Runware launched the modular Sonic Inference Pod data center for AI inference
Runware launched Sonic Inference Pod, a modular portable data center for AI inference with water-free cooling that can be built within days. There are 10 pods operating in the USA, Europe, and Asia-Pacific, clients include Higgsfield AI and Wix, and 160 additional locations are available.
Runware launched Sonic Inference Pod, a modular and portable data center designed for AI inference. The unit is designed as a standalone transportable unit that, according to the company, can complement large-scale data center projects operated by hyperscalers. The pods use a closed-loop cooling system that consumes no water, and according to the company, they take days to build, while traditional data centers take months to years.
According to Runware, Sonic Inference Pod offers higher-quality inference at lower prices than competing serverless inference platforms and GPU clouds. The modular design is intended to enable rapid capacity expansion by adding new pods instead of expanding a fixed data center building, rapid deployment anywhere electricity is available, and easy adaptation to new generations of hardware. CEO and co-founder Flaviu Radulescu said that all pods operate as a single network — requests are routed to where capacity is available, and if one pod goes down, traffic shifts to another, so a failure affects only that pod, rather than the entire facility.
Runware currently operates 10 pods in the USA, Europe, and the Asia-Pacific region and provides inference to companies including Higgsfield AI and Wix. Another 160 locations are available for deploying new pods. In December 2025, the company raised 50 million dollars in Series A funding to build infrastructure for image generation and sees its expansion into pods as part of its broader mission to provide inference to businesses in general, rather than just one specific product.
Radulescu mentioned that demand for AI computing power will grow regardless of who meets it, and in his view, the approach taken by Runware — with no energy transmission losses, no water used for cooling, and the use of existing electrical capacity — means less demand for new grid infrastructure and water for the same amount of computing power.
Why it matters
For companies running AI products, this is a potential alternative to traditional data centers and serverless GPU clouds — according to Runware, modular pods can be deployed within days anywhere electricity is available, and the distributed network of pods also eliminates a single point of failure because traffic is redirected to another pod during an outage. Companies seeking inference capacity can therefore consider a dedicated pod as a complement or replacement for renting GPUs from major cloud providers, although claims of lower prices and higher quality so far come only from the company itself.
Relevant practical impact
What this means
For a business
Companies running AI products gain another option for deploying inference capacity outside traditional data centers and hyperscalers — according to Runware, modular pods can be deployed within days anywhere electricity is available, and customers can also rent an entire pod exclusively for themselves.
DevelopmentCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.