WHERE FORTRESS AMERICA LANDS
The Southern Spine and the Race to Build It
WHERE FORTRESS AMERICA LANDS | Blue Collar Analytics | June 2026 (Revised)
WHERE FORTRESS AMERICA LANDS
The Southern Spine and the Race to Build It
Author: Adam Wood | Publication: Blue Collar Analytics | Series: Fortress America, Paper III | Date: June 2026 | Version: 1.1
Revision note (v1.1): This version corrects a citation error in Section 4 regarding TSMC’s Arizona water reclamation system. The original version cited figures (10,000 cubic meters per day scaling to 36,000 cubic meters per day by 2026) that describe TSMC’s Taiwan facilities, not its Arizona operations. This version also substantially rebuilds Section 4 and the paper’s conclusion. Every public water-recycling claim made by an operator in this buildout — TSMC, Samsung’s water partner EPCOR, Intel, the city of Sherman, GlobalWafers, Texas Instruments, and the water-engineering firm Gradiant — was checked individually against primary sourcing. One of these, TSMC’s, held up as an honest precedent, voluntarily-adopted standard against which the others can fairly be measured; the rest describe something narrower, softer, or less verifiable than their original framing suggested. Section 4 is rebuilt around that finding, and the paper’s conclusion is revised to match. Substantive conclusions elsewhere in the paper are unchanged.
EXECUTIVE SUMMARY
The Fortress America framework established in Papers I and II predicted that hemispheric consolidation would produce a domestic infrastructure buildout traceable in real time through capital flows, regulatory actions, and physical deployment. Paper III maps where that buildout is physically landing — a California-Texas-Arizona southern spine emerging as the industrial core of the American economy for the next generation. The rare earth supply chain from mine to magnet to motor is now domestically traceable for the first time. More than a dozen semiconductor fabs are at some stage of production or active construction across this corridor. The binding constraint on the entire buildout is water in an arid region, and public claims about the engineering response to that constraint vary widely in how much scrutiny they can withstand — from one operator’s precedented, honestly-stated target to several others whose claims describe less than their framing implies.
SECTION 1 THE SOUTHERN SPINE
Arizona is the most concentrated single-state semiconductor investment in American history. Since 2020, the Arizona Commerce Authority has documented more than 60 semiconductor expansions representing more than $205 billion in investment — a figure Governor Katie Hobbs cited directly at SEMICON West 2025, the semiconductor industry’s own annual trade conference, held in Phoenix for the first time. [1] By early 2025, Industrial Info Resources tracked nearly $50 billion in active industrial construction across the state, with semiconductor fabs, data centers, power plants, and mining operations leading the pipeline. [2]
Texas is building a different but complementary concentration — multiple companies across a contiguous corridor rather than one dominant anchor. Samsung’s Taylor fab on 1,200 acres is the largest foreign direct investment in Texas on record, targeting operational status at the end of 2026 with 1,500 permanent employees and production of 2-nanometer leading-edge chips. [3] In July 2025, Samsung filed a formal regulatory disclosure on the Seoul Stock Exchange confirming a $16.5 billion semiconductor supply agreement running through December 31, 2033; Tesla CEO Elon Musk confirmed the same day that Samsung’s Taylor facility would produce Tesla’s next-generation AI6 chip, used across Tesla’s AI products including autonomous vehicles and the Optimus humanoid robot. [4] The robotics manufacturer and the semiconductor fab are already contractually linked in the same state through the end of the decade.
Both states are absorbing pressure from different directions, and the pattern reveals which part of the buildout is durable. Semiconductor fabs faced real delays in 2023-2024 — Taiwan Semiconductor Manufacturing Company’s (TSMC) first Arizona fab slipped roughly a year due to skilled-labor shortages, and its second fab moved from a 2026 target to 2027-2028 — but both are now active, with TSMC reporting pilot production ahead of its most recent revised schedule as of April 2026. [5] Data centers in the same state have faced a different kind of setback entirely: data center opposition nationwide blocked or delayed at least 75 projects worth roughly $130 billion in the first quarter of 2026 alone — a single-quarter total matching all of 2025 combined, and more than double the $64 billion blocked cumulatively between 2023 and early 2025. Arizona has been part of this wave throughout, where Chandler’s city council unanimously rejected a proposed AI data center in December 2025 and a $14 billion West Valley project was withdrawn after the city declined to approve rezoning. [6] The fabs slipped on labor and equipment timelines and recovered. The data centers are being stopped by the communities they are built in.
