Runware Tests a Portable Future for AI Data Centers With Sonic Inference Pod

Runware announced Sonic Inference Pod, a modular data center designed as a single transportable unit for AI inference. The pitch is flexibility: instead of expanding a fixed facility, capacity can be added by creating and deploying more pods.

The company frames the pod as a complement to hyperscalers’ massive data center projects, not simply a replacement. For readers tracking AI infrastructure, the key takeaway is that inference demand is pushing companies to test smaller, distributed formats alongside traditional builds.

Runware says the Pod can offer inference at higher quality and lower cost than other serverless inference platforms and GPU clouds. CEO Flaviu Radulescu told TechCrunch that distributed compute positioned closer to end users is what the company believes will win over the long term.

Runware also says the system can deploy anywhere there is power, adapt quickly to new hardware releases, and use a closed-loop cooling system rather than water. The company says that cooling approach can be built in days, compared with the longer timelines associated with traditional data centers.

Radulescu said Runware currently has 10 pods in deployment across the U.S., Europe, and Asia-Pacific. The company also says it has 160 sites available to power pods right now.

Runware already provides inference to companies including Higgsfield AI and Wix. The pod rollout follows Runware’s previously announced $50 million Series A in December, tied to infrastructure for companies generating images.

AI data centers remain controversial because of resource use, and communities around data centers have reported rising utility costs. Runware says it wants a future where its systems can run on renewable power and avoid drawing on resources communities need, but the article notes that future is not necessarily here today.

Radulescu argued that AI power use will keep rising because inference demand is growing, regardless of who supplies it. Runware’s position is that pods can reduce some infrastructure pressure by using existing power, avoiding water in cooling, and reducing transmission losses.