Taiwan Semiconductor Manufacturing Co. is weighing a new multibillion-dollar chip campus in Texas, according to a Bloomberg report published October 1, 2026, that cited people familiar with the matter. The move would mark TSMC’s second major US manufacturing region, stacked on top of an Arizona build-out that has already grown from a $12 billion pledge in 2020 to a stated $265 billion commitment as of July 2026. Nothing about Texas is finalized. But the fact that the world’s largest contract chipmaker is even studying a second American mega-site shows how tight AI chip capacity has become, and how much leverage Nvidia and Apple now hold over where their silicon gets made.
What Bloomberg’s Report Actually Says About TSMC’s Texas Plan
Bloomberg’s October 1 story, sourced to people familiar with TSMC’s internal planning, describes a proposal for multiple fabrication plants on a single Texas site, not a single standalone fab. The report frames the dollar figure as multibillion, with people briefed on early discussions suggesting the eventual price tag could run into the tens of billions once land, utilities, and equipment are counted. TSMC has not confirmed the plan publicly, and Bloomberg’s sourcing makes clear the project remains under internal review rather than approved by TSMC’s board.
That caveat matters. TSMC has a pattern of studying sites for years before breaking ground, and Arizona itself took multiple rounds of expansion before reaching its current scale in the hardware industry’s biggest reshoring bet. A Texas campus, if it happens, would not open this decade. What the report signals instead is demand: TSMC’s existing capacity, even after five years of expansion, is not enough to keep up with orders for the chips that power Nvidia’s GPUs and Apple’s processors.
How Arizona Grew From $12 Billion to $265 Billion
TSMC’s Arizona commitment did not arrive as one number. It climbed in five distinct steps over six years, each announced separately as demand outpaced the prior plan. The timeline below reconstructs those increments from TSMC’s own announcements, White House statements, and Bloomberg’s reporting.
| Date | Announcement | Increment | Cumulative total |
|---|---|---|---|
| May 14, 2020 | TSMC’s original Arizona fab announcement | $12 billion | $12 billion |
| December 6, 2022 | Second Arizona fab added | +$28 billion | $40 billion |
| April 2024 | Third Arizona fab confirmed | +$25 billion | $65 billion+ |
| March 3-4, 2025 | White House and TSMC jointly announce expansion | +$100 billion | $165 billion |
| July 15-16, 2026 | Bloomberg and NIST confirm new US buildout deal | +$100 billion | $265 billion |
Each jump tracked a surge in orders for advanced logic chips rather than a single grand plan drawn up in 2020. The $65 billion mark covered three Phoenix fabs for leading-edge wafers. The March 2025 jump to $165 billion, announced alongside the White House, added three more fabs plus two advanced packaging plants and an R&D center. The July 2026 increase to $265 billion followed the same pattern: more orders, more promised capacity, no finished buildings to show for most of it yet. A Texas campus would be a sixth increment on top of a number that already doubled twice in eighteen months.
Nvidia and Apple Are the Customers Driving the Math
TSMC does not design chips. It manufactures them for customers, and two of those customers, Nvidia and Apple, are large enough that their roadmaps effectively set TSMC’s capacity planning. Nvidia’s data center GPUs and Apple’s iPhone and Mac processors both depend on TSMC’s most advanced nodes, and both companies have spent 2026 locking down supply years in advance. Nvidia’s own buyback and spending decisions this year, covered in our look at Nvidia’s $150 billion buyback, reflect a company flush with cash and racing to secure wafers before rivals do. Apple briefly overtook Nvidia atop the market-cap rankings earlier in 2026 before slipping back, a swing detailed in our report on how Apple’s Nvidia crown lasted two months, not the year some analysts expected.
Both companies need TSMC’s leading-edge nodes specifically, not just any fab capacity. That narrows TSMC’s options. Building generic chip plants would not solve Nvidia’s or Apple’s bottleneck. A Texas site, like Arizona, would need to run the same advanced process technology Taiwan uses, which costs far more per fab than legacy chip production and takes longer to bring online.
Why a Single Customer Can Reshape a Foundry’s Map
Foundries rarely build ahead of demand at this scale. TSMC’s expansion has instead followed its two biggest customers’ growth curves almost in lockstep. When Nvidia’s data center order book grows, TSMC’s committed capacity grows within a year or two. That dependency cuts both ways: it gives TSMC predictable demand, but it also means a slowdown at either customer could leave a half-built Texas campus without enough orders to fill it.
