Powering The AI Era: High-Density Racks, Cutting-Edge Systems, And Advanced Cooling From Wiwynn

Author: Marcus, Leader of Power Engineering Team, Wiwynn Collection: 2025 OCP Global Summit - Expo Hall Sessions

Summary

Wiwynn, a Taiwanese AI server and data center solution provider, presents a comprehensive overview of its rack-level, system-level, and component-level technologies for AI data centers. The presentation opens with an AI-generated promotional video (produced in one week for $100 in tokens) before diving into market context: AI CapEx investment continues growing at 24%+ in 2025, with a compound annual growth rate of 23% through 2028. Computing power demand has shifted from 8-fold growth per two years (pre-Transformer era) to 250-fold per two years, while chip TDP has reached 1,000W and is projected to hit 2,000-3,000W per chip in the near future.

Wiwynn's rack-level solutions feature a double-wide rack designed to accommodate chipsets approaching 15 cm x 15 cm, providing more PCB space for cable routing and liquid cooling components (cold plates, quick connectors, hoses). The double-wide rack uses a liquid-cooled green busbar (to address thermal-limited bus bar current), features modular design at the rear end for easier service, and incorporates chassis stiffness enhancements to prevent sag in wider form factors. System-level solutions include platforms for both NVIDIA and AMD, with chassis compatible with both power supplies and busbar solutions. Component-level cooling solutions span package-level (PLC in/out for direct chipset heat dissipation), module-level (double-sided cold plates for ASIC and VPD cooling, two-phase cold plates), system-level (fanless servers with extended cold plates, immersion cooling chassis), and rack-level (AOC solutions for minimal-impact liquid cooling upgrades).