Wistron
Wistron builds the HGX baseboards and DGX systems for NVIDIA itself, which is about as deep in the supply chain as a vendor can sit. Its own-brand server business is smaller; its significance is as the design-manufacturer behind other people's badges, now expanding US production in Texas.
- Self-described world leader in AI baseboard and GPU platform manufacturing; NVIDIA HGX Partner Program member with early reference-architecture access
- For Vera Rubin racks, Wistron (with Quanta and Foxconn) handles assembly up to L10 (bootable server) before NVIDIA ships direct, a change from the prior full-rack ODM model
- GB200/GB300 NVL72 rack-level liquid-cooled systems developed with sister company Wiwynn
- Collaborated with NVIDIA on Omniverse factory digital twins and physics-informed AI heat-flow simulation for GB200 production
| Model | Platform | Form factor | Cooling |
|---|---|---|---|
| HGX baseboard systems Platform capability rather than named retail SKUs | HGX B200 / HGX B300 | server | air or DLC |
| GB200/GB300 NVL72 racks (with Wiwynn) Rack-level liquid cooling engineered jointly with Wiwynn | GB200 NVL72 / GB300 NVL72 | full rack | DLC |
| RTX PRO 6000 Blackwell platform Added to the AI server platform Aug 2025 | MGX | server | air |
- NVIDIA HGX Partner Program member; named RTX PRO Server partner
- NVIDIA reportedly booked the entire output of Wistron's new Zhubei plant through 2026 for Blackwell and Rubin systems
- 2023 precedent: NVIDIA shifted ~10% of A100 PCB orders to a Foxconn subsidiary on Wistron capacity constraints, a supply-reliability data point
- Zhubei, Taiwan: new AI supercomputer plant (opened June 2025), output reportedly booked by NVIDIA through 2026; second Taiwan plant planned
- Fort Worth (AllianceTexas), Texas: two sites, $761M, ~1.1M sq ft, operational H1 2026, among the first US plants making AI supercomputers domestically
- Dallas and Houston, Texas: additional assembly and test sites
- Ciudad Juarez, Mexico
- Builds the HGX baseboards and systems behind other vendors' badges; predominantly build-to-order for hyperscalers and NVIDIA, not a direct-to-enterprise seller
- Wistron, Foxconn, Quanta and Inventec together held 53.2% of the server market in Q4 2025
- Capacity risk history: 2023 order shift and reports its capacity may not meet 2026 demand
Is Wistron an OEM, ODM or integrator for NVIDIA AI servers?
Wistron sits in the "ODM system designers" tier of our index. Self-described world leader in AI baseboard and GPU platform manufacturing; NVIDIA HGX Partner Program member with early reference-architecture access
Does Wistron design its NVIDIA servers in-house?
Wistron builds the HGX baseboards and DGX systems for NVIDIA itself, which is about as deep in the supply chain as a vendor can sit. Its own-brand server business is smaller; its significance is as the design-manufacturer behind other people's badges, now expanding US production in Texas.
Which NVIDIA platforms does Wistron ship?
Wistron ships systems on HGX B200 / HGX B300, GB200 NVL72 / GB300 NVL72, MGX. The systems table on this page lists the specific models.
Where does Wistron manufacture its AI servers?
Zhubei, Taiwan: new AI supercomputer plant (opened June 2025), output reportedly booked by NVIDIA through 2026; second Taiwan plant planned. Fort Worth (AllianceTexas), Texas: two sites, $761M, ~1.1M sq ft, operational H1 2026, among the first US plants making AI supercomputers domestically. Dallas and Houston, Texas: additional assembly and test sites. Ciudad Juarez, Mexico