Musk’s AI Push Is Increasingly Tied to Samsung, Tesla Earnings Call Shows

<Tech X 110>

Tesla earnings call lays bare Elon Musk’s concerns

Samsung Foundry is a key production base for Tesla’s next-generation AI chips

Naming Samsung outright was unusual and highlights the importance of the AI supply chain

Elon Musk, chief executive officer of Tesla and SpaceX, publicly thanked chip suppliers including Samsung Electronics during Tesla’s earnings conference call. Samsung’s role in Tesla’s AI semiconductor strategy is expanding, from producing next-generation AI chips to developing follow-on products.

On Tesla’s recent second-quarter earnings conference call, Musk thanked Micron for allocating memory to the company as he outlined Tesla’s in-house AI chip plans. He also said memory prices had climbed to unusually high levels. Later, while discussing Tesla’s semiconductor roadmap, he said development was going very smoothly and again thanked TSMC, Samsung Electronics and Micron for their support.

Benzinga, a US investment news outlet, said it is rare for Musk to name specific suppliers and thank them during an earnings call. His repeated references to TSMC, Samsung Electronics and Micron showed Tesla’s growing reliance on its AI semiconductor supply chain, the outlet said.

Benzinga interpreted Musk’s remarks as a sign of difficulties in securing memory supply. It said his comments underscored the role of memory, which has received less attention than other parts of the AI supply chain. Nvidia’s graphics processing units have become a symbol of the AI boom, but advanced memory, including high-bandwidth memory, is also emerging as a critical component in AI systems.

Musk also disclosed Tesla’s plan for a chip development facility called Terafab. He said the company wants to dramatically shorten development time for custom AI chips by handling logic design, memory, lithography mask development, packaging and testing in one place.

Benzinga said that plan also hinges on securing a stable supply of sufficient memory. Because AI systems require far more high-speed memory than conventional computing, memory availability is becoming a key variable in AI competitiveness.

“Musk is focusing not only on vehicles or AI processors, but also on another core part of AI hardware that is becoming harder to secure: memory,” Benzinga said. “His thanks suggest the next bottleneck in the AI race may not be the processor itself, but the memory paired with it.”

Musk also said AI5 is slated to enter full-scale mass production around the middle of 2027 and that development is advancing quickly. AI5 is Tesla’s next-generation AI chip and is set to be installed first in Optimus, the humanoid robot the company is preparing to mass-produce. Samsung Foundry and TSMC will split production. Samsung became a key Tesla partner last year after signing a foundry supply contract with the company worth about $16.4 billion.

Samsung Foundry is preparing to mass-produce AI5 at its Taylor, Texas, plant using an advanced 2-nanometer process. AI5 recently completed tape-out, meaning the design phase has been finalized. After prototype production, performance verification and yield stabilization, mass production is projected to begin around the middle of 2027. Musk has previously described AI5 as potentially becoming “one of the highest-volume AI chips in history.”

Musk said suppliers had already made massive investments to support Optimus and robotaxis. He added that Samsung and TSMC are building semiconductor plants in Texas and Arizona, respectively, and are investing tens of billions of dollars to build the AI computing capacity needed for Optimus and robotaxis.

Tesla also plans to produce AI6, the successor to AI5, at Samsung Electronics’ Taylor plant. “I think AI6 will be the best edge-computing chip in the world,” Musk said. Kan Budhiraj, Tesla’s vice president of supply chain, said Samsung’s semiconductor plant would be used extensively for upcoming projects.

Kang Kyung-ju, Hankyung.com reporter, qurasoha@hankyung.com