Tesla and Nvidia fund Samsung to secure their AI chips

For their AI chips, Tesla and Nvidia also need lesser-known components: substrates. The two groups reportedly plan to invest nearly $3 billion in Samsung's production capacity, in exchange for long-term supply.

Designing ever more powerful chips is good. Being able to put them together is even better. According to the South Korean daily Maeil Business Newspaper cited by Reuters, Tesla and Nvidia would finance the expansion of Samsung Electro-Mechanics factories to the tune of 3.7 trillion won, or approximately $2.8 billion.

These investments would concern FC-BGA substrates, which are supports ensuring the electrical connections between a processor and the electronic card which hosts it. Their name does not make you dream, but their availability is as important as that of the other components necessary for assembling the chips.

Tesla would contribute around $830 million for a dedicated line in the Sejong factory. The substrates would be intended for the AI ​​chips of its autonomous vehicles and its humanoid robot Optimus. In exchange, the manufacturer would benefit from priority access to production for five years, from 2029 to 2033. This contribution would be included in the investment program of 4,270 billion won planned for the expansion of the site: it would therefore not be added to it.

The calendar still leaves a little time before the first deliveries. Construction should be completed at the end of 2027, then installation of equipment and testing would take up the first half of 2028. Commercial production would begin in the second half, before deliveries to Tesla in 2029, according to information from the Korean daily.

Nvidia, for its part, would spend around $1.9 billion for the expansion of the Sejong and Busan facilities. The partnership was officially announced on October 21. The deal provides lines dedicated to substrates for AI processors and data center networking chips. Supplies are to begin in 2026 as part of a multi-year commitment.

Why finance your supplier directly? Because AI processors impose increasing demands: more connections, high consumption and heat that complicates manufacturing. There are few suppliers capable of producing these high-end substrates; Ibiden and Unimicron are also among them. For Samsung Electro-Mechanics, these agreements make it possible to reduce the cost of new installations while reserving volumes for major customers. For Tesla and Nvidia, the interest is just as concrete: having the necessary components on hand when their chips are ready.

Machine-translated from French