Elon Musk wants Tesla and SpaceX to test 10 chip designs per year at Terafab, according to a report carried by Yahoo Finance — a volume of in-house silicon iteration that would put the two companies among the most aggressive custom-design customers in the U.S. semiconductor ecosystem.

The same report says TSMC is eyeing Texas, adding a geographic wrinkle to a story that is otherwise about design cadence. Neither claim has been confirmed by Tesla, SpaceX, TSMC or xAI, and no dollar figures, capacity numbers or timelines appear in the report.

What is Terafab?

The Terafab name ties the story to xAI's compute buildout ambitions. Musk has repeatedly framed advanced packaging and AI accelerator production as a bottleneck for training frontier models, and a dedicated fabrication or advanced-packaging facility would give Tesla, SpaceX and xAI a captive path to silicon outside the standard foundry queue.

For Tesla, the motivation is concrete. The company already designs its own FSD (Full Self-Driving) inference chips and Dojo training silicon, and a faster test loop — 10 designs a year implies roughly one new mask set every five weeks — would compress the iteration cycle between architecture ideas and validated silicon.

SpaceX designs custom silicon for its Starlink user terminals, where cost and power efficiency per unit matter at shipment volumes in the millions. Faster design-turnaround capacity would directly affect terminal economics.

Why Texas matters for TSMC

The report's second thread — TSMC reportedly looking at Texas — fits a broader pattern. TSMC's first U.S. fab in Phoenix, Arizona, has moved toward volume production on 4nm-class processes, and the chipmaker has committed to additional U.S. sites under Washington's industrial-policy push. Texas, with its existing semiconductor workforce and power infrastructure, has courted fab investment alongside Ohio, Arizona and New York.

If TSMC did land a Texas footprint, it would sit in the same state where Tesla maintains major manufacturing operations — a proximity that, in the most optimistic reading of the report, could connect Musk's test-volume ambitions with leading-edge domestic capacity.

That reading remains speculative. The report offers no confirmed investment amount, no acreage, no process node and no timeline for any Texas site.

The design-cadence context

Ten designs a year is an unusual number for a non-merchant chip company. For comparison:

• Traditional automakers iterate flagship SoCs on multi-year cycles.
• AI accelerator vendors typically tape out a small number of flagship designs annually.
• A 10-per-year test cadence implies heavy use of shuttle runs, multi-project wafers and rapid prototyping rather than full production masks.

That distinction matters commercially. Test silicon at volume requires either dedicated capacity at a foundry or an in-house facility — which is precisely what the Terafab framing suggests Musk is weighing.

What remains unconfirmed

Everything in the report sits at the roadmap-and-rumor stage:

• No confirmation from Tesla, SpaceX, xAI or Terafab-affiliated entities on the 10-designs-per-year target.
• No TSMC statement on Texas site evaluation.
• No capital-expenditure figures, incentives packages or municipal negotiations disclosed.

What is firmly established is the direction: Musk's companies continue to pull more silicon design in-house, and TSMC continues to evaluate U.S. expansion beyond Arizona. If the Texas thread hardens into a confirmed site selection, the commercial picture for domestic advanced-node capacity — and for Musk's ability to iterate custom silicon at speed — changes materially.