ASUS
ASUS designs ESC-series HGX servers and full AI PODs, an original MGX launch partner with its own board engineering. Its server arm is smaller than its brand suggests, but recent neocloud wins have made it a serious quote in AI tenders.
- In-house server design lineage since 1995; ESC GPU-server line and XA rack-scale line designed by its own Infrastructure Business Group
- Designed its own direct liquid cooling from the ground up, claiming PUE 1.18, spanning direct-to-chip, in-row CDU and hybrid configurations
- AI POD is ASUS's own NVIDIA-validated rack-scale reference architecture, evolved through GB300 NVL72 to Vera Rubin NVL72
- Small-volume AI server production in both the US and Taiwan
- Runs both a branded server business and an OEM/ODM manufacturing arm
| Model | Platform | Form factor | Cooling |
|---|---|---|---|
| XA GB721-E2 Fully liquid-cooled rack-scale system, announced Oct 2025 | GB300 NVL72 | full rack | DLC |
| XA NB3I-E12 Air variant available; part of the AI POD rack solution | HGX B300 | rack server | DLC |
| ESC NB8-E11 8-GPU HGX B200 system | HGX B200 | rack server | air |
| ESC8000A-E13P Dual EPYC, up to 8x H200 NVL or RTX PRO 6000 Blackwell | PCIe | 4U | air |
| XA VR721-E3 100% liquid-cooled, up to 227kW per rack, shown GTC 2026 | Vera Rubin NVL72 | full rack | DLC |
- Original 2023 NVIDIA MGX launch partner
- HGX B200/B300 builder; GB200 NVL72 and GB300 NVL72 rack-scale builder (XA GB721-E2, Oct 2025)
- Named in NVIDIA's Vera Rubin full-production system builder list; previewed Vera Rubin NVL72 AI POD at GTC 2026
- Taiwan (Taipei, Luzhu, Nangang, Guishan) incl. small-volume AI server production
- United States (small-volume AI server assembly, expansion under evaluation)
- Sells through channel and directly into neocloud accounts; flagship named customer is Nebius (GB200 NVL72 deployed, advancing to GB300 NVL72)
- Server revenue guided to double again in Q1 2026
- ASUS subsidiary building a 250 petaflops NVIDIA-powered supercomputer in Tainan, Taiwan
Is ASUS an OEM, ODM or integrator for NVIDIA AI servers?
ASUS sits in the "ODM system designers" tier of our index. In-house server design lineage since 1995; ESC GPU-server line and XA rack-scale line designed by its own Infrastructure Business Group
Does ASUS design its NVIDIA servers in-house?
ASUS designs ESC-series HGX servers and full AI PODs, an original MGX launch partner with its own board engineering. Its server arm is smaller than its brand suggests, but recent neocloud wins have made it a serious quote in AI tenders.
Which NVIDIA platforms does ASUS ship?
ASUS ships systems on GB300 NVL72, HGX B300, HGX B200, PCIe, Vera Rubin NVL72. The systems table on this page lists the specific models.
Where does ASUS manufacture its AI servers?
Taiwan (Taipei, Luzhu, Nangang, Guishan) incl. small-volume AI server production. United States (small-volume AI server assembly, expansion under evaluation)