Rebellions.ai

Rebellions is an AI accelerator company that sells hardware and software for running AI inference in data centers. Rather than a data analysis application, it is infrastructure: chips, cards, servers and rack scale systems paired with a proprietary full stack software layer that natively supports PyTorch and vLLM. The product family is called Rebel. Rebel100 is the silicon, described as four chiplets acting as one die and designed for inference from the ground up. RebelServer is the building block for running inference efficiently at scale and is aimed at agentic AI. RebelRack is a fully integrated platform for generative AI and agentic workflows. RebelPOD is turnkey infrastructure that integrates into an existing data center without new construction. The site says systems can run fully on premises in existing air cooled data centers, and it promotes sovereign AI deployments where the customer keeps control of data, security and operations. For developers, the company provides a Model Zoo, SDK documentation, tutorials and white papers. Its news pages describe running a sovereign large language model from SK Telecom on RebelServer and a partnership to bring inference infrastructure to Japan. The vendor claims 3.8 times performance per watt and lower capital and operating costs compared with a current generation GPU node. Buyers are enterprises, AI service providers and national AI programs. Pricing is not published, and purchases go through contact with sales.