Interview published November 6, 2026

In a recent interview, Elon Musk described his plans for domestic semiconductor production, large-scale data center construction, and computing in space. He also gave estimates of economic growth driven by artificial intelligence (AI) and robotics. The interviewer asked questions on each topic. Figures below are Musk's statements unless otherwise attributed.

Domestic Chip Manufacturing

Musk stated that the United States has limited chip fabrication capacity relative to Asia. He cited the following:

• TSMC is building multiple fabrication plants (fabs) in Arizona. Some are operational.
• Samsung operates fabs in Texas. It is also building or has built a fab for Tesla dedicated to the AI5 and AI6 chip designs.
• Even if all announced U.S. fabs are completed, Musk estimated their output at about 10% of Asian production, at best.
• Taiwan, Samsung, and Micron perform most of global chip production. Micron also contributes to memory production.

Musk said that geopolitical risk in Asia could disrupt supply. He also said that the rate of U.S. chip production is too slow even without such a disruption.

The Terafab Project

Musk's target is one terawatt (TW) of chip production per year. He named the project "Terafab." He said the name can be read as "terror," "terra" (Earth), or "tera" (the prefix for terawatt), and that all three readings apply.

Musk expects that the project will produce economic value comparable to the GDP of the 20th-largest country. The interviewer suggested the project could place among the top 10. Musk said that the project would fail in a dramatic way if it does not succeed. He estimated that it will succeed to some degree.

Production Efficiency

Musk identified low utilization of production time as a major inefficiency. He said that more than half of the time a wafer spends in a chip fab is spent being transported or waiting for a machine. He estimated that a useful process changes the wafer's atoms for only about 20% of its time in the fab.

He said chip manufacturers avoid process changes because fabs require large capital investment. Companies with working recipes prefer minor changes to proven methods over new approaches. Musk said his approach differs in that it accepts failed experiments in order to test more methods.

Research and Production Sites

• Research lab: Musk said a research lab under construction at Gigafactory Texas in Austin will test many fabrication approaches. Most are expected to fail. Successful methods will move to production.
• Production site: Scaled AI chip production will take place at a facility in Grimes County, Texas, near Texas A&M University.
• Integrated design and testing: Musk said the Austin lab may combine logic, memory, packaging, and lithography under one roof. This would allow engineers to design, fabricate, and test a complete chip within the same facility.
• Iteration speed: Musk estimated that a development cycle that normally takes one year could be shortened to one month.

Musk argued that technological progress depends on the number of iterations per unit of time and the progress made in each iteration. He said that an order-of-magnitude increase in iteration speed would produce a similar increase in the rate of technological advance.

Data Centers

The interviewer raised public opposition to data centers. Musk said that opposition is strongest where no data centers exist. He said that in Memphis, the company's operations have increased business activity. He stated that restaurants and hotels are full and that the company pays at times twice the prevailing wage. He also said the company has supported local schools and that tax revenue for the county and city has increased substantially.

Construction Cost and Speed

Musk compared the economics of SpaceX's data center construction with industry norms.

Musk said that industry data center builders must sign long-term leases to secure financing. He said SpaceX builds data centers without contracts because it uses them internally for its cars, robots, and rockets. He also said that competitors' completed data centers are idle because they lack power.

Power Generation

Musk said SpaceX built a power plant at its Memphis site that produces over 1 gigawatt. He said the company plans to reach 1.5 to 2 GW of internal power generation. He estimated the plant was built in about one year, compared with a typical five years.

Management Approach

Musk said he walks the Memphis site four to five times a day. He said he performs manual work himself, including pulling power cables, installing fiber optic cable, working on electrical switches, and working on power plant management software. He said this helps him understand the work and identify improvements.

The interviewer relayed a statement from a former Ford production head, who had joined a Tesla car factory in Texas under construction. The executive said that at Ford, a change required approval through six levels. At Tesla, he said, the change required only a visit to the executive above him. Musk said he has no upper floor at the Memphis site. He also said he lives in an Airstream trailer parked next to the power plant.