The southern spine is the operational core of the domestic semiconductor buildout today, with Arizona in high-volume production and Texas ramping to production in 2026-2027. New York and Ohio represent a second wave of the same institutional pattern — committed capital, steel in the ground, state and federal funding locked in — coming online in the 2030-2032 timeframe. [7] The fab-by-fab production status confirming this geography as the industrial core of Fortress America is detailed in Section 3.
SECTION 2 MINE TO MAGNET TO MOTOR: THE DOMESTIC RARE EARTH SUPPLY CHAIN
Semiconductor capacity is the precondition for advanced manufacturing; without the chip, a robot has no intelligence to actuate in the first place. But intelligence without actuation is just as inert — a robot that can think but cannot move is a research demonstration, not a defense asset. The rare earth supply chain is the second non-negotiable bottleneck, and for the first time, it is now traceable end to end on American soil.
Rare earth elements are not a single-application input. They are foundational across the entire advanced manufacturing stack: permanent magnets containing neodymium, praseodymium, dysprosium, terbium, and samarium power the precision actuators inside semiconductor fabrication equipment itself, optical crystals doped with yttrium and lanthanum enable the lasers used in advanced lithography, and the same magnet chemistry that moves a wafer with micron-level precision inside a fab is what moves a robotic joint with the torque and control a humanoid platform requires. [8] The supply chain risk is not confined to one product category. China accounts for over 90% of global rare earth production broadly, and approximately 85 to 90% of neodymium-iron-boron (NdFeB) magnet production specifically — concentration that touches the fabs documented in Section 1 as much as the robots documented in Paper II. [9]
The chain that resolves this dependency, for the first time, exists entirely within U.S. borders. Mountain Pass, California, operated by MP Materials, is the only operational rare earth mining and processing facility of scale in the United States — the extraction point. Independence, MP Materials’ magnet manufacturing facility in Fort Worth, Texas, is already producing at an initial 1,000 metric ton annual capacity and expanding — the conversion point, where processed rare earth oxide becomes finished magnet. [10] A robot built in Texas with American rare earth magnets processed from California ore is a fundamentally different strategic asset than one assembled with components Beijing can restrict on short notice.
The federal government’s commitment to this chain is structured with more depth than a typical subsidy. In July 2025, the U.S. Department of Defense (DoD) entered a multibillion-dollar agreement with MP Materials that made DoD the company’s largest shareholder, acquiring approximately 15% of outstanding shares through a $400 million purchase of newly created preferred stock, convertible at $30.03 per share. [11] The package extends well beyond equity: a $150 million loan to expand heavy rare earth separation capacity at Mountain Pass; a 10-year price floor of $110 per kilogram for MP’s neodymium-praseodymium (NdPr) products, insulating the company from Chinese price manipulation; and a 10-year offtake agreement under which DoD guarantees that 100% of magnet output from a new “10X” production facility — targeting 10,000 metric tons of annual capacity by 2028 — will be purchased by defense and commercial customers. [12] MP secured an additional $1 billion in commercial financing from JPMorgan Chase and Goldman Sachs on the strength of the federal commitment. As a binding condition of the agreement, MP Materials ceased all sales of rare earth products to China in July 2025 — a cessation already reflected in the company’s own reported revenue — and committed not to renew its existing offtake agreement with China’s Shenghe Resources at its January 2026 expiration, ending the one remaining contractual link between America’s primary rare earth producer and a Chinese state-affiliated buyer. [13]
This is not an isolated transaction. It is the first and most fully developed example of a repeating federal instrument: the government taking direct equity positions in companies that control the extraction and early-stage processing of strategic minerals, rather than in the factories that turn those minerals into finished goods. The same Department of Defense office that structured the MP Materials deal has since taken a 10% stake, with warrants for an additional 7.5%, in Trilogy Metals to advance copper and cobalt development in Alaska, and a separate equity position in Lithium Americas. [14] What this pattern signals about how the government is choosing to secure the upstream end of the supply chain, rather than only the downstream manufacturers, will be explored in full in the final paper of this series.
