AI infrastructure management software "AI-Stack"

In fact, AI-Stack has attractions that go beyond "GPU resource partitioning." It can automatically allocate a resource pool composed of single or multiple GPUs to individuals or teams based on policies. GPU resources can be set for each team, including minimum and maximum usage amounts. For example, if GPU resources are completely unused, one team can monopolize their use. When other teams start using them, they are automatically allocated based on policies, thereby increasing the utilization rate of valuable GPU resources.

INFINITIX, a company specializing in GPU orchestration and AI infrastructure management software, has completed a new funding round of 225 million Taiwanese dollars. This raises the company's valuation to over 1 billion Taiwanese dollars, further accelerating its growth stage towards an IPO. This investment round saw participation from several strategic investors, including HH-CTBC Capital, a joint fund of Foxconn and CTBC Financial HD, Fubon Financial Venture Capital, Cathay Venture, and JU-SHAN Capital. The funds raised will primarily be used to expand the research and development framework, deepen product technology, and strengthen the development of overseas markets, thereby boosting INFINITIX's global competitiveness and market share expansion.

On November 13, INFINITIX, in collaboration with related companies such as SignalPro Technology, Supermicro, Macnica, Stark Technology, and Hetone, launched a project to build the next-generation "SiGTRON AI-FACTORY" around the Southern Taiwan Science Park. The project aims to strengthen AI infrastructure in the region as part of the smart computing ecosystem promoted by Neocloud (Next-generation Cloud). The AI-FACTORY will utilize 128 SMCI HGX B200 server nodes and 24 SMCI RTX Pro6000 server nodes, achieving a total computing performance of 40.96 ExaFLOPS. INFINITIX's independently developed AI-Stack GPU resource management software will serve as the operational core, handling GPU partitioning, integration, cross-node computing, and flexible scheduling, thereby creating an environment that allows multiple users to utilize resources efficiently and optimizing real-time inference.