Terafab Targets 1 Million Wafers a Month From 2032

By Grace Kim

Terafab has told Japanese chip-equipment makers it wants 1 million wafers a month from 2032, Nikkei reported on Monday, citing people familiar with the matter.

Tesla and SpaceX run Terafab, a semiconductor project whose proposed mix assigns about 700,000 wafers each month to memory and more than 300,000 to logic.

The figure is a procurement target, not disclosed operating capacity for a built fab.

Tesla CFO Vaibhav Taneja put the projected cost at $20 billion to $25 billion.

What is Terafab asking Japanese suppliers to deliver?

During its procurement team's autumn visit to Japan, Terafab was placing orders with Japanese equipment makers, according to Nikkei.

The reports did not name the suppliers.

Seoul Economic Daily's account of the Nikkei report said some requested delivery deadlines fall only months away, while a leading-edge fab usually takes more than three years from planning to volume output.

That timing leaves a narrow margin before Terafab's second-half 2028 mass-production target.

Industry officials told Nikkei it "remains unclear whether mass production can be achieved as planned."

A supplier executive described the attention surrounding the project: "Right now, the biggest topic in the semiconductor industry isn't Kioxia or Rapidus. It's Elon Musk."

How large is the 1 million-wafer target?

Nikkei compared Terafab's proposed monthly output with 40 times the capacity of government-backed Rapidus.

Rapidus's plant in Chitose, Hokkaido, is sized for 25,000 wafers a month.

TSMC said its annual capacity exceeded 17 million 12-inch-equivalent wafers in 2025, or about 1.4 million a month.

Nikkei estimated that Terafab's 2032 goal equals roughly 70% of TSMC's current capacity across every node.

The proposed 700,000-wafer memory allocation would also match an estimated 650,000 to 700,000 wafers a month of Samsung Electronics' DRAM capacity.

Memory would account for about 70% of the stated allocation, making it the larger manufacturing requirement.

These comparisons describe a long-range target. They do not show that Terafab has installed equipment, qualified processes or secured the volume needed to deliver the plan.

What did Musk announce in March?

Musk announced Terafab on March 22 in Austin as a joint effort by Tesla, SpaceX and xAI.

The announced goal was to produce up to 1 terawatt of computing power each year for Tesla vehicles, Optimus robots and AI satellites.

The March plan called for an initial 100,000 wafers a month and a rise to 1 million at full capacity, without a timeline.

Nikkei said its report attaches the first date to that full-scale goal.

It also reports that Terafab's team is aiming for mass production in the second half of 2028.

Jiji Press reported that the project is leaning toward Intel's 2-nanometer or 1.4-nanometer process.

The wording signals a process preference under consideration, not a confirmed production-node commitment.

Taken together, the disclosures describe three distinct stages: an initial 100,000 wafers a month, mass production in the second half of 2028 and full capacity from 2032.

What must happen next?

Terafab must convert near-term equipment requests into installed tools, qualified production and a credible ramp before 2032.

Missing the delivery deadlines would place the second-half 2028 production goal at risk before the company tests its much larger 2032 capacity plan.

The 2032 target would put Terafab at 40 times Rapidus's stated monthly scale if delivered, but the nearer 2028 production date depends first on those tool deliveries.