The one legislative risk to this chain is the Section 45X Advanced Manufacturing Production Credit, which subsidizes domestic production of critical minerals and components. [15] For most 45X-eligible products, the credit phases down starting in 2030 and expires entirely in 2033. Critical minerals, however, sit on a separate track within the same credit: the phaseout for critical minerals does not begin until 2031, with full expiration in 2033 — and unlike solar and wind components, critical mineral production was carved out from the general sunset schedule precisely because of its strategic designation. [16] Even within that more favorable schedule, non-renewal in 2033 would still represent a real long-term risk to the economics of expanding domestic magnet production at the pace the DoD offtake agreement assumes.
But the DoD offtake agreement itself is what makes that risk politically difficult to realize. The Pentagon is now MP Materials’ largest shareholder, has committed to a 10-year price floor, and has guaranteed purchase of 100% of a facility’s output through approximately 2035 — two years past the 45X expiration date. A Congress that allowed the critical minerals credit to lapse without renewal would be undermining an investment the Department of Defense itself structured, sized, and continues to hold equity in. Attacking the credit’s renewal becomes, in practice, attacking the defense department’s own balance sheet and its own stated rare earth independence strategy — a substantially higher political bar than allowing a generic manufacturing subsidy to expire on schedule.
The mine-to-magnet-to-motor chain described here is not a projection. Mountain Pass is producing today. Independence is producing today. The DoD’s equity, loan, and offtake commitments are signed, public, and filed with the U.S. Securities and Exchange Commission (SEC). [17] What remains is scale — the 10X facility in Northlake reaching its 10,000-metric-ton target by 2028, and the broader expansion of domestic magnet capacity beyond MP Materials alone, an effort already extending into Wyoming and Alaska on separate tracks. The chain that supplies the humanoid robotics manufacturing imperative identified in Paper II now exists, end to end, without crossing Chinese territory at any point.
SECTION 3 THE SEMICONDUCTOR SPINE
The strategic vulnerability this buildout addresses is stark and well documented. The United States declined from producing roughly 37 to 40% of global semiconductors in 1990 to approximately 10 to 12% today, and as of the passage of the CHIPS and Science Act in 2022, none of the world’s most advanced chips were manufactured domestically. [18] The correction underway since is not a slow policy response — it is one of the fastest industrial buildouts in American history, and its pace is best understood by naming what is actually under construction, what is already producing, and what remains delayed.
TSMC’s Arizona campus is the clearest evidence of acceleration. Fab 21’s first facility entered high-volume 4-nanometer production in Q4 2024 and is now manufacturing chips for Apple and Nvidia — the first time TSMC has produced cutting-edge AI silicon outside Taiwan. [19] The second fab’s construction was completed in 2025, with equipment installation beginning in the third quarter of 2026 and 3-nanometer production targeted for 2027 — a full year ahead of the original schedule, according to TSMC’s own CEO. [20] A third fab broke ground in April 2025, targeting 2-nanometer and A16 process technology by the end of the decade. [21] Total committed investment has grown from an initial $12 billion in 2020 to $165 billion across six planned fabs, with TSMC describing long-term framework expansion reaching approximately $465 billion — the largest single foreign direct investment in American history. [22]
Texas Instruments offers the clearest example of speed. Construction began on its Sherman, Texas site in mid-2022; the first fab, SM1, began production on December 17, 2025 — three and a half years from groundbreaking to output. [23] The facility produces foundational analog and embedded processing chips, the components used in nearly every electronic device, including industrial robotics and automotive systems. [24] SM2’s exterior shell is already complete, with cleanroom installation underway in 2026. The Sherman campus is part of a $60 billion investment across seven planned fabs in Texas and Utah. [25]
Samsung’s Taylor, Texas fab presents a more complicated picture, and an honest accounting requires saying so. Originally targeted for 2024 production, the facility was delayed twice — once to upgrade its process technology from 4-nanometer to the more advanced 2-nanometer node, and again reportedly due to insufficient near-term customer demand. [26] As of May 2026, Samsung’s own foundry leadership confirmed that customer production, including for Tesla, is scheduled to begin in 2027, with the facility’s third-generation 2-nanometer process now in installation. [27] The delay is a demand and technology-upgrade story, not a capital withdrawal — Samsung’s total investment in the site grew from $17 billion to $44 billion over the same period, and the company received $4.75 billion in direct CHIPS Act funding to proceed. [28]
