Quanta / QCT
Quanta is the original hyperscale ODM and QCT is its direct-sale arm. It designs boards and rack systems for the largest clouds, and the same engineering is available over the counter through QCT for buyers who can self-support.
- Designs its own server boards and compute trays; builds complete GB200 NVL72 racks from 18x QuantaGrid D75B-1U trays
- Markets solutions validated at L10 (system), L11 (rack) and L12 (cluster) integration levels
- Dual-track model: custom ODM designs for hyperscalers (Meta, Google, Microsoft, AWS) plus NVIDIA reference systems sold under the QCT brand to enterprises and telcos
- Own rack-scale liquid-cooling line (QoolRack) removing up to 75kW of heat per rack, cutting cooling power up to 66.8% vs air
| Model | Platform | Form factor | Cooling |
|---|---|---|---|
| QuantaGrid D75B-1U 2 Superchips per tray, building block of the QCT NVL72 rack | GB200 NVL72 | 1U compute tray | DLC |
| QuantaGrid S74G-2U First Grace Hopper MGX server | MGX | 2U | air |
| QCT QoolRack Drops into existing air-cooled rows without new DC infrastructure | GB200 NVL72 | rack-level cooling | DLC |
| QuantaGrid HGX systems Sold under the QCT brand to enterprises and telcos | HGX B200 / HGX B300 | rack server | air or DLC |
- Long-standing HGX and MGX design partner; QuantaGrid S74G-2U was the first server combining Grace Hopper with MGX (COMPUTEX 2023)
- First shipment of its GB200 NVL72 announced May 2025; GB300 NVL72 production from Q3 2025
- Named among adopters of NVIDIA DSX accelerating the Vera Rubin ramp
- Industry tracking put Quanta at roughly 300-400 GB200 NVL72 racks shipped in April 2025 alone
- Taoyuan, Taiwan (HQ, R&D and manufacturing)
- Shanghai, Changshu and Chongqing, China
- Fremont, California (expanding Bay Area footprint, ~$45.6M additional lease investment 2025)
- La Vergne, Tennessee (Quanta Manufacturing Nashville)
- Anchor customers: Microsoft Azure, AWS, Meta, Google Cloud, plus NVIDIA itself as a reference-system partner
- QCT exists specifically to sell Quanta's hyperscaler-honed engineering to enterprises and research institutions over the counter
- AI server orders stated secured through 2027; AI server capacity expected to double in 2026
Is Quanta / QCT an OEM, ODM or integrator for NVIDIA AI servers?
Quanta / QCT sits in the "ODM system designers" tier of our index. Designs its own server boards and compute trays; builds complete GB200 NVL72 racks from 18x QuantaGrid D75B-1U trays
Does Quanta / QCT design its NVIDIA servers in-house?
Quanta is the original hyperscale ODM and QCT is its direct-sale arm. It designs boards and rack systems for the largest clouds, and the same engineering is available over the counter through QCT for buyers who can self-support.
Which NVIDIA platforms does Quanta / QCT ship?
Quanta / QCT ships systems on GB200 NVL72, MGX, HGX B200 / HGX B300. The systems table on this page lists the specific models.
Where does Quanta / QCT manufacture its AI servers?
Taoyuan, Taiwan (HQ, R&D and manufacturing). Shanghai, Changshu and Chongqing, China. Fremont, California (expanding Bay Area footprint, ~$45.6M additional lease investment 2025). La Vergne, Tennessee (Quanta Manufacturing Nashville)