Musk officially announced the construction of the largest chip factory in history: the annual production target is 50 times the current global capacity, with 80% directly serving space missions.

Musk announced the launch of the ultra-large-scale chip manufacturing plan codenamed TERAFAB, with an annual production target of 1 terawatt (TW) of computing chips—about 50 times the current global annual chip production capacity, with approximately 80% of capacity directly serving space missions. The project is jointly led by Tesla, SpaceX, and xAI, located in Austin, Texas, and is the largest public manufacturing plan in human history to date.

On March 21, Musk held a press conference in downtown Austin, attended by Texas Governor Greg Abbott. Musk stated that current global annual chip production capacity is about 20 gigawatts (GW), which is only about 2% of his target demand; existing suppliers—including TSMC and Micron Technology—can no longer meet Tesla's accelerating demand in the robotics, autonomous driving, and AI fields. “Either build TERAFAB or there are no chips,” Musk said. He plans to first build an advanced factory near the Austin Giga Texas site, integrating logic chips, memory chips, and advanced packaging into one facility and equipped with photomask manufacturing equipment, forming a closed-loop for design, production, and testing.

TERAFAB is the first unified strategic project jointly announced by Tesla, SpaceX, and xAI, with the timing coinciding with SpaceX’s planned large-scale IPO this summer.

According to Bloomberg, SpaceX is expected to raise up to $50 billion, with a valuation possibly exceeding $1.75 trillion, and launching AI data center satellites into space is one of the core financing logics of this IPO. Musk did not provide a specific timeline for factory construction or capacity targets.

Analysts generally remain cautious about the engineering feasibility of the plan, pointing out that semiconductor manufacturing involves tens of billions of dollars in investment, years of construction, and highly scarce professional talent. However, some believe that TERAFAB’s strategic significance extends beyond the chip industry itself—it integrates computational manufacturing and space expansion into a unified industrial logic, directly providing industrial-level support to SpaceX’s valuation narrative.

Computing Power Gap: The Inevitable Logic of In-House Production

The core driver behind Musk’s push for TERAFAB is his extreme assessment of future computing power demand.

According to his roadmap, just the Optimus humanoid robot will consume 100 to 200 GW of chip computing power; the demand for space solar AI satellite clusters reaches the terawatt level. He expects annual production of humanoid robots will eventually reach 1 to 10 billion units—currently, global annual car production is about 100 million units, and robot production may be 10 to 100 times greater than cars.

In Musk’s view, the expansion of terrestrial computing power is approaching a physical ceiling. The total capacity of the US power grid is about 0.5 TW, unable to simultaneously support large-scale AI training, robot operation, and data center superimposed demands. Good construction locations are increasingly scarce, and NIMBY effects are increasing. Space is completely different—no day-night cycles or atmospheric attenuation, solar efficiency is more than five times that of the ground, and the larger the scale, the lower the marginal cost. Musk believes that within 2 to 3 years, the cost of deploying AI chips to space will be lower than deploying them on the ground.

Existing suppliers can no longer fill this gap. Tesla has a cooperation agreement with Samsung’s Austin factory for the next generation of chips and maintains a supply relationship with TSMC and Micron, but Musk said these suppliers’ supply growth “is far below our expectations.” He pointed out that while the semiconductor industry is expanding output as a whole, the expansion rate is still insufficient.

Factory Architecture: Closed Loop and Two Types of Chips

TERAFAB plans to complete the entire process of photomask manufacturing, chip production, packaging and testing, and design iteration within a single building, forming a high-speed recursive iteration closed loop.

Musk stated that, to his knowledge, such integration is unprecedented globally, and the iteration speed is expected to be an order of magnitude faster than existing solutions. He previously mentioned the target process being 2 nanometers.

The factory will produce two types of chips: one optimized for edge inference, mainly for Optimus robots and Tesla vehicles; the other designed specifically for the space environment, high-power chips that must withstand high-energy particle bombardment, radiation accumulation, and extreme temperatures, allowing chips to operate at higher temperatures than on Earth to reduce cooling system weight requirements. In terms of demand structure, space chips will be absolutely dominant—Musk expects terrestrial computing power to stay at 100 to 200 GW, while space will ultimately reach the terawatt level.