Intel’s Chandler, Arizona campus adds two more fabs to the spine. Fab 52 is in high-volume production on Intel’s 18A process — the first U.S. facility to cross the 2-nanometer threshold, with Intel’s own chief technology officer describing it as capable of more than 10,000 18A wafer starts per week — while Fab 62 is under construction and expected to be ready around 2028. [29] Intel’s commitment to Arizona has held even as the company cancelled comparable projects in Germany and Poland, consolidating its advanced manufacturing investment domestically rather than abroad. [30]
Taken together, the southern spine’s fab buildout shows a pattern consistent with the rest of this paper’s findings: delays have been common, but they have been delays of one to two years driven by labor availability, technology upgrades, and customer demand — not cancellations, and not capital withdrawal. Every fab named above is either producing today or under active construction with a committed completion date. That distinguishes this buildout sharply from the data center cancellation wave described in Section 1, where the setbacks have come from local political rejection rather than engineering or market timing.
SECTION 4 THE WATER CONSTRAINT AND THE ENGINEERING RESPONSE
The binding constraint on the southern spine is water, in a region already under acute and well-documented stress. Arizona’s water supply depends heavily on the Colorado River, which has experienced sustained shortage conditions for over two decades, and on groundwater aquifers facing long-term depletion as the state’s population and industrial base both grow. [31] Texas faces a parallel and independently documented problem: the state’s own water planning authorities project a long-term supply deficit of nearly 6.9 million acre-feet, with officials warning that without expanded infrastructure, demand could outstrip supply during the next prolonged drought. [32] Into both constrained systems, the fab buildout documented in Section 3 introduces a new and significant industrial water demand layered on top of an existing shortage.
The scale of that demand is best understood by direct comparison. Estimates of a single fab’s daily water draw vary by facility size and process node — independent technical sources put the range anywhere from roughly 3 million to 10 million gallons per day, with the largest individual facilities and multi-fab campuses running higher still. [33] The World Economic Forum’s commonly cited industry figure, 10 million gallons per day, sits at the top of that range rather than describing a typical fab. A single large data center, by contrast, consumes roughly 1.5 to 5 million gallons per day. [34] Even using the low end of the fab range, a fab remains the larger water consumer — and at the high end, cited most often in public discussion, the gap widens to a factor of two to four.
One operator’s public commitments stand apart from the rest of this section, and it is worth establishing why before turning to the others. TSMC states plainly, in its own public materials, that its Arizona fabs “aim to achieve” a 90% water recycling rate, pursued through a purpose-built Industrial Reclamation Water Plant (IRWP) with a stated “design goal” of near-zero liquid discharge. [35] The IRWP broke ground in August 2025 and will not be operational until 2028. TSMC’s own language has always reflected that timeline honestly — an aim, a goal, a plant under construction. Their aims and designs have been clearly stated and appear to model the existing capabilities already demonstrated in their Taiwanese operations, where the company maintains recycling rates exceeding 85%. [36] That standard is also the harder and more expensive of the two paths available to a fab operator: recycling the core ultrapure water that touches the wafer, rather than the comparatively simpler work of recycling water used in cooling towers and air scrubbers. Nothing required TSMC to choose the harder path. The City of Phoenix’s own public position is that it has sufficient water capacity for TSMC regardless of whether the 90% target is ever reached, and the figure does not appear among the enforceable milestones attached to TSMC’s $6.6 billion CHIPS Act award. [37] TSMC set this target voluntarily, stated it in public, and did so in a community where water has been, by local reporting, “the flash point in many neighborhood meetings and planning sessions” — in front of the audience most likely to notice if the company fell short. [38] Whether TSMC’s Arizona plant reaches 90% recycling on the 2028 timeline the company has set for it is a separate, later question, and this paper takes no position on it in advance. What can be said now is that TSMC’s public description of its own progress has not, at any point, claimed more than the company has actually built.
