According to Tweaktown, a South Korean semiconductor engineer alleged that, during negotiations with Samsung for a chip manufacturing agreement, Tesla sought extremely favorable terms: demanding access to sensitive manufacturing processes, audit rights, and pressuring for lower pricing. The engineer, who posted on the X platform under the handle @TTHUEONE, also claimed that Tesla’s Terafab project poached numerous employees from Samsung’s factory in Taylor, Texas.
Before these allegations emerged, Elon Musk had just completed a busy week. He publicly praised Micron Technology, stating that it was “rational in its pricing approach for future memory chips, unlike other manufacturers who engage in excessive pricing,” a comment widely interpreted as a veiled reference to Samsung. He also hinted that TSMC might be involved in the Terafab project—a major semiconductor initiative by Tesla and SpaceX—while ruling out the possibility of TSMC holding equity or participating in operations.
IT Home noted that claims about talent poaching are not new. A March report by Business Post highlighted that the proximity of the two facilities poses a structural risk to Samsung, noting that Terafab would compete for labor and supplier resources in the same region. At that time, Tesla was already recruiting semiconductor R&D engineers for the project.
Recent reports indicate that Samsung has dispatched advanced process yield engineers from its domestic operations to the Taylor facility to bolster staffing in preparation for the mass production of 2-nanometer chips scheduled to begin in 2027.
Regarding the core dispute over the chip manufacturing contract, Tesla's AI5 chip has already entered pilot production at Samsung's Taylor facility, with a target of mass production by 2027. The chip's design is approximately half-mask size, featuring a large central compute die surrounded by 12 memory modules, aiming for a computational power of 2,000 to 2,500 TOPS.
Musk stated that the chip's performance is approximately three times that of comparable NVIDIA chips, at only one-tenth the cost, though these figures have not been independently verified by third parties. The AI6 and AI6.5 chips are expected to be manufactured using TSMC's and Samsung's 2-nanometer processes, respectively, with significantly increased on-chip SRAM capacity and reduced memory capacity to 144GB.
Neither Tesla nor Samsung has confirmed or denied the allegations against the engineer. This incident also highlights how tensions in the semiconductor supply chain are intensifying between chip designers and foundries amid the race for AI computing power.