Runware, a company specializing in AI infrastructure, is once again in the spotlight, just months after the introduction of its Sonic Inference Pod, a modular data center manufactured in a factory and delivered to clients ready for installation. The firm has ambitious plans ahead. According to hinews, Runware aims to establish a factory dedicated to assembling these pods on a production line, akin to automotive manufacturing. While the concept is not entirely novel—companies like Crusoe and Schneider Electric also offer prefabricated pods—Runware asserts that it can expedite the shipping of its products.
Flaviu Radulescu, co-founder and CEO of Runware, informed hinews that the company intends to scale production at its factory in Romania to achieve a data center capacity of 30 megawatts per day by 2028, equating to nearly 1 gigawatt monthly. This significant capacity means that, at this rate, Runware could generate as much in a year as the entire U.S. is projected to add in data center power this year, approximately 10 gigawatts. This ambition also mirrors a broader industry trend aimed at deploying data centers more rapidly and efficiently.
For instance, Nvidia recently unveiled the Vera Rubin AI factories and the DSX POD, both designed to facilitate quicker and simpler construction of AI data centers. However, Runware distinguishes itself by claiming it has actually built a factory rather than merely conceptualizing the infrastructure that would enable others to do so.
Radulescu stated that Runware plans to incrementally boost production starting next year, noting that the company currently operates two factories: one that has been utilized for several years to manufacture the pods and a second facility that began construction earlier this year. The latter is expected to be operational within months and will commence production early next year.
Runware will produce the modular data centers on an assembly line. Clients can either provide their own site for assembly or allow Runware to locate one for them. Radulescu mentioned that the company aims to have the pods operational within 48 hours, a timeframe that includes installation and setup, but not construction. Previously, a pod took three weeks to build, but with the ramp-up in production, this duration is expected to reduce to just one day, according to the company. Runware will also supply spare parts in the event of component failures. Customers have the option to rent computing power from Runware or supply their own GPUs, with a monthly fee applicable for the pods and operational costs. Power and internet services will be billed separately, he added.
The pods can be tailored to meet customer specifications, Radulescu explained. While they feature a standard design, they can be configured to accommodate customer needs and the latest GPU chips. Given that chip delivery can be time-consuming, this aspect works to Runware’s advantage. “Our pod delivery times are also shorter than current GPU lead times, giving us time to adapt the infrastructure to the hardware being ordered,” he stated. “For new GPU generations, our in-house hardware team can design, build, and test components itself, allowing us to adapt quickly.”
At this stage, these plans remain aspirational—successful execution will be the true measure of viability. The factory will require a substantial workforce, inventory, supply chain management, and financial resources. There are cautionary examples of challenges in reinventing manufacturing, such as the electric vehicle startup Arrival, which, once valued at over $13 billion, faced difficulties in producing electric vehicles at scale and cost-effectively compared to traditional manufacturing plants. Runware, which has secured over $60 million in funding to date, plans to pursue a Series B round, but is currently focused on completing and launching the factory with the $50 million raised in its Series A last December.
“Our cost structure is very different,” Radulescu explained, detailing how Runware aims to sidestep Arrival’s pitfalls. “We do have headcount and materials costs,” he continued. “However, our factory staff is cost-effective relative to the value of deploying compute for AI.” He also noted that designing its own systems and sourcing widely available, affordable components helps maintain low costs. “We’ve brought in a highly experienced plant manager from the automotive industry to collaborate with our chief of hardware and ensure effective execution.”
Sourcing components may pose challenges. Although Runware claims to design and manufacture its own cooling and other pod systems using standard components, certain parts, such as pumps and fans, are sourced from Germany, and the company still relies on Nvidia for networking. Nevertheless, Runware asserts that it has garnered $1 billion in committed interest for its Sonic pods since they became available to external customers a month ago, “representing over 20 megawatts, which we’re now working to deliver,” Radulescu stated.
Data centers, both in the U.S. and internationally, have faced scrutiny regarding their energy consumption and its impact on the grid and other resources. When questioned, Radulescu clarified that the company’s pods require only power and an internet connection. They will utilize a closed-loop liquid cooling system with air-cooled radiators to dissipate heat, eliminating the need for a water connection. “We’re collaborating with power providers on a site pipeline covering 397 megawatts, with the potential to expand to 1.2 gigawatts,” he mentioned, referring to its list of planned sites and adding that the company is deploying its modular data centers on sites significantly smaller than those utilized by hyperscalers like Amazon, Microsoft, and Google. He further explained that these sites typically already have power, thus minimizing additional strain on the grid or disruption to the community. The pods operate “completely silent up to temperatures of 86 degrees Fahrenheit. We utilize energy from renewable sources whenever feasible,” he continued. “We are actively working on expanding our network of energy providers and sites, aiming to make renewables our primary energy source.”
Looking ahead, Runware plans to bring its new factory online early next year, followed by expansion into the U.S. and other global regions. Radulescu asserted that modular systems represent the future and will soon become “the standard for large AI deployment.” He added, “Customers should be able to plan a hyperscale data center around factory-built pods from the outset and have it operational in months rather than years.”