During the press conference, Musk also showcased a 100-kilowatt-level AI micro-satellite prototype and stated that “future satellites might reach megawatt levels.” In January this year, SpaceX applied to the FCC for permission to launch one million data center satellites into orbit.

SpaceX IPO: Financing Mainline of the Space Computing Power Strategy

The release point of TERAFAB aligns closely with SpaceX’s IPO preparations.

According to Bloomberg, SpaceX plans to complete its IPO this summer, with expected financing up to $50 billion. If successful, it will set an IPO financing record, with company valuation possibly exceeding $1.75 trillion. Deploying AI data centers in space is one of the core logics for this financing, and TERAFAB’s announcement provides specific industrial support for this logic.

SpaceX completed its acquisition of xAI in February this year, and xAI now operates as its wholly-owned subsidiary. Musk said that most chips produced by TERAFAB are expected to be consumed by xAI, mainly for space AI model training and satellite data processing.

The synergy between Tesla and xAI has already materialized on several levels: Tesla sells Megapack energy storage products to xAI, some vehicles have integrated xAI’s Grok chatbot, and in January Tesla announced a $2 billion investment in xAI and signed a framework cooperation agreement. The announcement of TERAFAB is a further upgrade of the synergy between the three companies—moving from product-level cooperation to jointly leading the same industrial project.

Three-Company Synergy: Complete Chain from Chip to Orbit

The strategic significance of TERAFAB lies in integrating the dispersed capabilities of Musk’s three companies into a complete industrial chain.

Within this framework, Tesla handles the chip demand side for Optimus robots and electric vehicles, SpaceX is responsible for launching chips and computing power infrastructure into orbit with large transport capacity, and xAI operates space AI satellite systems, consuming most of the chip output. Together, they constitute a closed loop from chip manufacturing to orbital deployment to AI computing. This is also why Musk defines the project as a “joint project of three companies” rather than a single enterprise action.

Musk positions TERAFAB as “the first step for humanity to become a solar system civilization,” and outlines a subsequent roadmap: build an electromagnetic mass launcher on the Moon, utilizing the Moon’s low gravity and lack of atmosphere to accelerate materials directly to escape velocity, increasing computing power by another thousand times to reach the petawatt (PW) level. He said he hopes to see the construction of the lunar mass launcher in his lifetime.

In the financial report meeting in January this year, Musk already stated that building TERAFAB is “to resolve a likely production capacity bottleneck in three to four years.” If the project materializes, its impact will extend beyond the semiconductor industry itself—the global computing power supply structure, orbital data center infrastructure, and even SpaceX’s deep space mission engineering foundation will all change.

Challenges: Funding, Technology, and Talent

Analysts’ reservations focus on three areas.

In terms of funding, Morgan Stanley estimates that building a factory with a monthly output of 100,000 advanced logic chip wafers costs about $45 billion; UBS’s estimate starts from $30 billion. Baird analyst Ben Kallo directly raised the market’s main concern: “Where does the money come from?” Musk has not disclosed any financing arrangements yet.

Regarding supply chain, high-end extreme ultraviolet (EUV) lithography machines are almost entirely dependent on Dutch company ASML, with delivery cycles of 1–2 years, and new customers usually wait even longer. Integrating logic chips, memory chips, and advanced packaging processes into one factory will greatly increase system complexity; building new semiconductor facilities typically costs tens of billions of dollars and often takes years to reach full production.

In terms of talent, Bernstein semiconductor analyst Stacy Rasgon stated: “Because it’s Musk, I won’t easily dismiss it, but I suspect this is actually harder than sending rockets to Mars.” He cites TSMC’s Arizona factory as an example—this project experienced years of delays, requiring engineers to be airlifted from Taiwan to the US to boost production. “These talents are not cabbages,” said Rasgon.

It’s worth noting that Musk himself has no semiconductor manufacturing background and is seen as having a history of over-promising targets and timelines—at this press conference, he gave no specific milestones for factory construction or capacity ramp-up.

Risk Warning and DisclaimerThe market carries risk, and investment requires caution. This article does not constitute personal investment advice, nor does it take into account individual users’ special investment objectives, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article fit their specific circumstances. Investment based on this is at your own risk.