Partnership aims to build a vertically integrated U.S. AI chip base
Intel Chief Executive Officer (CEO) Lip-Bu Tan (left) and Tesla CEO Elon Musk shaking hands. (Photo courtesy of Intel X)
As U.S. semiconductor company Intel decides to join Tesla CEO Elon Musk’s artificial intelligence (AI) chip production project, “TerraFab,” attention is focusing on whether the competitive landscape of the global foundry (contract chip manufacturing) market will change. With AI chip volumes from global big tech companies pouring in, and Intel emerging as Tesla’s key partner as well, competition for orders in the 2-nanometer-class segment is expected to intensify.
According to the industry on April 8, Intel announced on April 7 (local time) that it will participate in the TerraFab project together with SpaceX, xAI, and Tesla. TerraFab is a mega-scale semiconductor production base pursued by Musk. It is expected to serve as an “integrated semiconductor manufacturing base” that performs AI semiconductor design, manufacturing, testing, and improvement all in one.
Recently, demand for AI chips from big tech companies has increased explosively, creating a situation in which existing foundry companies such as TSMC and Samsung Electronics alone cannot handle demand. Reservations for production capacity at TSMC, the world’s largest foundry company, are reportedly already fully booked through 2028.
Amid this, Musk’s plan is to accelerate semiconductor self-reliance by vertically integrating everything from semiconductor design to production and testing. In that case, the idea is that the project’s execution speed can be increased while reducing dependence on external companies.
In the meantime, Intel has struggled to secure customers due to low yields in advanced processes such as 2-nanometer nodes. However, by securing a mega-scale customer this time, Intel is expected to establish a foothold to expand semiconductor production capabilities within the United States. Intel is expected to manufacture Tesla’s AI chips and robot semiconductors using its 18A (1.8-nanometer) process. While Intel previously applied the 18A process mainly to its own processors, it plans to focus on securing external customers starting this year.
Attention is also on whether the foundry market structure—led by TSMC with Samsung Electronics holding second place—will see changes. TSMC and Samsung Electronics have been stepping up production of 2-nanometer-class advanced processes starting this year. TSMC has secured major big tech companies such as Nvidia, Apple, and AMD as customers. They are reportedly expected to use TSMC’s leading-edge processes to develop the latest AI chips.
Samsung Electronics will begin full-scale operation of its Taylor, U.S. plant in the second half of this year. Samsung Electronics plans to mass-produce semiconductors based on a 2-nanometer process at the Taylor plant. Tesla’s next-generation AI chip, “AI6,” will also be mass-produced at this plant.
Until now, the structure has been that Samsung Electronics takes the volume that TSMC cannot absorb, but there are observations that a market restructuring could occur. As moves spread to reorganize semiconductor supply chains around the United States, Intel could emerge as a strong alternative for U.S. big tech companies. Kim Hyung-joon, Head of the Next-Generation Intelligent Semiconductor Project Group (Professor Emeritus at Seoul National University), analyzed, “As the Trump administration advocates a semiconductor nation-first policy, the interests of Musk and Intel aligned.”
However, many point out that, due to Intel’s technical limitations so far, it will be difficult for Intel to catch up with Samsung Electronics and TSMC right away. Kim said, “Not only TSMC but also Samsung Electronics’ 2-nanometer yields have risen a lot,” adding, “Even if Intel mass-produces a 2-nanometer-class process, there is a gap of 2 to 3 years compared with TSMC and Samsung.”
Kim Eun-jin
댓글
0