Altman secretly visits Hon Hai and TSMC: discusses collaboration details for Stargate and self-developed ASIC chip project

Recently, OpenAI CEO Sam Altman was reportedly making a low-profile visit to Foxconn and TSMC.

According to Taiwan Economic Daily, citing industry sources, Altman arrived in Taiwan on September 30 and held separate meetings with Foxconn and TSMC. The core topics discussed centered on OpenAI’s two strategic pillars: the "Stargate" project and finalizing details of outsourced production for its self-developed AI chip ASIC.

Neither TSMC nor Foxconn responded to inquiries regarding the meetings.

Foxconn becomes key partner for "Stargate"

One of Altman’s main objectives for his visit was to secure a stable hardware foundation for the ambitious "Stargate" project. According to reports, the project aims to build five new data centers in the US, making it one of the largest investments in computing power in the country.

Sources revealed that Foxconn is the largest AI server supplier for the "Stargate" project. Partners in the project include Oracle and SoftBank, and Foxconn happens to be a major partner for both companies in AI servers and data centers.

Therefore, Altman’s talks with Foxconn focused mainly on ensuring the production and delivery capacity of AI servers to support the extensive infrastructure demands of "Stargate".

Self-developed chip project accelerates, with focus on TSMC

Aside from servers, another key focus of Altman's visit was OpenAI's self-developed AI chip plan. According to supply chain sources, OpenAI is actively developing its own custom AI chips (ASIC) to reduce heavy reliance on external suppliers like NVIDIA.

Reports indicate OpenAI has recruited teams formerly responsible for AI chip design at Google, and plans to have TSMC manufacture the results of its chip design. The focus of discussions between the two parties is believed to be on specific details of chip production at TSMC, with the aim of achieving mass production by 2026.

Supply chain analysis suggests the initial versions of OpenAI's self-developed chips may be used mainly for inference rather than training, but successful mass production will serve as a critical test of whether tech giants’ in-house chips can replace current market solutions.

Risk Warnings and DisclaimerThe market has risks, investment must be cautious. This article does not constitute personal investment advice, nor does it take into account special investment goals, financial situations, or needs of individual users. Users should consider whether any opinions, viewpoints, or conclusions in this article fit their specific circumstances. Investment is based on own judgment and at your own risk.