Why CUBE 5.0 is different

Traditional data center construction often involves installing major systems in sequence. Alibaba Cloud’s approach moves more of that work into factories, where modules can be produced and tested in parallel before being transported to the site for assembly. Reports describe the process as similar to putting together building blocks.

CUBE 5.0 applies this model to five major systems: power supply, cooling, security, intelligent management and fire protection. Alibaba Cloud says their overall modularization rate has risen from roughly 30% in earlier designs to 90%.

That distinction matters because modularization is not simply about using containers. It changes when and where work is done: more integration and testing happen before the equipment reaches the data center site, reducing the amount of sequential construction required there.

How the 100-day delivery schedule works

Alibaba Cloud’s reported timeline is divided into three stages:

1. 30 days: Factory production and site preparation
2. 50 days: On-site installation
3. 20 days: Commissioning and testing

Together, these stages total 100 days. The reported measurement begins with a prepared foundation or stacked modular units and ends after the power, cooling, security, intelligent management and fire-protection systems are deployed and tested. Servers and network equipment can then enter the facility.

This scope is an important caveat. The 100-day figure describes delivery of the data center infrastructure and its core systems; it should not automatically be read as a promise that a fully equipped, production-loaded AI campus will be running from an empty site in 100 days.

For comparison, reports cited conventional delivery cycles of six to 12 months in China and 12 to 18 months in the United States for comparable large-scale projects.

Air and liquid cooling in the same architecture

CUBE 5.0 is designed to support both air-cooled and liquid-cooled configurations. Alibaba Cloud reported a PUE of no more than 1.15 for air cooling and no more than 1.10 for liquid cooling.

The ability to switch between cooling approaches is intended to give operators more flexibility as chip power density, local climate and workload requirements change. It also reduces the risk of locking a new facility into a single cooling configuration before the hardware mix is known.

Power and chip compatibility

The architecture incorporates high-voltage direct-current power systems alongside what reports describe as a highly flexible design. Alibaba Cloud says this configuration is intended to accommodate mainstream heterogeneous chips and computing architectures across at least three generations of hardware.

That forward-compatibility claim is particularly relevant for AI infrastructure, where accelerator designs and server configurations can change rapidly. A modular facility that can be reconfigured for different chip generations may require fewer major physical changes as computing requirements evolve.

Reported efficiency and density gains

Beyond faster delivery, Alibaba Cloud reported several additional benefits compared with its previous generation. These include construction costs more than 10% lower, computing density per unit area five to 10 times higher, and a near-halving of transformer numbers.

The company also reported first-time testing pass rates approaching 100%. These figures are company claims reported by secondary sources, rather than independently verified benchmarks, so they are best understood as targets or reported results rather than universal performance guarantees.

What the announcement signals

Alibaba Cloud’s capacity expansion and CUBE 5.0 claims point to a broader shift in AI infrastructure: speed of construction is becoming almost as important as the chips installed inside the facility. By standardizing more of the power, cooling and safety infrastructure in factories, cloud providers can try to shorten deployment cycles and make future expansions more repeatable.

The strongest takeaway is not simply the headline number of 100 days. It is the combination of 90% modularization, parallel factory production, switchable cooling and multi-generation hardware support. Those design choices are aimed at making AI data centers faster to build, denser and easier to adapt—although the reported performance and cost figures still come from Alibaba Cloud and associated reports and should be validated project by project.

Why CUBE 5.0 is different

Traditional data center construction often involves installing major systems in sequence. Alibaba Cloud’s approach moves more of that work into factories, where modules can be produced and tested in parallel before being transported to the site for assembly. Reports describe the process as similar to putting together building blocks.

CUBE 5.0 applies this model to five major systems: power supply, cooling, security, intelligent management and fire protection. Alibaba Cloud says their overall modularization rate has risen from roughly 30% in earlier designs to 90%.

That distinction matters because modularization is not simply about using containers. It changes when and where work is done: more integration and testing happen before the equipment reaches the data center site, reducing the amount of sequential construction required there.

How the 100-day delivery schedule works

Alibaba Cloud’s reported timeline is divided into three stages:

1. 30 days: Factory production and site preparation
2. 50 days: On-site installation
3. 20 days: Commissioning and testing

Together, these stages total 100 days. The reported measurement begins with a prepared foundation or stacked modular units and ends after the power, cooling, security, intelligent management and fire-protection systems are deployed and tested. Servers and network equipment can then enter the facility.

This scope is an important caveat. The 100-day figure describes delivery of the data center infrastructure and its core systems; it should not automatically be read as a promise that a fully equipped, production-loaded AI campus will be running from an empty site in 100 days.

For comparison, reports cited conventional delivery cycles of six to 12 months in China and 12 to 18 months in the United States for comparable large-scale projects.

Air and liquid cooling in the same architecture

CUBE 5.0 is designed to support both air-cooled and liquid-cooled configurations. Alibaba Cloud reported a PUE of no more than 1.15 for air cooling and no more than 1.10 for liquid cooling.

The ability to switch between cooling approaches is intended to give operators more flexibility as chip power density, local climate and workload requirements change. It also reduces the risk of locking a new facility into a single cooling configuration before the hardware mix is known.

Power and chip compatibility

The architecture incorporates high-voltage direct-current power systems alongside what reports describe as a highly flexible design. Alibaba Cloud says this configuration is intended to accommodate mainstream heterogeneous chips and computing architectures across at least three generations of hardware.

That forward-compatibility claim is particularly relevant for AI infrastructure, where accelerator designs and server configurations can change rapidly. A modular facility that can be reconfigured for different chip generations may require fewer major physical changes as computing requirements evolve.

Reported efficiency and density gains

Beyond faster delivery, Alibaba Cloud reported several additional benefits compared with its previous generation. These include construction costs more than 10% lower, computing density per unit area five to 10 times higher, and a near-halving of transformer numbers.

The company also reported first-time testing pass rates approaching 100%. These figures are company claims reported by secondary sources, rather than independently verified benchmarks, so they are best understood as targets or reported results rather than universal performance guarantees.

What the announcement signals

Alibaba Cloud’s capacity expansion and CUBE 5.0 claims point to a broader shift in AI infrastructure: speed of construction is becoming almost as important as the chips installed inside the facility. By standardizing more of the power, cooling and safety infrastructure in factories, cloud providers can try to shorten deployment cycles and make future expansions more repeatable.

The strongest takeaway is not simply the headline number of 100 days. It is the combination of 90% modularization, parallel factory production, switchable cooling and multi-generation hardware support. Those design choices are aimed at making AI data centers faster to build, denser and easier to adapt—although the reported performance and cost figures still come from Alibaba Cloud and associated reports and should be validated project by project.