Tesla's Terafab project is accelerating, with the company targeting substantial in-house chip production to support autonomous driving, robotaxis, humanoid robots, and AI infrastructure. The push is already forcing a split among its potential foundry partners, with divergent responses that could reshape supplier relationships and competitive dynamics across the semiconductor industry.
The ambition is sweeping: a pilot line processing 3,000 wafers per month by 2029, scaling toward an annual production capacity capable of delivering one terawatt of compute power — all aimed at weaning Tesla off external chip suppliers.
Samsung draws the line
Bloomberg reported that Tesla's Terafab team approached Samsung Electronics seeking help to build a semiconductor fabrication facility. Samsung declined, instead proposing that Tesla expand its use of existing foundry capacity — notably the wafer plant under construction in Taylor, Texas.
The refusal is rooted in self-preservation. Semiconductor competitiveness hinges on process capabilities such as etching, deposition, cleaning, and yield stabilization — skills cultivated over decades that foundries guard closely. Helping Tesla master those capabilities, observers note, would risk turning a major customer into a direct competitor.
Samsung's caution is also commercially rational. The Yonhap News Agency reported that the company recently secured next-generation Tesla chip orders valued at US$16.5 billion. With the Taylor plant in proximity to Tesla's Austin headquarters, Samsung has every incentive to deepen its contract manufacturing role rather than share the technology underpinning it. For Tesla, building its own fab would demand massive capital expenditure and a lengthy yield ramp-up — making established foundries the more pragmatic near-term option.
Intel bets on Tesla
Where Samsung retreated, Intel stepped in. After Samsung's refusal, Elon Musk turned to Intel, which has taken a markedly different approach. According to the Seoul Economic Daily, Intel is actively participating in the Terafab plan and recruiting talent with Samsung-related expertise — a move observers read as an effort to rebuild its technical credibility through a high-profile partnership.
It is a calculated risk. TSMC and Samsung dominate advanced process markets, and questions remain about Intel's ability to meet Tesla's performance and yield expectations, given its current capabilities in leading-edge mass production. But Intel's willingness to engage — where Samsung would not — signals a deliberate bid to leverage Terafab as a credibility accelerant and close the gap with its rivals.
A three-way power play
The contrasting responses from Samsung and Intel expose a wider tension running through the foundry sector: the tradeoff between protecting proprietary know-how and pursuing market share through bold partnerships. Samsung is playing defense, preserving its technological edge and locking in existing revenue. Intel is playing offense, willing to absorb risk for a shot at relevance.
For Tesla, the road to chip self-sufficiency remains long and capital-intensive. Entrenched technological barriers and the realities of yield ramp-up mean that sourcing from established foundries remains the more feasible path in the near term, even as Terafab development continues. How each party navigates this dynamic — Tesla pushing for independence, Samsung protecting its crown jewels, Intel chasing a comeback — will be one of the defining storylines at the intersection of automotive autonomy and advanced chip production.
Article translated by Jingyue Hsiao and edited by Jerry Chen
