Tesla CEO Elon Musk announced that the tape-out for the next-generation artificial intelligence (AI) chip, the AI5, has been completed. He also expressed his gratitude to Samsung Electronics and TSMC, which will be responsible for manufacturing the AI5 chip.
On April 15, Musk stated, “Congratulations to the Tesla AI chip design team on completing the AI5 chip tape-out.” He added, “We are developing exciting chips, including the AI6 and Dojo 3.” Taping out signifies that the chip design has been finalized and handed over to the production department. Musk told Samsung Electronics and TSMC, which will be responsible for producing the AI5 chip: “Thank you for your support in production. This chip will become one of the highest-volume AI chips in history.”
According to industry insiders, the AI5 is expected to enter mass production at TSMC’s facilities in Taiwan and Arizona, as well as at Samsung Electronics’ Taylor, Texas, plant. Full-scale mass production is projected for 2027, approximately one year after the tape-out is completed. Musk mentioned both the AI5 and AI6 that day, stating: “These chips will be used in Tesla’s products, such as self-driving cars and humanoid robots.” It is reported that the AI5 will be jointly produced by TSMC and Samsung Electronics, while the AI6 will be manufactured entirely by Samsung Electronics. Musk previously indicated that the AI6’s first silicon could be ready as early as December 2026.
Particularly noteworthy is that the chip photo released by Musk bears the inscription “KR2613.” This indicates that the chip was manufactured at Samsung Electronics’ Korean foundry during the 13th week of 2026. This suggests that Samsung Electronics’ domestic production lines in Korea played a critical role in the production of the AI chip prototype.
Collaboration among South Korean semiconductor companies within the supply chain (SCM) is also notable. The prototype shown in the newly released photo is equipped with SK Hynix memory products. It is understood that Samsung Electronics’ memory products are also included in the overall supply chain for the AI5. In areas such as high-bandwidth memory (HBM), which determines the performance of Tesla’s AI chips, the technological capabilities of both Samsung and SK Hynix appear to have played significant roles.
Samsung Electronics’ cutting-edge foundry process, “SF2T,” was applied to this AI5 chip. At last year’s Seoul Advanced Foundry Forum (SAFE 2025), Samsung Electronics unveiled its third-generation 2nm process, “SF2P+,” and announced that it was developing “SF2T”—a custom process designed for Tesla’s upcoming AI6 chip and other chips.
Analysts point out that since the SF2T process—previously thought to be reserved for next-generation vehicle models—was brought forward for use in the AI5 chip, the foundry alliance between the two companies has become even more solid.
Meanwhile, a curious incident occurred during the press conference: when Musk mentioned TSMC, he tagged an unrelated account. While thanking TSMC, Musk mistakenly directed the mention to Taiwan’s “TSC” Semiconductor Company. In fact, TSMC does not yet have its own official X account.
With the announcement that the AI5 design has been completed, Tesla’s roadmap for developing its own AI semiconductors is expected to accelerate further. Musk stated that in addition to the AI5, subsequent product lines—such as the next-generation supercomputer chip Dojo 3 and the AI6—are already in development, indicating that Tesla is committed to reducing its reliance on NVIDIA.