Measured against that standard — public claims that describe no more than what has actually been achieved or actually been committed to, and that point to demonstrated precedent where precedent is claimed — the other claims made across this buildout hold up less consistently.
Samsung’s water partner in Taylor, EPCOR, announced a target in a July 2023 press release: reclaiming or reusing 75% of the Blue Sky Water Reclamation Facility’s process water. [39] Unlike TSMC’s figure, this one has not been paired with an operational timeline, an engineering description, or any subsequent public update. A local outlet that sought detail on how the target would be met, roughly six weeks after the announcement, received no response beyond the same press-release language, and none has surfaced publicly in the nearly three years since. The claim may be accurate. It has not been made possible to check.
Intel describes its Ronler Acres, Oregon operations as having reached “net positive” water status as of 2022 — returning more treated water to the local watershed than the company withdraws. [40] This claim is true, and it is a genuine achievement. It is also, by industry reporting on the underlying mechanism, an “end-of-pipe” result: water is treated and returned to secondary systems and to the watershed, rather than recycled back into the core ultrapure loop that touches the wafer — the same category of water TSMC’s existing Arizona system currently handles at a 65% rate, achieved more completely. “Net positive” is accurate. It answers a different question than the one this section is asking.
In Sherman, Texas, three separate things get discussed together and are worth pulling apart. The city’s Post Oak Wastewater Treatment Plant is real, built, and operating — but its 16 million gallon per day permitted capacity is citywide infrastructure, serving “domestic, commercial and industrial wastewater” for the whole community, not a figure specific to either fab. [41] Its stated purpose is to treat wastewater to a standard “suitable and safe for reintroduction back into the natural stream environment” — it returns water to a creek, not to TI’s or GlobalWafers’ own production lines. Separately, GlobalWafers carries a water commitment NIST itself lists as a condition of its CHIPS Act award: recycling “at least 50% of the process water used onsite” within a year of completing its project, industry convention suggesting “process water” refers to the core production loop rather than secondary systems, though the government’s own award language does not spell out that distinction explicitly. [42] That commitment is qualified by “commercially reasonable efforts,” a standard that does not require the 50% figure actually be reached. TI has separately stated, in its own press materials rather than in NIST’s listed award conditions, that it is “endeavoring to achieve a 70% water reuse capability” across Sherman and its Lehi, Utah site — a company statement with no equivalent government-tracked commitment behind it. [43] TI’s own most recent disclosed company-wide water reuse figures — 27% in 2020, 29% in 2023 — show little movement over that period and sit far below the 70% target, with no demonstrated site achieving anything close to that rate publicly disclosed as of this writing. [44] No public reporting yet confirms whether either company’s Sherman-specific water target has been met.
Independent water-engineering firms serving the semiconductor sector, including Gradiant, advertise recycling capabilities of up to 99% for individual clients, at investments running into the hundreds of millions of dollars. [45] The company has not disclosed the client or site behind its highest-profile version of this claim, describing it only as “one of the world’s largest semiconductor manufacturers.” No public source connects this specific claim to any facility in the southern spine.
None of this means the companies making these claims are being dishonest. It means that, measured against the plainest and most verifiable standard available in this section — a company that stated a hard goal, described it honestly, pointed to demonstrated precedent, and made that statement in public to the community most likely to hold it accountable — most of the other claims in this buildout describe something narrower, softer, less precedented, or less verifiable than their framing suggests: a true result answering a different question, a target that has gone unconfirmed for years, a commitment qualified by its own contract language and unsupported by any comparable track record, or a capability whose relevance to this region has not been established. The water constraint documented at the start of this section has not been resolved by any of it. TSMC’s Arizona fabs have been drawing water and producing chips since late 2024. The plant designed to substantially close that gap will not exist until 2028 — and of everyone whose public claims this section has examined, TSMC is the only one who has said so plainly, in advance, without being asked.
