Investing the equivalent of this year’s total market size in the next five years? Musk’s “chip super factory” Terafab makes a massive push in semiconductor equipment

Author | Zhao Ying

    Editor | Hard AI

Elon Musk is assembling an unprecedented vertically integrated semiconductor complex, with a potential investment scale sufficient to reshape the global semiconductor equipment market landscape.

According to a UBS research report released on July 7,The wafer fabrication equipment (WFE) procurement volume for SpaceX’s Terafab project over the next five years is expected to be equivalent to the total size of this year’s global WFE market.Analysts predict that Terafab could launch pilot production lines as early as 2027, with equipment orders already placed with suppliers and an initial scale of about $5 billion.

The progress of this project means the ceiling of the global WFE market will be significantly raised. The report highlights that if Terafab advances on schedule, global WFE expenditures could approach $300 billion by 2029 and see a single customer purchasing over $50 billion annually by around 2030-2031.

Terafab: A "Full Industry Chain" Super Factory for Chips

The core logic of Terafab is highly similar to how Tesla built its own battery supply chain back in the day. During Terafab’s project release in March 2026, Musk stated that the combined AI computing output of all current global wafer fabs only meets about 2% of SpaceX’s target demand. He pointed out that current suppliers—including TSMC, Samsung, and Micron—are ramping up their production far slower than the growth in SpaceX’s demand, saying, "Either build Terafab ourselves, or there will be no chips."

In terms of product planning, Terafab will focus on two types of chips: chips for edge inference mainly used in the Optimus humanoid robot, and high-power chips optimized for space environments. Musk expects ground-based computing demand to be about 100 to 200 GW annually, while space-based computing demand could reach as high as 1 TW per year.

Regarding the factory structure, Terafab plans to bring mask making, front-end logic and memory processes, advanced packaging, and testing entirely under one roof, enabling an ultra-fast closed loop of "design—manufacturing—testing—iteration." The report identifies it as a truly vertically integrated semiconductor complex.

WFE Expenditure Estimate: $135 Billion in Five Years, Over $50 Billion at Peak Year

Analyst John Hodulik predicts thatSpaceX’s AI business (excluding satellite internet and rocket launches) will have total capital expenditures of around $1.1 trillion over the coming five years, of which about 20%, or approximately $225 billion, will go to Terafab. With about 60% of industry capital expenditure typically converted to WFE, that means $135 billion in WFE procurement over five years—roughly equivalent to the current entire global WFE market.

In terms of timing, Terafab’s WFE expenditure is expected to start at around $5 billion for the pilot production line in 2027, rising to about $10 billion in 2028 (already included in the forecast of about $250 billion for global WFE that year), and then jumping to over $50 billion annually by 2030-31—adding a new WFE buyer of a scale similar to TSMC to the market.

On site selection, SpaceX has submitted a tax abatement application in Grimes County, Texas. Relevant documents show initial semiconductor fab investment of $55 billion, and if all planned phases are realized, total investment may reach $119 billion. Based on this, initial Terafab capacity may include about 80,000/month memory wafer capacity and two logic/contract fabs of about 20,000/month each, with supporting mask workshops and advanced packaging/testing lines.

Intel May Be "Technology Knowledge Holder", Memory Suppliers Face Potential Impact

On the technical cooperation side, the report notes Intel is engaging with SpaceX about Terafab. Its role may be similar to the previous technology transfer framework between IBM and AMD—that is, through technology licensing, to provide Terafab with process flows, manufacturing IP, PDK design rules, and tool formulas, while retaining core technology ownership and collecting licensing or patent fees. UBS also considers another possibility: if the pilot line is successful, Intel may incorporate its "Ohio One" campus into Terafab as a joint venture. This site is large enough to support two advanced process fabs. Furthermore,if Terafab ultimately cannot succeed independently, the possibility of SpaceX further investing in Intel will increase.

For memory, Musk clearly included memory chips as one of Terafab’s production targets during the launch, but the source of memory IP is still unclear, and current suppliers have scant incentive to license core IP to potential competitors. However, the report also points out that if Terafab successfully establishes large-scale memory production, it could conversely force Korean firms to accelerate the buildout of US-based memory capacity—Samsung already has ample land for expansion in Taylor, Texas, while Micron and SK hynix are also paying close attention.

In summary, regardless of what form Terafab ultimately takes, progress on this project as a whole is a positive boost for semiconductor equipment suppliers and is likely to become a major market theme for earnings seasons to come.