Three months after TSMC told investors its Arizona buildout would total 12 fabrication and advanced packaging facilities plus an R&D center for $265 billion, the company is reportedly shopping for a second American site in Texas. Before anyone celebrates a bigger American footprint, it's worth asking the question this newsletter asks every time TSMC opens its mouth: is this a funded plan, or is it a hedge dressed up as expansion?
What's Actually Been Said
According to Reuters, two sources familiar with the matter say TSMC is "evaluating" a potential Texas investment — language doing a lot of work here. The plans "have not been finalised," and the company would not confirm anything to Reuters directly. CFO Wendell Hu is quoted mid-sentence in the available reporting, but the operative fact is this: Taiwanese media reported the story first, citing unnamed industry sources, and Reuters is confirming the existence of deliberations, not a decision.
That matters because I've tracked this company's Arizona number inflate from $12 billion to $65 billion to $165 billion to, as of July, $265 billion — each jump accompanied by the same kind of "additional four or more fabs will be built" language CEO C.C. Wei used again in the Reuters report. A pledge that keeps growing before the prior pledge is built is not evidence of momentum. It's evidence that announcing capacity is cheaper than pouring it.
Trade coverage out of Taiwan adds texture without adding certainty. Digitimes frames the potential Texas campus as running to "tens of billions of dollars" and possibly serving as a second major U.S. hub alongside Arizona, with Singapore reportedly also in talks — though the site's own reporting sits behind a paywall and doesn't specify fab count or timeline beyond that framing. A separate Digitimes item puts a ceiling of "up to 6 advanced wafer fabs" on the Texas idea, attributed to unnamed sources, which should be read as an upper bound floated in the rumor mill, not a blueprint.
Samsung's Taylor Fab Shows What Real Progress Looks Like
The contrast worth sitting with is Samsung, which is not talking about a hypothetical third site — it's describing specific milestones at a site it already committed to years ago. Per Seoul Economic Daily, Samsung plans to begin initial operations at its first Taylor, Texas fab late this year, and break ground on a second Taylor fab within the year, targeting mass production there in 2030. Samsung's foundry VP Noh Mi-jung also laid out a process roadmap with dates attached: SF2P (its third-generation 2nm node) in 2028, SF1.4 in 2029, and a goal of lifting HPC's share of foundry revenue from 28% today to 66% by 2029.
None of that is a guarantee Samsung hits its numbers — this publication has flagged plenty of "on track" language from foundries that didn't survive contact with yield reality. But there's a structural difference between a company naming a construction date for a facility it's already building and a company "evaluating" a facility it hasn't sited. Samsung's Taylor fab has been under construction long enough to have an "initial operations" date attached to it. TSMC's Texas idea doesn't have a site, a dollar figure sourced to the company, or a fab count TSMC itself has confirmed.
The Accountability Gap
Here's the pattern this newsletter keeps finding: capacity announcements travel faster than capacity itself. TSMC's Arizona pledge went from $12 billion to $265 billion across several years of earnings calls, and Reuters' own tally shows the current plan already covers 12 facilities and an R&D center — a very large unbuilt backlog before a single additional Texas shovel goes in the ground. Floating a second U.S. region while the first one is still mid-construction reads less like aggressive expansion and more like a company managing its political optics with Washington and its customers simultaneously, at minimal present-day cost, since "evaluating" commits nobody to anything.
I'd bookmark this one specifically for follow-up: if TSMC's Q4 or Q1 earnings call attaches a dollar figure, a fab count, and a site to the Texas idea, that's a real commitment to track against. If it stays at "evaluating" through two more earnings cycles, treat it as what it looks like today — a hedge against Arizona delays, a pressure-release valve for the price-hike conversations TSMC is already having with customers, and a headline, not a construction schedule. Samsung's Taylor dates are the ones with a calendar attached. TSMC's Texas story, so far, has a rumor and a maybe.