CONCLUSION
Across this paper and the series it belongs to, the forces shaping the buildout of Fortress America are moving with a speed that is itself a finding. Section 1 documented more than $205 billion in announced semiconductor investment concentrated in Arizona alone, with Texas building a complementary corridor anchored by a $16.5 billion contractual link between Samsung’s Taylor fab and Tesla’s robotics and vehicle silicon. Section 2 traced the rare earth supply chain from Mountain Pass to Independence to a $400 million Department of Defense equity stake — the federal government’s largest and most structurally complete bet on any single company in this series, complete with a 10-year price floor and an offtake agreement guaranteeing the purchase of an entire facility’s output. Section 3 showed the semiconductor spine itself: fabs producing today in Arizona, fabs ramping in Texas, delays measured in months and quarters rather than cancellations. Section 4 turned to the constraint that ties all of it together — water — and found something more specific than a shortfall: a set of public claims that, examined individually, describe the state of that constraint less completely than their framing suggests.
What several billion-dollar projects already producing, and several more under active construction, demonstrate is that these are not speculative bets being made by capital hoping a policy environment holds steady. They are commitments structured to survive a change in administration, because the surrounding institutions — state legislatures funding water infrastructure before the fabs are finished, the Department of Defense taking equity rather than simply writing a grant, municipal governments rebuilding their utilities ahead of demand — are treating the underlying strategic problem as a fixed feature of the next decade, not a four-year policy preference.
The clearest evidence of that institutional permanence is the rare earth chain itself, and the timeline is worth stating plainly because it spans three administrations without reversing direction once. The Department of Defense’s first rare earth investment in MP Materials came in 2020, under the first Trump administration, a $10 million Defense Production Act award to a company struggling against Chinese retaliatory tariffs. [46] The Biden administration not only continued the relationship but expanded it — $45 million for Mountain Pass processing, more than $288 million to a second rare earth company, and a formal five-year mine-to-magnet investment strategy announced in 2024. [47] The second Trump administration then escalated that foundation into the $400 million equity stake, the price floor, and the offtake guarantee detailed in Section 2 — by far the largest single commitment in the chain’s six-year history. Three administrations, two parties, one direction, each handoff larger than the last.
This is the new mechanism this series has identified — equity rather than grants, ownership rather than subsidy — and it will be explored in full in the final paper of this series, which examines how the federal government is securing the raw material end of the supply chain across rare earths, copper, cobalt, and lithium, and what that ownership costs in dollars and control.
The water constraint is the one piece of this picture without a finish line, and it deserves a more precise closing than “underway but incomplete.” One company in this section set the standard by which the rest can fairly be judged. TSMC’s Arizona target — 90% recycling of the water that actually touches the wafer, not just the water that cools the building around it — was pointed explicitly at a demonstrated capability the company already runs in Taiwan, and was stated in public, in a community watching closely, with no regulator or contract requiring the company to say anything at all. Measured against that same bar, the rest of the buildout falls short in different and specific ways. Samsung’s reclamation partner announced a target three years ago and has answered no question about it since. Intel’s celebrated “net positive” result is real, and it is not the result this section is asking about. Sherman’s wastewater plant is a genuine, completed exception — but it treats water for the whole city, not for the two fabs it gets credited with serving, and it returns that water to a creek, not to either company’s production line. The one Sherman commitment actually aimed at a fab’s own process water, GlobalWafers’ federally tracked 50% target, is real and government-listed, and is also softened by language that does not require the number be reached. Texas Instruments’ matching 70% figure carries no such federal tracking at all, and the company’s own most recent disclosed results — 27% company-wide in 2020, 29% in 2023 — sit nowhere near it, with no comparable site TI has pointed to as precedent. And the water-engineering industry’s own flagship claim of 99% recycling, cited widely as evidence the private sector has this handled, cannot be confirmed to describe any facility in the region this paper is about.
None of this means the companies making these claims are lying. It means that TSMC’s Arizona fabs have been drawing water and producing chips since late 2024, that the plant designed to substantially close that gap will not exist until 2028, and that of everyone whose public statements this paper has examined on the hardest and most expensive part of the water problem, exactly one company has been consistently precise about what it has and has not yet done. Closing the mine-to-magnet-to-motor chain in Section 2 is a solved problem as of this writing. Closing the water gap in Section 4 is not — and the public record on how close anyone actually is to closing it is considerably thinner than it first appears.