AI Companies and OpenAI

Musk said he provided almost all initial capital for OpenAI. He said he proposed the idea, named the company, recruited its key people, and shared his knowledge with them. He said OpenAI began as an open-source nonprofit and became a closed-source, for-profit entity. He called this the opposite of his original intent. He said he did not expect the company to become for-profit.

The interviewer stated that seven people left OpenAI and founded Anthropic. Musk said he left OpenAI after disagreements with its direction. He said he declined an ownership interest offered to him because he saw no basis for a commercial stake in a nonprofit.

SpaceX Revenue and Growth

Musk gave the following figures for SpaceX revenue:

• Last year: About $15–16 billion.
• Current year: Annualized revenue above $100 billion by December.
• Next year: Annual recurring revenue (ARR) above $500 billion by December, according to Musk's estimate. He said a trillion-dollar ARR is "not out of the question," with a probability of roughly 2%.
• Quarterly growth: Revenue increased more than 80% from Q2 to Q3 this year. Musk estimated about 70% growth from Q3 to Q4.

Musk stated that the revenue estimates are predictions and not commitments.

Musk said that bringing compute capacity online at scale is difficult. He cited a meme showing one worker digging a hole while eight people watch. He said adding purchase orders does not increase output without additional skilled labor. He said SpaceX and Tesla have the strongest physical hardware capabilities outside China. He said Tesla's Model Y is the best-selling electric vehicle in China.

Starlink

Musk said that a previous estimate projected Starlink could reach $1 trillion in revenue and$700–800 billion in EBITDA within 10 years, giving it a value of $14–15 trillion. He said SpaceX is currently valued at less than$2 trillion.

Key figures for Starlink:

• Current constellation: About 11,000 satellites.
• Version 3 satellites: Over 10 times the capability of Version 2, with 10 times as many launched. Musk estimated Starlink's bandwidth will be 100 times higher within two to three years.
• Laser links: Current links operate at about 100 Gbps. The next generation will reach 10 Tbps, which Musk described as 100 times faster.
• Version 4: Musk expects 10 times the capability of Version 3.
• Long-term constellation: Musk said the constellation could reach 10,000 to 100,000 satellites. When satellites are replaced every five years, each replacement will provide 10 times the capacity of the satellite it replaces.

Musk said agentic AI traffic already exceeds human internet traffic, citing Cloudflare. He said Cloudflare predicts AI traffic will reach 1,000 times human traffic within five years. Musk said he believes this estimate is conservative. He said agentic AI will increase Starlink's addressable market beyond the consumer, government, and enterprise segments he had previously projected.

Musk said Starlink revenue doubles roughly every year at scale. He said AI can produce annual growth on the order of magnitude.

On spectrum, Musk said SpaceX can expand its use of its current spectrum. He said that if SpaceX can use spectrum more effectively than a legacy telecommunications carrier, the company could acquire that carrier.

Computing in Space

Musk said terrestrial power supply cannot support the scale of AI compute he expects. He cited these figures:

• U.S. average power consumption: just over 500 GW.
• U.S. peak power capacity: roughly 1.2 TW.
• An annual addition of 200 GW of AI compute would increase U.S. grid demand by about 40% each year.

Musk identified several advantages of space-based solar power:

• Solar panels face the sun directly for most of the time.
• Solar panels receive about 30% more power without atmospheric attenuation.
• The system experiences no seasonal variation.
• A small battery is sufficient because the system does not need to operate through nighttime.

Test Satellites and Timeline

• A satellite carrying an NVIDIA GPU has launched.
• A satellite carrying a Google TPU has launched. Musk said both operate effectively in space.
• SpaceX expects to begin launching its first space-based AI computers, co-designed with NVIDIA, in the third quarter of next year.
• Launches will start slowly, reach significant volume in 2028, and reach very large volume in 2029.

