Will Musk's Terafab Shake Up the Global Semiconductor Equipment Market? UBS Estimates $135 Billion in Purchases Over 5 Years

Elon Musk, CEO of Tesla (TSLA-US) and SpaceX (SPCX-US), is actively advancing an unprecedented vertically integrated semiconductor initiative, with potential investments large enough to reshape the global semiconductor equipment (WFE) market. A new research report released by UBS on Tuesday (7th) estimates that SpaceX's 'Terafab' project will procure an amount of wafer fabrication equipment (WFE) over the next five years equivalent to the total global WFE market procurement volume for this year.

Analysts expect Terafab to begin pilot line operations as early as 2027, with equipment orders already being placed. Initial scale is estimated at around $5 billion.

The core strategic logic of Terafab mirrors Tesla's earlier in-house battery supply chain development. Musk stated at the Terafab launch event in March that the total AI computing power output from all global wafer fabs combined amounts to only about 2% of SpaceX's target demand.

He emphasized that existing suppliers, including TSMC (2330-TW), Samsung, and Micron Technology (MU-US), are expanding far too slowly to keep up with SpaceX's demand growth, stating, 'Either we build Terafab ourselves, or we won't have chips.'

In terms of product planning, Terafab will focus on two main types of chips:

First, edge inference chips for the Optimus humanoid robot;

Second, high-power chips optimized for space environments.

Musk estimates that ground computing demand will be around 100 to 200 gigawatts annually, while space computing demand could reach approximately 1 terawatt per year.

In terms of factory architecture, Terafab plans to integrate mask manufacturing, front-end logic and memory processes, advanced packaging, and testing within a single site, aiming to achieve an ultra-fast closed loop from 'design—manufacturing—testing—updating.'

UBS characterizes this as a true vertically integrated semiconductor complex.

UBS analyst John Hodulik predicts that SpaceX's AI-related capital expenditures over the next five years will total approximately $1.1 trillion, with about 20% (over $225 billion) allocated to Terafab.

Using the industry-standard conversion rate of 60% of capital expenditure translating into WFE spending, the cumulative five-year WFE procurement is estimated at around $135 billion—roughly equivalent to the total global WFE market volume for this year.

In terms of timeline, Terafab's WFE spending is expected to start at around $5 billion for the pilot line in 2027, increase to approximately $10 billion in 2028, and surge to over $50 billion annually by 2030–2031.

At that point, the global WFE market will see the sudden emergence of a new buyer with procurement volumes comparable to TSMC.

The report also notes that if Terafab progresses as scheduled, global WFE spending could approach $300 billion by 2029.

Regarding site selection, SpaceX has already submitted a tax incentive application in Grimes County, Texas. Related documents indicate an initial wafer fab investment of $55 billion, with the total investment potentially expanding to $119 billion if all planned phases are realized.

Initial capacity planning is estimated to include approximately 80,000 wafers per month of memory wafer production, along with two logic/contract manufacturing wafer fabs each producing around 20,000 wafers per month, plus integrated mask shops and back-end packaging and testing capacity.

On the technology collaboration front, the report indicates that Intel (INTC-US) is actively engaging with SpaceX, potentially in a role similar to the historical technology transfer framework between IBM and AMD. This would involve licensing process flows, manufacturing intellectual property, PDK design rules, and tool recipes to Terafab, while retaining ownership of underlying technologies and collecting licensing fees.

UBS also presents a scenario: if the pilot line is successfully validated, Intel might incorporate its 'Ohio One' facility into the Terafab ecosystem via a joint venture. This facility is large enough to support the operation of two advanced process wafer fabs.

In the memory domain, Musk has explicitly listed memory chips as a Terafab production target, though the source of related process intellectual property remains unclear. Existing memory suppliers are unlikely to be willing to license core IP to a direct competitor.

However, the report also notes that if Terafab successfully establishes scalable memory production, it could conversely push Korean manufacturers to accelerate front-end memory capacity deployment in the U.S. Samsung already owns ample land in Taylor, Texas, suitable for expansion, while Micron and SK Hynix are also watching developments.

In summary, the analysis concludes that regardless of the final form Terafab takes, the advancement of this project will positively support semiconductor equipment suppliers overall and become a central market theme for the coming financial quarters.