Lenovo
Lenovo designs ThinkSystem AI servers around its Neptune liquid-cooling engineering, a genuinely in-house warm-water DLC capability it has iterated for over a decade. Global manufacturing footprint and aggressive pricing make it the usual third quote in enterprise deals.
- Lenovo Neptune: proprietary warm-water direct liquid cooling, now 6th generation; 5th gen claimed up to 40% lower power vs air
- ThinkSystem SR680a V4 described by Lenovo as designed fully in-house from the ground up
- Own chassis engineering: SC777 V4 uses a vertical 21-inch compute tray in the Lenovo-specific N1380 13U enclosure, not a stock NVIDIA reference chassis
- XClarity Controller: Lenovo's own BMC firmware across ThinkSystem
- Rack-scale integration of the 48U MGX-compliant GB300 NVL72 rack (~90% liquid cooled)
| Model | Platform | Form factor | Cooling |
|---|---|---|---|
| Lenovo NVIDIA GB300 NVL72 18x 1U compute trays, 9 NVLink switch trays, ~90% liquid cooled | GB300 NVL72 | 48U rack-scale | DLC |
| ThinkSystem SC777 V4 Neptune 2x Grace plus 4x Blackwell per tray | GB200 NVL4 | vertical 21-inch tray, 8 per 13U N1380 enclosure | DLC |
| ThinkSystem SR780a V3 Neptune water-cools the 8-GPU HGX board; claimed PUE 1.1 | HGX B200 | 5U | DLC |
| ThinkSystem SR680a V4 8x B300 (2.3TB HBM3e), ConnectX-8 SuperNICs, designed fully in-house per Lenovo | HGX B300 | 8U | air |
| ThinkSystem SR685a V3 Hopper-era system still sold | HGX H100/H200 | 8-GPU rack server | TBD |
- System builder for HGX B200 (SR780a V3), HGX B300 (SR680a V4) and GB300 NVL72 rack-scale
- Named in NVIDIA's Vera Rubin launch and full-production builder lists
- Multiple ThinkSystem models are NVIDIA-Certified; first partner to market with NVIDIA-Certified A100 servers
- AI Cloud Gigafactory offering with NVIDIA announced at CES 2026, targeting neocloud providers
- Ullo, Hungary (first European in-house plant, 49,000 sqm, ships servers and water-cooled systems for Europe)
- Monterrey, Mexico (plant plus AI Innovation Centre for North America)
- Saudi Arabia (new facility under construction with PIF-backed Alat)
- #1 global provider on the Top500 supercomputer list (169 systems, roughly 1 in 3)
- Liquid-cooled server revenue up 154% YoY in a recent quarter; ISG posted record $19.2B revenue
- Lenovo/NVIDIA claim Lenovo technology is used by 8 of the top 10 global cloud providers (vendor claim)
- No named hyperscaler mega-deal publicly attributed
Is Lenovo an OEM, ODM or integrator for NVIDIA AI servers?
Lenovo sits in the "Global OEMs" tier of our index. Lenovo Neptune: proprietary warm-water direct liquid cooling, now 6th generation; 5th gen claimed up to 40% lower power vs air
Does Lenovo design its NVIDIA servers in-house?
Lenovo designs ThinkSystem AI servers around its Neptune liquid-cooling engineering, a genuinely in-house warm-water DLC capability it has iterated for over a decade. Global manufacturing footprint and aggressive pricing make it the usual third quote in enterprise deals.
Which NVIDIA platforms does Lenovo ship?
Lenovo ships systems on GB300 NVL72, GB200 NVL4, HGX B200, HGX B300, HGX H100/H200. The systems table on this page lists the specific models.
Where does Lenovo manufacture its AI servers?
Ullo, Hungary (first European in-house plant, 49,000 sqm, ships servers and water-cooled systems for Europe). Monterrey, Mexico (plant plus AI Innovation Centre for North America). Saudi Arabia (new facility under construction with PIF-backed Alat)