Elon Musk has ruled out Taiwan Semiconductor Manufacturing Co. (TSMC) as a partner in his Texas-based "Terafab" semiconductor project, according to a Taipei Times report citing the executive's recent public remarks.

The decision narrows the field of potential foundry partners for a facility whose name suggests terawatt- or tera-scale manufacturing ambitions, though the company has not disclosed process nodes, capacity targets, capital expenditure estimates or production timelines for the project in publicly available materials.

What is the Terafab project?

The "Terafab" name surfaced in Musk's public commentary as shorthand for a vertically integrated fabrication complex tied to Tesla and xAI's custom silicon needs. Musk has described the planned Texas site as a multi-product facility combining AI training chips, power devices and advanced packaging. The project aligns with a pattern at Musk-led companies of pulling strategic components in-house, a strategy previously applied to battery cells at Tesla and rocket engines at SpaceX.

Why the TSMC exclusion matters

TSMC commands roughly 90% of the world's advanced-node foundry market, a position built on sustained process leadership from 28nm down to current 3nm production and a roadmap extending to A16 (1.6nm-class) by 2027-2028. The Taiwanese foundry manufactures Tesla's in-house AI training silicon on advanced process nodes. A split between Musk and TSMC would force a sourcing decision that has so far been implicit rather than explicit.

Industry analysts view the exclusion as a signal of Musk's preference for captive or near-captive production rather than shared foundry capacity. Tesla has invested heavily in chip design talent, and bringing wafer fabrication closer to that design work would shorten iteration cycles.

Texas as a semiconductor hub

The Terafab decision lands in a state already absorbing tens of billions of dollars in chip-industry capital expenditure. Samsung Electronics is building a multi-module fab complex in Taylor, Texas, with 4nm production targeted at the first module and 2nm GAA production planned at subsequent phases. Intel operates fabs in Austin and has converted its Rio Rancho, New Mexico site for advanced packaging, though that work falls outside Texas. Texas Instruments is expanding 300mm analog production in Richardson and Sherman.

Texas offers power-grid capacity, state-level incentives under the CHIPS and Science Act framework, and proximity to Tesla's Austin headquarters and Gigafactory Texas. The combination has made the state a focal point for fab investment over the past three years.

Who might fill the gap

With TSMC excluded, three foundry options remain plausible. Samsung Foundry brings the closest geographic fit through its Taylor complex, plus a 2nm GAA roadmap. Intel Foundry Services offers leading-edge process nodes through its 18A node, though its track record with external customers remains limited. GlobalFoundries focuses on mature and specialty nodes that do not align with AI training silicon requirements.

Musk has not named a replacement partner publicly, and the Taipei Times report does not identify any alternative foundry in his statements.

What remains undisclosed

No confirmed timeline, capex figure, process node or wafer-output target exists for the Terafab in public reporting. Equipment supplier selections, permitting milestones and workforce plans also remain opaque. Whether the project reaches groundbreaking within the next 12 to 24 months will depend on Tesla and xAI capital allocation decisions still pending in 2025-2026 corporate guidance.

For the foundry sector, the Terafab exclusion underscores a recurring dynamic: leading-edge capacity remains tight, but captive silicon ambitions at automakers and AI startups are reshaping the customer mix. Whether Musk's Texas bet becomes a third major US logic fab cluster or a long-tail experiment will become clearer as equipment orders, hiring announcements and CHIPS Act funding applications surface in the quarters ahead.