REFERENCES
[1] Arizona Commerce Authority; Office of Governor Katie Hobbs, remarks at SEMICON West 2025, Phoenix, October 2025.
[2] Industrial Info Resources, Arizona Industrial Construction Pipeline, 2025.
[3] Office of Governor Greg Abbott, Texas Semiconductor Innovation Fund Grant Announcement, September 17, 2025.
[4] Samsung Electronics, Seoul Stock Exchange Regulatory Filing, July 28, 2025; Bloomberg, Reuters, CNN, July 28, 2025.
[5] Tom’s Hardware, “TSMC accelerates production timeline for new Arizona factory,” December 2025; Data Center Dynamics, “TSMC says Arizona fab is now ahead of schedule,” 2026.
[6] Data Center Watch (10a Labs), Q1 2026 Report; NBC News, “Study shows state and local opposition to new data centers is gaining steam,” 2026; Tom’s Hardware, “More than 75 data center build-outs worth $130 billion have been successfully blocked in the first three months of 2026,” 2026; Fox Business, “Chandler, Arizona, city council unanimously votes against AI data center,” December 2025.
[7] UltraFacility, “Semiconductor in numbers: Global fab construction timelines,” April 2026; Intel Newsroom, Ohio project status statements, 2025-2026.
[8] Rare Earth Exchanges, “How Rare Earth Elements Enable Modern Semiconductor Manufacturing Equipment,” January 2026.
[9] World Population Review / USGS rare earth production data, 2025; Optimusk, “Tesla Optimus Supply Chain,” 2026; 36kr English, April 2026.
[10] MP Materials Corp., Form 8-K, January 22, 2025, U.S. Securities and Exchange Commission; Fort Worth Report, “Fort Worth manufacturer begins producing rare earth magnets,” January 2025.
[11] MP Materials Corp., Form 8-K, July 2025, U.S. Securities and Exchange Commission; CNBC, “Pentagon to become largest shareholder in rare earth miner MP Materials,” July 2025.
[12] MP Materials Corp., Form 8-K, July 2025, U.S. Securities and Exchange Commission; The Defense Post, “Pentagon Takes Stake in US Rare Earth Company,” July 2025.
[13] MP Materials Corp., Form 8-K, July 2025, U.S. Securities and Exchange Commission; Select Committee on the Chinese Communist Party, U.S. House of Representatives, witness testimony, November 2025; C&EN, “US invests in rare earth firm MP Materials,” July 2025.
[14] Investing News Network, “Trilogy Metals Shares Rocket as US Government Takes Stake in Alaska Project,” October 2025; Axios, “US to take 10% stake in Trilogy Metals,” October 2025; Mayer Brown, “US Government Equity and Equity-Linked Investments in Critical Minerals,” April 2026.
[15] Internal Revenue Code Section 45X, Advanced Manufacturing Production Credit, as amended.
[16] U.S. Department of the Treasury, guidance on Section 45X critical minerals phasedown schedule, 2025-2026.
[17] MP Materials Corp., Form 8-K filings, U.S. Securities and Exchange Commission, 2025.
[18] Council on Foreign Relations, “The CHIPS Act: How U.S. Microchip Factories Could Reshape the Economy,” 2024; Semiconductor Industry Association, Chip Incentives & Investments data.
[19] TSMC Arizona, company facility status page, 2026; Tech Insider, “TSMC’s $165B Arizona GigaFab: Reshaping US Chips,” 2026.
[20] Tom’s Hardware, “TSMC brings its most advanced chipmaking node to the US yet,” December 2025.
[21] TSMC Arizona, company facility status page, 2026.
[22] BlackRidge Research, “TSMC Arizona Fab: USD 165 Billion Semiconductor Project,” 2026.
[23] Texas Instruments, “Texas Instruments begins production at its newest 300mm semiconductor manufacturing facility in Sherman, Texas,” December 17, 2025.
[24] EE Times, “Inside Texas Instruments’ New 300mm Fab in Sherman, Texas,” December 2025.