TSMC vs. Samsung Foundry vs. Intel: Who’s Actually Building in the US
TSMC is not the only foundry chasing US capacity, but it is far ahead of its two closest rivals on both committed dollars and construction progress. Samsung’s Taylor, Texas campus and Intel’s Ohio and Arizona projects were all announced with fanfare, and both have run into delays that TSMC has largely avoided.
| Company | US site | Committed investment | Status as of October 2026 |
|---|---|---|---|
| TSMC | Phoenix, Arizona (plus proposed Texas site) | $265 billion | First fab producing advanced chips; later fabs under construction, Texas still unconfirmed |
| Samsung Foundry | Taylor, Texas | $17 billion for the fab itself; over $37 billion for the wider regional plan | Two leading-edge logic fabs plus R&D, targeting 2026 operations |
| Intel | New Albany, Ohio | Up to $28 billion planned | Delayed; production now generally expected around 2030-2031 |
| Intel | Chandler, Arizona (Fab 52 and Fab 62) | More than $32 billion | Fab 52 in leading-edge production; Fab 62 part of broader expansion |
Samsung’s Taylor Bet Hasn’t Closed the Gap
Samsung’s Taylor campus was supposed to put the company closer to TSMC’s heels in advanced US manufacturing. Instead, Samsung Foundry has struggled to win the same volume of leading-edge orders from Nvidia and Apple that TSMC commands, leaving Taylor’s two planned fabs running well below the order book TSMC enjoys in Phoenix. Samsung’s broader $37 billion-plus regional figure covers R&D and an Austin expansion alongside Taylor, not just the new fab itself.
The Memory Squeeze Sitting Behind TSMC’s Decision
TSMC’s capacity crunch is not happening in isolation. It is unfolding alongside the worst memory-chip shortage the AI hardware market has seen, and the two problems feed each other. Micron said on September 30 that it could not predict when supply would catch up with demand, according to DIGITIMES, and the company expects conditions to tighten further through 2027 and 2028. More than 75% of Micron’s 2027 output was already committed to customers as of October 1, 2026, 247 Wall St. reported, with a large share of that tied to high-bandwidth memory for AI accelerators.
SK Hynix’s 2026 memory capacity was described as essentially sold out for the year, Network World reported, and Gartner forecast a 47% increase in DRAM prices during 2026 on the back of that undersupply. Counterpoint Research separately forecast roughly 30% further price growth through the end of 2026, potentially followed by another 20% increase in early 2027, per TBS News. Our earlier coverage of how Micron posted a record $54.2 billion quarter while tightening supply, and of Nvidia’s Rubin Ultra losing a third of its memory to the HBM shortage, both trace back to the same root cause: manufacturers are redirecting advanced wafer capacity toward HBM at the expense of everything else.
Logic chips and memory chips are made in different kinds of fabs, so TSMC’s Texas plan would not directly add DRAM or HBM supply. But both shortages share a cause: AI accelerator demand has outrun the entire industry’s capacity to add leading-edge manufacturing fast enough, whether that’s Micron’s HBM lines or TSMC’s logic wafers. Our look at how wafer prices could surge 40% in 2027 on AI demand covers the raw-material side of the same squeeze.
How the CHIPS Act Fits Into a Second TSMC Site
The US CHIPS Act remains active in 2026, though its money has moved slower than its headlines. The Government Accountability Office reported on August 7, 2026, that the Commerce Department had awarded roughly $31.7 billion in direct funding to 24 companies across 49 projects, but had disbursed only about $13.1 billion of that as of April 2026. The Congressional Research Service noted in July 2026 that twelve companies holding preliminary funding memoranda still had not received final awards.
Samsung’s own CHIPS award, finalized December 20, 2024, came to $4.7 billion, which the CRS calculated as roughly 12.8% of the company’s expected Taylor project investment. Commerce was still issuing new commitments as recently as July 2026, when it announced letters of intent worth $874 million across seven companies, though those had not converted into final awards. Any Texas incentive package TSMC negotiates would likely draw on a mix of leftover CHIPS funding, state-level tax breaks, and the kind of direct federal deal-making that produced the March 2025 and July 2026 Arizona increases.
Why Texas, Specifically
Texas already hosts Samsung’s Taylor campus and a dense cluster of chip-adjacent suppliers, utility infrastructure, and skilled labor built up over the past five years. For TSMC, planting a second site near existing semiconductor infrastructure cuts risk compared with picking an entirely new state. It also diversifies TSMC’s US footprint away from a single region, reducing the odds that one state’s permitting delays, water shortages, or power grid constraints could stall its entire American expansion at once. Arizona has faced all three of those pressures during its own build-out, and spreading future capacity across two states is a hedge any manufacturer this size would consider.
Market Impact: What This Means for Chip Prices and Availability
A Texas announcement, if it comes, would not change chip prices or availability this year or next. Fabs take years to permit, build, and qualify for production, and TSMC’s own Arizona timeline shows second and third fabs still ramping years after groundbreaking. What a Texas plan does signal to markets is that TSMC expects demand to keep climbing well past its current $265 billion Arizona commitment, which investors read as a vote of confidence in sustained AI chip orders from Nvidia, Apple, and other leading-edge customers.
For buyers already squeezed by the DRAM and HBM shortage covered above, the more relevant near-term signal is that logic chip capacity, not just memory, is now a constraint companies are planning around years in advance. Our coverage of DRAM now costing more than TSMC’s own 2nm silicon captures how unusual it is for memory, historically a commodity, to outprice TSMC’s most advanced logic process on a per-unit basis. A second TSMC mega-site would eventually ease the logic side of that equation, but not before the end of this decade at the earliest.
