Rebellions Partners With Japan’s ai& to Supply More Than 100 RebelRack Units

Rebellions announced on the 16th that it has signed a partnership with Japan AI infrastructure company ai&. Through the collaboration, Rebellions plans to target Japan’s local AI inference infrastructure market. ai& is a vertically integrated global AI technology company.

The partnership is part of ai&’s AI inference service business in Japan. At the core of the collaboration is deploying Rebellions’ RebelRack on ai&’s heterogeneous computing infrastructure.

As a result, Japanese companies, government agencies and developers will be able to use Rebellions-based AI inference infrastructure built locally in Japan. Rebellions plans to expand local customers’ computing resource options and support Japan’s sovereign AI infrastructure strategy.

ai& has begun taking delivery of the initial batch and plans to deploy Rebellions infrastructure at its Tokyo data center. It also plans to expand RebelRack deployments to more than 100 units. The initial buildout scale could be expanded further depending on ai&’s overall data center expansion plan.

ai& has secured more than USD 2 billion in infrastructure investment funding and has set goals of operating 5 sites by the end of 2026 and securing 40MW of infrastructure capacity by the end of 2027.

For such infrastructure, Rebellions will combine an inference-optimized architecture with the heterogeneous infrastructure model operated by ai&. ai&’s infrastructure operation method combines different computing resources depending on the characteristics of each AI workload, and the purpose of applying Rebellions NPU is to improve power efficiency and reduce operating costs.

With the application of Rebellions NPU, flexible inference resource allocation will be supported, and ai& expects this will enable it to build a range of AI services and provide low-cost inference services. Rebellions said reducing inference costs is a key factor in broadening the scope of AI application and service use.

Rebellions hardware can also be easily integrated into ai&’s existing development environment. ai&’s technical team uses open-source software widely used in GPU environments, and Rebellions also supports major open-source frameworks.

This will allow ai& to introduce Rebellions infrastructure while using its existing development environment, without a separate large-scale system transition. As a result, ai& can bring in Rebellions infrastructure while maintaining its existing development environment, reducing the burden of migration.

Rebellions expects the first system could begin operating on the day it is delivered. Industry observers interpreted the adoption as part of the shift from the experimental stage of AI to the commercial stage. Matt Eastwood, IDC senior vice president of research, said the impact of inference economics, which encompasses performance, power efficiency and cost, on infrastructure selection is growing. He said ai&’s commercial-scale adoption of Rebellions products demonstrates that purpose-built AI semiconductors are emerging as a practical alternative for inference infrastructure.

David Bennett, cofounder and CEO of ai&, said the significance of heterogeneous infrastructure lies in using the optimal hardware for each workload. He said Rebellions is a partner that fits that model. He also said Rebellions provides high efficiency and allows service cost structures to be designed with flexibility.

Park Sung-hyun, CEO of Rebellions, said the economics of inference are a key factor that determines the spread and use of AI. He added that through this partnership, the company has secured direct contact with local customers in Japan and can now expand its business foundation in Japan’s AI infrastructure market.

Source: TECHWORLD · Park Gyu-chan
Original: https://www.epnc.co.kr/news/articleView.html?idxno=406987

References

• https://www.epnc.co.kr/news/articleView.html?idxno=406987