Smartbird Locks In 3.7 MW of Data Center Capacity for 2027 AI Push

Smartbird has signed a multi-year colocation agreement securing more than 3.7 megawatts of IT capacity at an Expedient U.S. data center, with dedicated AI infrastructure deployments planned for the first half of 2027. The company announced the deal on November 12. It follows Smartbird's earlier technology partnerships with Supermicro and WEKA, and it addresses a constraint that hardware purchases alone cannot solve: a place to power, cool, and operate the machines.

For context, 3.7 megawatts is roughly the continuous electrical load of several thousand average U.S. homes, though a megawatt figure describes only the power available to IT equipment — not GPU counts or inference throughput. Those depend on server selection, storage, networking, and workload behavior. Smartbird says the Expedient facility can accommodate high-density systems, including liquid-cooled racks.

"Access to power and AI-ready data center capacity is one of the most important requirements for delivering infrastructure at scale," said Nadia Carlsten, Smartbird's president and CEO.

What the announcement leaves open

The deal reserves space and power; it does not deliver commissioned computing capacity. Three specifics are missing from the disclosure:

• The announcement does not identify the facility.
• It does not disclose financial terms.
• It does not specify which accelerators will be installed.

Smartbird frames the capacity as support for opportunities in its active customer pipeline — a pipeline the company does not quantify. Reserved capacity establishes where future systems can run; hardware installation, integration, and commissioning still determine when a cluster begins useful work.

Why cooling shapes the computing architecture

Densely packed accelerators concentrate heat in a small footprint, which makes heat removal part of the system design rather than an afterthought. Liquid cooling transfers heat through a fluid circuit instead of relying entirely on air moving through servers. Supporting such racks expands the range of systems a facility can host, although Smartbird's eventual configuration remains open.

Cooling matters because AI bottlenecks move between layers of the stack. During training, accelerators need a steady supply of data and fast communication between participating systems. During inference, memory capacity, response latency, and the number of simultaneous requests determine how effectively hardware serves users. A cluster's useful output depends on the whole system working together, not on accelerator specifications alone.

How Smartbird's managed model works

On its website, Smartbird describes a service spanning:

• infrastructure design and procurement
• installation and testing
• operations, monitoring, and maintenance
• performance optimization and capacity expansion as workloads evolve

The model starts with the customer's workload and coordinates compute, storage, networking, power, and cooling around those requirements. The commercial pitch is dedicated infrastructure without forcing enterprises to assemble and operate every layer themselves.

The partner stack now covers three layers. Supermicro addresses computing platforms and rack-scale integration. WEKA adds NeuralMesh data and memory infrastructure to the dedicated-cluster offering. Expedient supplies the facility layer beneath that technology stack.

What Expedient contributes

Expedient operates a network of U.S. data centers and provides cloud, colocation, and managed infrastructure services. Its colocation offering lets organizations host equipment in its facilities while drawing on power, connectivity, physical security, and on-site operational support.

The provider describes N+1 or 2N redundancy for data-center components — approaches that keep spare capacity or duplicated systems available when equipment fails. It also offers private connectivity between its facilities and round-the-clock staffing.

The announcement does not specify the exact redundancy design or network configuration for Smartbird's installation. Nor does the agreement establish that Expedient's separate AI software products will become part of Smartbird's service.

What comes next

The planned first-half 2027 deployments are the next milestone, and the gap between now and then is where the risk sits: reserved megawatts are not running clusters. Smartbird's approach assembles specialist providers for computing systems, the data layer, and now facilities, with the company itself responsible for integrating and operating those elements around enterprise workloads. The practical significance of the Expedient agreement is that it connects Smartbird's managed AI proposition to the power and cooling required to turn that proposition into operating infrastructure — on a timeline still more than two years out.