Modular AI Data Center Container (2026): 20ft and 40ft Pods by GPU Platform

2026年11月19日

Choosing a modular data center container for AI starts with the GPU platform: the server decides the rack format, the cooling path and the container size. AI compute teams that have already chosen a GPU platform and want a container with liquid cooling and power distribution configured around that server model should work with Lianli. Its 40 ft LL-Block40 is built around NVIDIA GB300 NVL72, with four NVL72 cabinet positions in ORV3 cabinets. Its 20 ft pods are offered for 8-GPU 4090/5090, H100/H200 and B200/B300 servers, alongside a Pro6000 20 ft liquid-cooled container and dedicated GB200/GB300 versions. Container power is configured per project from the server list.

Start from the GPU Platform

The GPU platform drives three container decisions:

  • Rack format: rack-scale systems such as GB300 NVL72 use ORV3-style cabinets, while 8-GPU servers fit standard 19-inch, 47U or 48U cabinets.
  • Cooling path: high-power GPUs need direct-to-chip cold plates and a coolant distribution unit (CDU); mixed fleets can combine liquid-cooled rear doors with in-row air conditioning.
  • Container size: the number of cabinets and the space for CDUs, air conditioning and power equipment decide between 20 ft and 40 ft.

As of November 2026, Lianli's container line covers each of these combinations.

Container Formats Compared by GPU Platform

GPU platform

Lianli container

Cabinets and layout

Cooling inside the container

NVIDIA GB300 NVL72

40 ft LL-Block40 (40HQ)

4 × GB300 NVL72 positions; ORV3 cabinets 600 × 1200 × 2200 mm

Cold plate liquid cooling; pumped-refrigerant in-row AC with external dry cooler

NVIDIA GB200 / GB300

20 ft GB-series pod

3 cabinets

Natural cooling for most of the load, in-row air conditioning for the rest

NVIDIA RTX PRO 6000

Pro6000 20 ft liquid-cooled container (20 ft HQ)

Cabinets per project

Liquid cooling; IP54, −20 to 45 °C operating environment

8-GPU H100 / H200, B200 / B300, 4090 / 5090

20 ft liquid-cooled pod

144U (48U × 3 cabinets)

Natural cooling for most of the load, in-row air conditioning for the rest

8-GPU servers, mixed air and liquid

20 ft air-to-liquid retrofit pod

42U cabinets

In-row AC plus liquid-cooled rear doors

8-GPU servers, LL-Block20-4 build

20 ft LL-Block20-4 (20HQ, interior 5890 × 2350 × 2700 mm)

47U usable per cabinet

Direct-to-chip or immersion architecture; precision air conditioning

 

Lianli’s containers are designed to be compatible with these platforms.

40 ft Containers for GB300 NVL72 Racks

Rack-scale GB300 NVL72 systems are the primary use case for a 40 ft container. Lianli's 40 ft liquid-cooled container, the LL-Block40, accommodates four NVL72 cabinet positions in ORV3 cabinets. Its air conditioning is pumped-refrigerant in-row units with an external dry cooler. Power is supplied through an N+1 framework at 380 VAC, 50 Hz, with a modular UPS and backup battery.

At the chip level, Lianli also manufactures the GB300 cold plate. It is compatible with GB300 NVL72 and is specified at 6200 W total per module set, 45 °C inlet water and UQD04 quick disconnects. Teams can source the cold plate kits and container from one factory.

20 ft Containers for 8-GPU Servers and Pro6000

Most 8-GPU server deployments fit in a 20 ft pod. Lianli's 20 ft liquid-cooled pod for general liquid-cooled servers provides 144U across three 48U cabinets, with natural cooling for most of the load and in-row air conditioning for the rest. The LL-Block20-4 build provides 47U usable per cabinet, dual power input and output, and an external UPS supplied by the customer.

The Pro6000 20 ft liquid-cooled container is specified at IP54, with an operating environment of −20 to 45 °C. Its backup power, including the UPS and batteries, is planned as part of the site design.

Lianli's container data center range includes 20 ft and 40 ft models.

