Direct:
Chenbro is no longer selling sheet metal. The Taiwanese chassis maker, founded in 1983 and later built into an own-brand PC and server chassis business, now wants to be a rack-scale mechanical contractor for AI data centers, and the numbers say the market is letting them. CEO Corona Chen this week framed the pivot explicitly: from chassis to systems to racks to liquid-cooling cabinets and CDUs, on the back of mechanical-design expertise the company is repurposing for thermal infrastructure.
The financial backdrop makes this credible. Chenbro closed 2025 at NT$22 billion in consolidated revenue, up roughly 52% year-over-year, with December a single-month record. Q1 2026 then came in at NT$7.11 billion, up 71% year-over-year in what is normally a slow quarter, with net income roughly doubling. Server chassis accounted for ~99% of revenue in Q4 2025; AI-related projects are already more than half of sales. That is the cash flow funding a Texas plant announced last August (up to NT$2 billion) and a six-hectare Johor site in Malaysia near the Southeast Asian hyperscale buildout.
The technical driver is GB200 and what comes after it. NVL72 racks pull roughly 120 kW, Vera Rubin is specced for 45°C warm-water direct liquid cooling, and Nvidia has publicly floated 1 MW racks down the road. Air cooling fails at these densities for basic physics reasons (water moves heat ~25x better than still air), which is why CDU capacity at the rack level has become one of the gating infrastructure components for AI deployments, alongside power delivery, facility water, manifolds, and memory. Chenbro is already aligned with Nvidia's MGX ecosystem, with GB200 NVL72/NVL36 tray designs and liquid-cooled MGX chassis in its public product roadmap. Moving up into CDUs and full rack integration is the logical next step, and it puts them in direct contact with the thermal-module incumbents (AVC, Auras, Boyd, Delta, Motivair/Schneider).
The risks Chen has flagged are not subtle: US tariff policy, Red Sea shipping costs up 40-50%, memory supply gaps, and the persistent “AI bubble” question. None of those change the immediate trajectory, but they explain why the geographic diversification (Taiwan, China, Malaysia, Texas) is happening now.
For consumer storage readers this is mostly tangential, but worth noting: the same hyperscale capex driving Chenbro's rack business is the capex absorbing NAND and HBM allocation. AI server demand is pulling DRAM and HBM into long-dated contract commitments while enterprise SSDs eat up the high-density NAND, leaving client and consumer segments competing for what's left at rising prices. The chassis pivot to liquid cooling is downstream of the same demand curve squeezing client drive availability.
Sources
- DigiTimes, “Chenbro expands from AI server chassis into liquid-cooling CDU” (May 12, 2026)
- DigiTimes, “AI server demand pushes Chenbro 2025 revenue above NT$20 billion” (Jan 2026)
- DigiTimes, “Nvidia GB200 fuels chassis sector pivot to liquid cooling, rack integration” (Feb 2026)
- DigiTimes, “Chenbro highlights four risk factors amid promising AI server outlook” (Mar 2026)
- TVBS English, “Server Chassis Maker Chenbro Posts Record Q1 on AI Demand” (Apr 2026)
- Taipei Times, “Chenbro delivers record quarter” (Nov 2025)
- VnExpress International, profile of chairwoman Maggi Chen (Apr 2026)
- Chenbro corporate “Organizational” / company history page
- Chenbro/PRNewswire, OCP Summit 2024 product announcement
- Network World, “Why AI rack densities make liquid cooling nonnegotiable” (Mar 2026)
- Nvidia Developer Blog, GB200 NVL72 OCP contribution / DCI partner list (Nov 2024); Rubin platform technical blog; 800 VDC architecture blog
- TrendForce press release on memory pricing and AI-driven allocation (Mar 31, 2026)
Reddit: https://ift.tt/NWwTot7