October 3: Musk "Announces" the Talks on X

On October 3, Elon Musk confirmed on X that he is in discussions with TSMC about chip manufacturing cooperation for Terafab. That makes TSMC the second top-tier foundry at the table — Intel had been the only officially named partner, slated to supply the 14A (roughly 1.4nm-class) process. TSMC has not commented.

Terafab is the mega-scale, vertically integrated chip project pushed by Tesla, SpaceX and xAI. Its goal is not merchant foundry work, but a secured, massive supply of custom silicon for FSD, Optimus, orbital data centers and beyond.

1TW a Year: Roughly 50x the World's Annual AI Compute Output

Scale is the project's most striking footnote. By Musk's earlier "global compute of about 20GW" math, Terafab's planned capacity sits at roughly 50 times the world's current annual AI compute output — no single wafer fab on Earth has ever been designed to that magnitude.

Terafab at a Glance

· Potential long-term total investment: about $119 billion; planned floor area over 100 million sq ft

· Target capacity: about 1TW of chips per year, roughly 50x current global annual AI compute output

· Site: Grimes County, Texas; mass construction from December 2026; phase one targeted for end-2028

· Intel 14A: risk production around 2027, volume production targeted near 2028

Scale also dictates that Terafab cannot bet on a single supplier. Intel's 14A is still ramping toward volume production on a timeline tightly coupled with Terafab's end-2028 phase one — while TSMC happens to be expanding into Texas itself, a coincidence observers call a "two-way rush".

Markets Vote: Intel Slips, TSMC Hits a Record High

As the news spread, Quartz reported Intel shares under pressure, while Yahoo Finance reported TSMC stock at an all-time high. Media sketch several possible structures: an equity joint venture (the JASM/ESMC playbook from Japan and Germany), managed operation, or technology transfer — even a hybrid of "Intel 14A front-end plus TSMC back-end".

What Musk wants was never to pick a winner, but to keep chip choice in his own hands: control the core links in-house, and pull in whoever is strongest for the rest.

Takeaway: Compute Certainty Is Becoming the Hard Currency of AI Strategy

For the AI industry, Terafab's scale proves one thing: when the demand curves for models, robots and space-based compute are steep enough, supply-chain certainty itself becomes core competitiveness. Enterprises planning their own AI infrastructure likewise need multi-source redundancy across performance, resilience and cost — the starting point for NineZenith's Zenith-Control (Tianshu) platform in optimizing compute and agent resource scheduling.

(Compiled from public reports by QbitAI, Quartz and Yahoo Finance)