Runware, a company specializing in AI infrastructure, plans to radically change the way data centers are built by moving from a traditional construction model to a mass manufacturing model. The company aims to establish an assembly line-based factory to produce prefabricated data center pods, much like cars are manufactured in major factories.
According to Flaviu Radulescu, the company's co-founder and CEO, the plan aims to increase production capacity at the company's factory in Romania to reach 30 megawatts of data center capacity per day by 2028. This ambition implies reaching a production capacity of nearly 1 gigawatt per month, a massive figure roughly equivalent to what the entire United States is expected to add in data center capacity over an entire year.
Accelerating deployment and reducing lead times
Runware's strategy relies on units called Sonic Inference Pods, which are modular units manufactured in the factory and shipped to customers to be ready for immediate installation. The company's competitive advantage lies in speed; it aims to have the units operational online within just 48 hours of their arrival, a time that includes installation and setup, not the manufacturing time itself. The company also seeks to reduce the manufacturing time for a single unit from three weeks to one day as production pace increases.
These units are characterized by flexibility, as their design can be customized to suit customer requirements and the latest generations of graphics processing units (GPUs). The company leverages long chip supply lead times to adapt the infrastructure of the units to the required hardware, reducing technical compatibility gaps.
What changes in practice?
This approach represents a solution to several challenges facing the growing AI sector:
- Scalability speed: Instead of taking years to build massive data centers, companies can plan around ready-made manufactured units and activate them within months.
- Resource efficiency: The units address water consumption concerns by using a closed-loop liquid cooling system with air-cooled radiators, eliminating the need for external water connections.
- Sustainability: The company seeks to connect its sites to renewable energy sources to reduce the environmental impact of the data centers.
Implementation challenges and supply chains
Despite the high ambitions, Runware faces complex logistical and technical challenges. The process of reinventing manufacturing involves high risks, as happened with previous companies in the electric vehicle sector. The company relies in its supply chains on components from Germany (such as pumps and fans) and on Nvidia to provide networking technologies.
To face these challenges, the company has hired a factory manager with extensive experience from the automotive industry to ensure precision in execution. The company has already announced confirmed interest worth $1 billion for its Sonic units, reflecting the growing demand for fast and modular infrastructure solutions.