Musk said space-based compute will be cheaper than ground-based compute once launch costs fall. He said that scaling to 1 TW per year or 200 GW per year requires space deployment.

Economic Value of Space Compute

Musk estimated the value of compute at $10 per watt. At that value, 100 GW per year of space compute would generate about$1 trillion in annual revenue. Because satellites remain in orbit, the revenue would accumulate each year. At $50 per watt, Musk said, the value would be$5 trillion per year.

Launch Costs and Starship

Musk cited the following launch cost figures:

Musk said Starship, as the first fully reusable rocket, should reduce the cost of access to orbit by a factor of 100 or more. He estimated that within two to three years, the cost of a ton to orbit will be less than the cost of air freight from California to Australia.

Reusability

• Falcon 9: Musk estimated about 80% reusability. The booster lands, and the fairing halves are recovered by parachute into the ocean and reflown. The upper stage is not recovered. Its cost is roughly that of a medium-sized jet.
• Starship: The entire system is designed for reuse. The tower arms, which Musk called "chopsticks," catch the booster and the ship. The rocket lifts off from and returns to the same launch pad. Musk said this is the fastest possible reuse cycle.
• Launch rate: Musk said the booster could fly once per hour, which could amount to 10,000 launches per year. He said the vehicle weighs over 5,000 tons at liftoff and has the highest thrust of any object made by humans.

Heat Shield

Musk said no one has built a fully reusable orbital heat shield. He cited the Space Shuttle, which required nine months of heat shield repair between flights. SpaceX has worked on the problem for about a decade. Musk described it as one of the hardest technical problems humans have attempted to solve.

SpaceX has achieved several controlled soft landings of the ship in the ocean. Musk said the ship would have been caught if a tower had been present. He said SpaceX wants high confidence in the ship's reliability before attempting a tower catch. He said the company plans to attempt a catch with Ship 42 in about one week.

Musk said the Shuttle could not iterate quickly because crew was aboard. He said Starship's uncrewed test flights allow SpaceX to take more risks and iterate faster.

China

Musk made the following statements about China:

• China's power output is about three times that of the United States, and it is expected to reach at least four times by a future date.
• China adds about 300 GW of power per year. Musk compared this with a total of about 550 GW for the United States.
• Industrial capacity is roughly proportional to electricity output, according to Musk.
• China is behind in chips, but moving quickly.
• Normally, foreign companies in China may own up to 49% of an operation. Tesla received approval for 100% ownership after several years of negotiations with Chinese officials. Musk said the government changed the law to allow it.

Musk said the only way to compete with China is to move computing and intelligence into space. He said that if the United States reaches 1 TW per year of space-based digital intelligence, its economic output would exceed China's unless China reaches the same level.

Economic Outlook

Musk gave the following estimates:

• Digital AI: Digital superintelligence could increase global economic output by 20% to 30%. Using a global GDP of about $130 trillion, this equals roughly$25 trillion. Musk said this increase will occur quickly. He estimated that digital AI could double white-collar productivity.
• Physical AI: Musk said humanoid robots could be about five times more productive than human workers because they can work 168 hours per week. He said about 120 million Americans are productive workers. Matching their output would require 20 to 30 million robots.
• Robot cost: Musk said a robot would initially cost about the same as a car, then decline over time. He cited a target of $20,000 to$30,000 per robot. He said robots would likely be rented rather than sold.

Musk said demand for AI compute will come from the AI systems themselves. He said such systems could meet all human needs, provide universal high income and medical care better than the current best, and produce goods for individuals. He said current human energy use represents a small fraction of the sun's energy, leaving significant room for growth.

Musk said that in a world with advanced AI, the current monetary system would seem outdated.

Closing Statements

Asked how he wants to be remembered at age 83, Musk said there is a difference between perception and reality. He said he would like the reality to be that he made a significant contribution to the advancement of civilization.