Historical Context: TSMC’s Reluctant Pivot to Building Abroad
For most of its history, TSMC kept its most advanced manufacturing in Taiwan, treating geographic concentration as a strategic asset rather than a risk. That changed gradually under pressure from US trade policy, customer demands for supply-chain diversification, and direct federal incentives starting with the original 2020 Arizona announcement. Each subsequent expansion came with public assurances from TSMC executives that Taiwan would remain the company’s most advanced node, even as the dollar figures committed to Arizona climbed past $65 billion, then $165 billion, then $265 billion in under three years.
A Texas campus would extend that pivot further than any single company in the industry has gone, turning what began as a single fab into a two-state American manufacturing network. It would also test whether TSMC’s model, built around Taiwan’s dense cluster of suppliers and engineers, can be replicated a second time on US soil at a cost and speed investors will tolerate.
What Analysts and Investors Are Watching Next
Three things will determine whether the Texas plan becomes real. First, whether TSMC’s board formally approves a site before the end of 2026 or pushes the decision into 2027. Second, whether Nvidia and Apple sign long-term capacity commitments specific to a Texas facility, the kind of customer backing that turned Arizona’s early fabs from a political gesture into a fully booked operation. Third, whether federal and Texas state incentives match what Arizona received, since TSMC has shown it negotiates hard on subsidies before committing new capital.
Analysts will also watch TSMC’s next quarterly earnings call for any direct comment, since the company has historically confirmed major US expansions within one or two quarters of initial press reports rather than letting speculation run for long.
Five Predictions for TSMC’s Next Move
- TSMC will likely confirm or deny the Texas plan within two to three quarters, following its usual pattern of resolving press speculation quickly once a report gains traction.
- If confirmed, the initial public dollar figure will probably be presented as a standalone commitment rather than folded into the existing $265 billion Arizona total, mirroring how each prior increment was announced separately.
- Nvidia and Apple will remain the implicit justification for any new site, even if neither company is named directly in TSMC’s own announcement.
- Texas and federal officials will likely offer incentive packages comparable to or exceeding what Samsung received in Taylor, given the competition between states for high-profile chip projects.
- Actual production from a Texas site, if approved, will not begin before the early-to-mid 2030s, based on the multi-year lag seen at every prior TSMC Arizona fab.
What It Means for Buyers of AI Hardware Right Now
For companies buying GPUs, AI accelerators, or high-end consumer chips today, the Texas report changes nothing in the short term. The constraint buyers are actually running into this quarter is memory, not logic capacity, and that squeeze is driven by Micron, Samsung, and SK Hynix’s HBM allocation decisions rather than anything happening at TSMC. What the Texas story does confirm is that the companies closest to the AI buildout, TSMC chief among them, are planning for demand to keep growing well into the next decade, not plateau. That is a signal worth tracking for anyone budgeting multi-year hardware procurement, even if it has no effect on this quarter’s price list.
Frequently Asked Questions
Has TSMC officially confirmed a Texas fab campus?
No. Bloomberg’s October 1, 2026 report, based on people familiar with internal discussions, described the plan as under consideration. TSMC had not issued a public confirmation at the time of this report.
How much has TSMC committed to Arizona in total?
TSMC’s stated US investment reached $265 billion as of a July 2026 announcement covered by Bloomberg and the US National Institute of Standards and Technology, up from $165 billion announced in March 2025 and $65 billion before that.
Would a Texas site replace or add to the Arizona investment?
Based on Bloomberg’s reporting, Texas would be a separate, additional project rather than a replacement for Arizona. The $265 billion figure applies specifically to the existing Arizona program.
Why does TSMC’s expansion matter for Nvidia and Apple customers?
Both companies depend on TSMC’s most advanced manufacturing nodes for their chips. Any delay or shortfall in TSMC’s capacity plans can directly affect how many GPUs or processors Nvidia and Apple can ship, which is why their order volumes effectively drive TSMC’s US expansion decisions.
Is the memory chip shortage related to TSMC’s Texas plan?
Not directly. Memory chips like DRAM and HBM are made by companies such as Micron, Samsung, and SK Hynix in different fabs than TSMC’s logic chips. Both shortages share the same underlying cause, which is AI accelerator demand outpacing the industry’s ability to add leading-edge capacity fast enough.
How does TSMC’s US investment compare with Samsung and Intel?
TSMC’s $265 billion committed total is well ahead of Samsung Foundry’s roughly $37 billion regional Texas plan and Intel’s combined Ohio and Arizona commitments of roughly $60 billion. TSMC has also made faster production progress than either rival at its US sites.
When would a Texas fab actually start producing chips?
No timeline has been confirmed, since the project has not been approved. Based on how long TSMC’s Arizona fabs took from announcement to production, a Texas site would likely need five or more years before shipping chips.
What happened to CHIPS Act funding in 2026?
The program remained active but only partly disbursed. The Government Accountability Office reported in August 2026 that Commerce had awarded about $31.7 billion to 24 companies across 49 projects, with only about $13.1 billion actually paid out as of April 2026.
