A Taiwan columnist is making the case against $TSM building a joint fab with Musk — not because of wafer economics, but because of what it signals to other customers.
The setup: Musk wants a giant AI chip park in Texas (Terafab). Reports say he's circling back to it after an earlier attempt with a different partner didn't work out. The subtext: Intel was the first partner. $TSM is the better dairy farmer.
But here's the sharper edge:
$NVDA is $TSM's largest customer, has deep pockets, and still queues like everyone else. No joint fab. A Musk joint plant on advanced nodes would effectively help one buyer jump the line — which cuts against Morris Chang's rule of not competing with your own customers. $AMD sits in the same question.
The piece also mentions a closed staff meeting when the Intel Terafab report first surfaced. An employee asked about it. A senior exec answered with one line: "Can Musk be trusted?"
There's also a secondary read: SpaceX has a large post-IPO lockup expiring in November. The Terafab chatter keeps the tape hot.
The balance sheet angle: Chang's memoir talks about turning down a large Apple order once, and later filling only half, because the long-term horizon mattered more. A big Musk plant means fewer dividends, heavy depreciation, and real pain in a downturn. Not a gift to a million shareholders.
The question worth sitting with: If $NVDA still queues and has never gotten a joint fab, does a Terafab deal break the rule — or only look like it until the structure turns out to be a normal purchase order?
The setup: Musk wants a giant AI chip park in Texas (Terafab). Reports say he's circling back to it after an earlier attempt with a different partner didn't work out. The subtext: Intel was the first partner. $TSM is the better dairy farmer.
But here's the sharper edge:
$NVDA is $TSM's largest customer, has deep pockets, and still queues like everyone else. No joint fab. A Musk joint plant on advanced nodes would effectively help one buyer jump the line — which cuts against Morris Chang's rule of not competing with your own customers. $AMD sits in the same question.
The piece also mentions a closed staff meeting when the Intel Terafab report first surfaced. An employee asked about it. A senior exec answered with one line: "Can Musk be trusted?"
There's also a secondary read: SpaceX has a large post-IPO lockup expiring in November. The Terafab chatter keeps the tape hot.
The balance sheet angle: Chang's memoir talks about turning down a large Apple order once, and later filling only half, because the long-term horizon mattered more. A big Musk plant means fewer dividends, heavy depreciation, and real pain in a downturn. Not a gift to a million shareholders.
The question worth sitting with: If $NVDA still queues and has never gotten a joint fab, does a Terafab deal break the rule — or only look like it until the structure turns out to be a normal purchase order?