TAIPEI (Taiwan News) — Japan’s SoftBank is in talks with Nvidia and Foxconn to develop and manufacture domestically produced AI servers, Nikkei Asia reported Friday.

The company’s mobile unit plans to focus on high-performance AI servers capable of running advanced GPU systems at high speed. The initiative is expected to be included in SoftBank’s medium-term business strategy, which could be unveiled as early as next week.

SoftBank has begun discussions with Nvidia on GPU server certification and with Foxconn on manufacturing cooperation. Foxconn is one of the world’s largest AI server manufacturers, while Nvidia dominates the global market for AI chips.

SoftBank aims to begin assembling externally sourced server components by the end of the decade before expanding into full in-house production. The company is also considering building a manufacturing base at a former Sharp plant in Sakai, Osaka, acquired last year.

The servers are expected to support SoftBank’s own data centers and its AI-RAN initiative, which combines mobile communications infrastructure with AI computing platforms. The company also plans to design and develop non-semiconductor components while leveraging outside expertise.

Demand for sovereign AI systems developed and operated domestically to reduce data security risks has been rising as AI adoption expands across governments and industries, per Nikkei Asia. Concerns over supply chain security and potential backdoors in AI servers have further accelerated interest in locally controlled infrastructure.

The global AI server market is projected to reach NT$16.45 trillion (US$524 billion) by 2030, according to ABI Research. Japanese companies have limited presence in the sector outside of firms such as Fujitsu and NEC, while global manufacturing remains dominated by major players, including the US’s Dell and China’s Inspur.