TSMC Stock Record $485.80|Musk Terafab Talks or Chip Capacity Pricing Power?

A record on just discussions

TSMC's U.S.-listed shares closed at an all-time high of $485.80 after Elon Musk confirmed talks over his Terafab chip project. But Musk's own words were "Just discussions, but something may come of it." No agreement has been reached. The shares rose about 2.7% that Monday. Intel, which joined Terafab in April as its first manufacturing partner, fell 2.6%.

Several reports read the rally as a bet on what TSMC might gain from the venture's Texas factories. That reading puts the weight in the wrong place. This record is better read as a price on TSMC's power to charge for scarce chipmaking capacity. As for Terafab, its value to shareholders turns on one thing. Who pays for the factory. Start with the size of the move. The record day added about 2.7%.

Over the year so far, the U.S.-listed shares had already gained about 60%. The rest of the chip group did not follow. The iShares Semiconductor ETF slipped 0.3% during the session. That points to something specific to TSMC, and Terafab is the obvious candidate. In Taipei, though, the whole market rallied. The TAIEX index closed at a record, up 2.55%.

Crypto Briefing credited two forces: a rebound in U.S. tech stocks and continued demand tied to AI. Put together, Terafab may explain part of one day's gain. It cannot explain a 60% climb. That climb was built on something else. That something shows up in sales. Second-quarter revenue reached $40.2 billion, up 36% from a year earlier. August sales then rose 53.3% year over year.

High-performance computing, the segment AI has reshaped, made up 66% of second-quarter revenue. Management expects full-year growth slightly above 40% in dollar terms. Growth alone does not tell shareholders how much of each sale TSMC keeps. That depends on whether customers can go anywhere else.

Why customers pay more

AMD's chief executive, Lisa Su, was back in Taiwan as AMD commits more than $10 billion there. The spending is expected to run through 2029. Only part of that money goes to TSMC. Much of it goes to packaging, chip substrates and system makers. But AMD's new Venice server chip is already ramping on TSMC's 2-nanometer process. TSMC raised its 2-nanometer production target by about 20%, to roughly 120,000 wafers a month.

One report says locked-in client commitments drove that increase, not speculation. Its 5-, 4- and 3-nanometer lines ran at full capacity through the second quarter, according to TrendForce. Scarce capacity shows up in price. One report says a 2-nanometer wafer could command about $30,000. That is a 50% premium over a 3-nanometer wafer. Put another way, the older wafer sits near $20,000.

Nikkei reported that TSMC finalized price increases of 5% to 10% for 2027, for advanced and mature chips alike. The report says management describes its pricing as strategic rather than opportunistic. The test is whether customers can walk away. Qualcomm confirmed it is evaluating Samsung's 2-nanometer process for future chips. But TrendForce and Digitimes report those talks are stuck on pricing and yields.

Qualcomm's announced 2-nanometer products remain slated for TSMC. When the best alternative stalls, a price increase sticks. That is what the record is pricing.

What scarcity costs

Price increases are not pure profit. Nikkei said they help offset higher spending on materials, equipment and electricity. TSMC guides 2026 capital spending at $60 billion to $64 billion. New factories also weigh on margins before they pay off. Invezz estimates the 2-nanometer ramp dilutes gross margin by 3 to 4 points in the second half of 2026. It says overseas fabs add more.

AD HOC News made the same point: overseas fabs and new process ramps raise depreciation and startup costs. Even so, TSMC guides third-quarter gross margin at 65% to 67%. For now, pricing is covering the cost. And the overseas bill is large. TSMC has committed $265 billion to a 12-facility complex in Arizona. Reuters reported, citing two sources, that TSMC is also weighing a separate campus in Texas.

No decision has been made, and TSMC has not issued a statement. So the shareholder question sharpens. A Texas campus funded by TSMC alone would be one more cost carried before revenue arrives. Terafab matters if it changes who carries that cost.

Who pays in Texas

Here Terafab arrives with a different shape. Tim Culpan's Culpium newsletter, which first reported the talks, laid out two structures. Culpan wrote: "The most likely scenario is for TSMC to try to own and operate the new factory." SpaceX and Terafab could then invest, commit to buying a guaranteed volume of chips, or both. In the second structure, SpaceX would hold majority ownership.

TSMC would invest less but still provide process technology and operating expertise. The ownership structure remains under discussion. Culpium compared the first plan to TSMC's ventures in Japan and Germany. There, local partners provide capital and demand, while TSMC contributes technology and runs the fabs. That template has a fresh example.

Nikkei, citing Reuters, reported that Sony and TSMC plan a roughly $6.3 billion venture in Japan. Sony would own about 60%, and TSMC about 40%. Terafab's scale makes the structure matter more. Tesla and SpaceX committed $16.8 billion to the first phase. Total spending across all phases could reach as much as $119 billion, about seven times the first phase.

Culpan also reported that a Texas hub could include up to six advanced fabs, depending on federal tax incentives and guaranteed customer demand. Both blueprints hand part of the capital, or the demand risk, to Musk's side. If the talks keep that shape, Terafab is growth TSMC does not fund alone. That is where its value lies.

Why the option is small

Terafab, though, is not an ordinary customer. It aims to handle raw silicon, logic chips, memory, packaging and testing under a single 100-million-square-foot roof. Business Today notes that this vertically integrated push clashes with TSMC's pure-play model, built around fabless customers like Nvidia, Apple, AMD and Broadcom. It adds that the talks may reflect setbacks in Musk's work with Intel and Samsung.

Musk has also capped TSMC's role. He replied "true" to a post saying Terafab would be "absurdly massive" and anything TSMC contributes "supplemental." He acknowledged TSMC chips would likely add to Intel's, not replace them. Intel's 14A process is still the only manufacturing process named for the project. The partner TSMC is courting wants to do most of this itself. Hear the two sides on the same talks.

Musk: "Just discussions, but something may come of it." TSMC declined MarketWatch's request for comment. TSMC chairman C.C. Wei, as reported by Taiwan Plus, said a new wafer fab takes about two to three years to build. Expanding its output could take one to two more. He called that a "fundamental" of the foundry industry. Neither side said who pays, or how much. What Wei's timeline does say is when.

Add the two stages, and a new fab needs three to five years before it runs at scale. That narrows the claim. Terafab cannot lift TSMC's earnings for years. Only a structure where Musk's side funds a large share makes it clearly additive. Until then, the record rests on pricing.

October 15

TSMC reports before the market opens on October 15. Monthly sales are already public, and September revenue is due October 8. The quarter's sales number holds little surprise. What matters, as 24/7 Wall St. put it, is what management says about AI demand and capacity heading into 2027. Here is the judgment. The record leans on TSMC's power to price scarce capacity.

Terafab is a small and distant option, worth most if Musk's side funds the fab. So watch one thing on the call: the gross-margin outlook. The third-quarter guide is 65% to 67%, with the 2-nanometer ramp and overseas fabs already weighing on it. If the outlook holds near that range, price increases are covering the new costs, and the scarcity reading stands.

If it falls clearly below, scarcity is not turning into profit the way this record assumes. In that case, customers such as Nvidia, Apple and AMD keep more of the value from the 2-nanometer shortage, and TSMC shareholders keep less. A record set on just discussions would then rest on thinner ground.

Informational only, not investment advice. Figures and quotes come from the linked reports.