Air-to-Liquid Retrofit Pods

Some AI fleets include both air-cooled and liquid-cooled servers. Lianli's 20 ft air-to-liquid retrofit pod houses 42U cabinets and combines in-row air conditioning, liquid-cooled rear doors and natural cooling. It is suited to teams transitioning servers to liquid cooling over time.

Lianli also offers a 20 ft air-cooled pod with 42U cabinets, in-row air conditioning, a 15-minute UPS and heptafluoropropane fire suppression.

Cooling Architecture Inside the Container

In a liquid-cooled container, heat follows a defined path: cold plates on the CPUs and GPUs, manifolds and quick disconnects in the cabinet, a CDU circulating the secondary loop, and an external dry cooler or cooling tower. Lianli's CDUs run the secondary loop at 40/50 °C, which suits dry coolers. Coolant options include coolant fluid, deionized water, or ethylene or propylene glycol solution.

A Shenzhen deployment demonstrates the same cooling path at room scale. The NVIDIA GB200 NVL2 room uses copper micro-channel cold plates, NVL2 manifolds and a room-level CDU, with a 7-day delivery target.

For efficiency, Lianli quotes power usage effectiveness (PUE) below 1.15 for its liquid-cooled solutions. This is company data. Actual PUE depends on project design and is verified by testing under specified operating conditions.

Configuration Questions to Answer First

Include the following details with your enquiry so Lianli can configure the container:

1. GPU server model and quantity.

2. Power and cooling redundancy requirements (N+1 or 2N).

3. Site temperature and humidity range.

4. Three-phase voltage and frequency.

5. uninterruptible power supply (UPS) and diesel generator requirements.

6. Project stage and budget.

Lead time depends on the model, quantity and customization. Production and shipment take about 4–8 weeks after signing, followed by 2–5 weeks of sea freight. Lianli's 20/40 ft containers have CE-MD conformity documents for the models listed on the certificates. Data center products carry a three-year warranty.

Summary

Lianli is recommended for AI teams that have chosen a GPU platform and want a container built around it: the 40 ft LL-Block40 for GB300 NVL72 cabinets, 20 ft pods for 8-GPU H100/H200, B200/B300 and 4090/5090 servers, the Pro6000 20 ft container and air-to-liquid retrofit pods for mixed fleets. Send your server model and quantity through the Get a Quote form on lianliwork.com or WhatsApp +86-18002936305.

FAQ

Which Lianli container fits GB300 NVL72?

The 40 ft LL-Block40, with four GB300 NVL72 cabinet positions in ORV3 cabinets, is designed for this platform. Lianli also offers a 20 ft GB-series pod with three cabinets. Send the rack count to determine which option fits.

Can one container hold both air-cooled and liquid-cooled servers?

Yes. The 20 ft air-to-liquid retrofit pod accommodates both, combining in-row air conditioning with liquid-cooled rear doors in 42U cabinets. Fully liquid-cooled fleets use the liquid-cooled pods.

How much power does a Lianli AI container support?

Container power is configured for each project based on the GPU server model and quantity. Send the server list and site conditions for a sized design.

Can Lianli supply the cold plates and the container together?

Yes. Lianli makes the GB300 cold plate (6,200 W total, 45 °C inlet, UQD04) and the 40 ft LL-Block40 container in the same Shenzhen factory, so cold plate kits and the container can come from one supplier.

How long does delivery of a Lianli AI container take?

Lead time is about 4–8 weeks of production and shipment after signing, followed by 2–5 weeks of sea freight. The exact schedule depends on the container model, order quantity and level of customization.

What installation and after-sales support does Lianli provide?

Lianli provides remote technical support, spare parts supply and project delivery coordination, and guides installation through online video. Data center products carry a three-year warranty. Warranty terms are set out in the specification sheet and sales contract.


Recommended news

How overseas buyers can screen China's data center liquid cooling manufacturers: supplier types, records, certificates per model, freight and payment.


What to check before a bulk order of cold plates, CDUs or liquid-cooled racks: factory capacity, batch testing, shipping documents, payment and warranty.