AI infrastructure firm Runware has launched the Sonic Inference Pod, a modular, transportable data centre designed to sit alongside the vast facilities being built by hyperscalers such as OpenAI. The company argues that distributed compute positioned closer to end users, rather than a handful of giant fixed sites, will ultimately win out, offering faster and cheaper AI inference while allowing capacity to be added quickly by deploying new pods rather than expanding existing buildings. This matters as demand for AI inference is outstripping the pace at which traditional facilities can be constructed, and Runware's model promises a quicker, less resource-intensive alternative.

The pods use a closed-loop cooling system that requires no water and can reportedly be built in days rather than the months or years needed for conventional data centres. Runware already has ten pods running across the US, Europe and Asia-Pacific, serving clients including Higgsfield AI and Wix, and has identified 160 further sites where pods could be deployed. The firm, which raised a $50 million Series A round in December, says each pod operates as part of a single network so traffic can reroute if one unit fails, and argues that competitors will struggle to replicate the approach quickly given the specialist hardware expertise required.