Elon Musk used a company event in Austin to describe a rhythm for the chipmaking project he has been assembling in Texas. Tesla and SpaceX will aim to make and test about 10 chip designs a year at the Terafab research fab at Giga Texas, he said on October 11, against the roughly two years a significant chip normally takes to travel from a blank sheet to volume production.
He was explicit that the goal is iteration, not output. The ten designs would be manufactured and tested, not pushed into mass production. “We want to do more like 10 designs per year,” he said, describing a faster loop rather than ten commercial products a year.
The research fab, as Musk describes it, would gather fabrication, testing, packaging and photomask production under one roof. The value he claims lies in the loop: test a chip, cut new masks, build another version, and repeat without shipping wafers across an ocean or waiting in another company’s queue. “You’re making a complete computer, testing it, then making new masks and then a new chip, and having that iterative loop be super fast in the same building,” he said.
“I think what we want to do with Tesla and SpaceX on the chip fab is really try some of these high-risk, high-payoff bets,” Musk said. He acknowledged the failure rate that approach implies. “Some of these, when you’re trying a bunch of different things, a bunch of them won’t work out.” He added a note about his own expertise — “I have no idea how to make chips, but we’ll figure it out” — and a timeline measured in years. “It might take us quite a few years, but over time I’m pretty sure we could do it.”
Tesla intends to start with conventional tools, modify them to raise throughput, then build machines of its own around the steps that limit output. That is a plan to buy into an industry and then rebuild it, a strategy with few precedents in a business where process knowledge is guarded closely and depreciated over decades.
Speed in design has value because everything downstream of it is slow. A modern chip needs a set of photomasks that costs millions of dollars and takes weeks to produce, so a team that revises a design pays for every change in money and calendar time. Companies plan around those costs by designing less often and validating more thoroughly before committing, which is why roadmaps are measured in years and why a start-up that misses a window often misses a generation of products.
The gap between a tested design and a manufacturable one is where most of the money sits. A prototype that runs correctly in a laboratory can fail on a production line because of defects, variability or heat, and closing that distance takes repeated runs through the same equipment. Musk’s claim is that housing every step in one building shortens each cycle; the industry’s answer is that the cycles are slow for physical reasons that a shorter walk between departments does not remove.
Terafab is a larger undertaking than the research fab. Musk has described it as two factories, each making one chip design, supplying the company’s cars and Optimus robots and, eventually, data centers in orbit. SpaceX proposed an initial $55 billion investment in a Texas semiconductor plant in a filing made public in May. Intel joined the effort in April, and Tesla has said it plans to use Intel’s 14A manufacturing process.
Taiwan Semiconductor Manufacturing Co. remains in the picture in a supporting role. Musk confirmed talks were under way — “just discussions, but something may come of it” — before saying Tesla and SpaceX would build and run Terafab themselves while allowing TSMC to sublease part of the site. Operational control would stay with the two companies. The discussions have a history. In December 2024, TSMC chairman C.C. Wei told Musk, “As long as you’re willing to pay, your chips will definitely be there.” Reporting has traced more serious talks with Texas officials to the first half of 2026, as TSMC weighed whether it needed another American fab and an anchor customer to justify it.
Analysts said the design target will be judged against yields, not intentions. A research fab that produces working prototypes but cannot reach commercial yield rates would leave Tesla dependent on outside foundries for the chips its vehicles and robots need, which is the position it is trying to escape.
The ambition runs against the economics of the industry Musk is entering. Leading-edge fabrication consumes capital in the tens of billions before a single wafer ships, employs process engineers that number in the thousands, and rewards scale above everything else. Tesla has advantages other entrants lacked: a captive customer, a balance sheet, and a willingness to experiment in public. It also has no track record in a discipline where the learning curve is measured in decades and the incumbents are still climbing it.
Musk framed the effort as a bet on iteration speed, and said the company would accept a high rate of failure along the way. Whether ten designs a year is a cadence or a slogan will be settled at the same place every chipmaker is judged: by what comes off the line, at what yield, and at what cost.
