[AI device demand underpins global trade, logistics giant raises this year's commodity trade growth forecast to 4.6%]

Tesla and SpaceX CEO Musk said on Wednesday that his companies will be responsible for building and operating the Terafab AI chip complex, and TSMC (TSM.N) will at most sublease a portion of it. Terafab is a project Tesla and SpaceX plan to jointly advance in Texas, aiming to build "a giant chip factory" to produce AI chips. In response to claims that "TSMC is most likely to own and operate the factory," Musk said: "No, we will build and operate the factory. There is absolutely no question about that." He added that TSMC could sublease part of Terafab if it wishes, "but that's it." Last week, Musk said he had discussed potential cooperation with TSMC on the Terafab project but did not reveal further details.

According to the Korea Economic Daily, it has been confirmed that Samsung Electronics' next-generation high-bandwidth memory "12-layer HBM4E" product has passed quality validation by Nvidia and major hyperscale data center operators. The testing is a process of validating the product against standards required by clients. However, what has been confirmed so far is only that the product has passed quality validation, and the supply quantities for each client, contract scale, and the timing for mass production and supply cannot yet be confirmed. Industry insiders analyze that as the industry transitions to next-generation products, Samsung is expected to expand its supply opportunities ahead of schedule in the high-performance AI memory market, where it competes with SK Hynix and Micron.

According to The Wall Street Journal, citing people familiar with the matter, Broadcom (AVGO.O) has in recent weeks been seeking more than $50 billion in financing for custom AI chips co-developed with OpenAI, and institutions including Apollo and Blackstone have held discussions about participating in the financing. Negotiations are still in an early stage, and the financing size may be adjusted. The financing could cover several gigawatts of OpenAI custom chip capacity and is expected to be completed as early as the end of this year. OpenAI internally calls the chip project "Nexus," with the first- and second-generation chips codenamed Jalapeño and Serrano, respectively. OpenAI and Broadcom previously announced a partnership to develop a 10-gigawatt custom chip system, with deployment planned to run from H2 2026 through the end of 2029.