interest.

Ecosystem and Manufacturing Context

Wiwynn describes its end-to-end capabilities as spanning server design, system integration, and high-volume rack delivery at what it designates L10 and L11 levels.

The company maintains a manufacturing footprint across Taiwan, the United States, Mexico, and Malaysia. The Computex exhibition reflects Wiwynn's stated positioning as a cloud IT infrastructure provider serving major data centers and cloud service providers globally.

The demonstrations at Computex 2026 run across all four of the technical domains Wiwynn has identified as integral to modern AI data center design: compute, thermal management, optical data transmission, and power delivery, with each presented as part of a co-designed rack-scale system rather than standalone component advances.

Taiwan Data Center Pipelines Are Moving Fast. Is Your Intelligence Keeping Up?

Every week, new hyperscale commitments, land acquisitions, and colocation expansions are reshaping the Taiwan data center landscape at a pace that leaves many market participants reacting rather than anticipating. If your team is still piecing together project intelligence from fragmented sources, the gap between you and better-informed competitors is widening with every announcement you miss.

The Global Project Tracking (GPT) platform by Blackridge Research was built to close that gap. It consolidates the full lifecycle of data center development activity across Taiwan into a single, structured intelligence environment, from the earliest planning signals through to project completion, so your business development, investment, and procurement teams are always working from a complete picture.

Whether you are targeting hyperscale campuses in Taiwan or the rapidly expanding digital infrastructure opportunities, the platform surfaces the opportunities that matter before they become common knowledge.

• Upcoming Projects

• Tender Notices

• Contract Awards

• Projects Under Construction

• Completed Projects

See the full depth of Taiwan data center project coverage for yourself. Contact Blackridge Research today to Book a Free Demo of the Global Project Tracking (GPT) platform.