Tesla-Samsung Clash Over Chip Prices and Talent Poaching

According to recent reports, Tesla CEO Elon Musk has praised Micron for offering reasonably priced memory products while criticizing other memory makers for charging excessive prices, remarks widely interpreted as targeting South Korea's Samsung Electronics. The comments have sparked a public dispute, with a Samsung semiconductor engineer accusing Musk and his companies of applying exploitative business practices, including aggressive foundry price negotiations and recruiting semiconductor talent from Samsung's Texas facility.

Musk's Memory Pricing Comments Trigger Dispute With Samsung

The controversy began when Musk criticized rising memory prices on X and revealed plans to invite Taiwan Semiconductor Manufacturing Co. (TSMC) to participate in Terafab, a chip manufacturing initiative jointly planned by Tesla and SpaceX. TSMC subsequently confirmed that it was in discussions with SpaceX and stated during an earnings briefing that it had no intention of giving up business opportunities to competitors. However, Musk has reportedly ruled out the possibility of TSMC taking an ownership stake in Terafab or participating in its operations.

In response, a Samsung semiconductor engineer based in South Korea publicly criticized Tesla's negotiations over its AI5 and AI6 chips. The engineer alleged that Tesla had demanded substantial foundry price reductions, access to internal semiconductor fabrication information, and acceptance of factory inspections. The engineer also claimed that Terafab, led by Musk, had recruited a significant number of local semiconductor professionals from Samsung's chip facility in Taylor, Texas.

Despite the public dispute, Tesla continues to maintain foundry relationships with both Samsung and TSMC. The two companies are expected to play roles in manufacturing Tesla's next-generation AI5 and AI6 processors.

Tesla AI5 Chip Targets High Computing Performance

Tesla's AI5 processor is expected to use 2-nanometer manufacturing technology from TSMC and Samsung's Taylor facility. Its projected computing performance ranges from 2,000 to 2,500 trillion operations per second (TOPS). According to Musk, the AI5 chip could deliver approximately three times the performance of comparable NVIDIA processors at just 10% of the cost.

A single-chip configuration is expected to offer performance comparable to NVIDIA's H100 GPU, while a dual-chip configuration could compete with NVIDIA's Blackwell platform. These performance comparisons and cost estimates are claims attributed to Musk and have not been independently verified.

The AI5 chip is reportedly undergoing trial production at Samsung's Taylor, Texas, facility, with mass production targeted for 2027. The chip is designed at approximately half-reticle size, with a central computing die surrounded by 12 DRAM modules. Its memory capacity has reportedly been adjusted to 96 GB.

The subsequent AI6 and AI6.5 processors are also expected to be manufactured by TSMC and Samsung. These designs are anticipated to incorporate substantial amounts of SRAM to improve memory bandwidth, while their DRAM capacity is expected to be reduced by one-third to 144 GB.

Terafab Project Aims to Accelerate AI Chip Manufacturing

At the center of the dispute is Terafab, a proposed semiconductor manufacturing project involving Tesla and SpaceX, with technical support from Intel. The project reportedly plans to invest $16.8 billion in a semiconductor campus covering approximately 100 million square feet in Grimes County, Texas.

Terafab's long-term objective is to produce AI computing capacity equivalent to 1 terawatt, or 1,000 gigawatts, annually. The initiative aims to integrate multiple stages of semiconductor manufacturing within a single facility, including mask fabrication, lithography, advanced chiplet packaging, and testing.

By consolidating these processes, Terafab seeks to reduce the chip design and development cycle from the current six to nine months to just a few weeks. If successful, the project could reshape how AI processors are designed and manufactured, while increasing demand for advanced semiconductor fabrication, high-bandwidth memory, SRAM, DRAM, and advanced packaging technologies.

As competition for AI computing capacity intensifies, the dispute highlights the growing tensions between chip designers and foundries over manufacturing costs, intellectual property, access to technical information, and semiconductor talent. The outcome of the Terafab initiative and Tesla's ongoing foundry partnerships could have significant implications for the next generation of AI hardware.