[25] TI.com, “Sherman, Texas: 300mm wafer fabs,” company site, 2026.
[26] Tom’s Hardware, “Samsung delays $44 billion Texas chip fab,” July 2025; Electronics360, “Report: Full production of Samsung’s Texas fab possibly delayed to 2027,” March 2026.
[27] TechTimes, “Samsung Taylor Fab Production Confirmed for 2027,” May 2026.
[28] MLQ.ai, “Samsung Delays Completion of $44 Billion Texas Chip Plant,” July 2025.
[29] CNBC, “Intel aims to find clients and catch TSMC with new chip fab in Arizona,” December 19, 2025; Tom’s Hardware, “Intel’s fab roadmap examined — Arizona, Ohio, Ireland, and the two deadlines deciding 14A process node,” June 2026.
[30] Ibid.
[31] U.S. Bureau of Reclamation, Colorado River shortage condition declarations, 2022-2026.
[32] Texas 2036, “Foundation for Economic Growth: Assessing Texas’ Water Infrastructure Needs,” 2024.
[33] World Economic Forum, “Semiconductor manufacturing and big tech’s water challenge,” 2024; CWR, “8 Things You Should Know About Water & Semiconductors”; IDE Tech, “Water Sustainability in the Semiconductor Industry”; Semiconductor Engineering, “How Semiconductor Fabs Use Water,” August 2025; SAMCO Technologies, “Industry Focus: Semiconductor industry trends and the importance of water resource management,” April 2025.
[34] EESI, “Data Centers and Water Consumption”; MOST Policy Initiative, “Data Center Water Use,” April 2026.
[35] TSMC, “TSMC Arizona and U.S. Department of Commerce Announce up to US$6.6 Billion in Proposed CHIPS Act Direct Funding,” pr.tsmc.com, April 2024; TSMC Arizona, “Sustainability” statement, tsmc.com, 2026; Data Center Dynamics, “TSMC breaks ground on water reclamation project in Phoenix, Arizona,” September 2025; Arizona Technology Council, “TSMC Breaks Ground on ‘Near-Zero’ Discharge Water Plant to Back Fabs,” September 2025.
[36] Ahwatukee.com / Times Media Group, “TSMC Arizona’s Water Reclamation Initiative Forwards Sustainability in Semiconductor Manufacturing,” September 2025, citing TSMC’s 2024 Sustainability Report.
[37] City of Phoenix, statement of Mayor Kate Gallego, August 2025; SemiWiki forum discussion, “In the city of Phoenix, TSMC Arizona will represent the world’s most advanced semiconductor technology in the United States,” January 2025; Senator Mark Kelly, “Kelly and Arizona Leaders Celebrate Finalized $6.6 Billion CHIPS and Science Act Award to TSMC,” press release, November 2024.
[38] Hoodline, “North Phoenix Scores Big As TSMC’s Second Chip Plant Hits Finish Line,” May 2026; Fortune, “Water-guzzling chipmaker TSMC and drought-plagued Arizona are an unlikely pair, but Phoenix says it has enough water,” April 2024.
[39] EPCOR USA, “EPCOR Tapped as Water Partner in Central Texas,” July 13, 2023; DNA Systems, “EPCOR – Sandow Water Project,” project completion record, October 2024; East Wilco Insider, “Looking for Enough Water,” September 1, 2023.
[40] ScaleBan Equipments, “How Semiconductor Industry is Tackling Wastewater Challenges”; industry reporting on Intel Ronler Acres end-of-pipe treatment mechanism and Intel’s 2022 net positive water milestone.
[41] City of Sherman, Texas, “Wastewater,” official city website; KXII, “Sherman unveils new multi million dollar wastewater treatment plant,” October 15, 2025.
[42] National Institute of Standards and Technology, “GlobalWafers (Texas),” nist.gov/chips, CHIPS award terms and environmental commitments.
[43] Texas Instruments, “Texas Instruments announces award agreement for CHIPS and Science Act funding,” ti.com, December 20, 2024; National Institute of Standards and Technology, “Texas Instruments (Utah),” nist.gov/chips.
[44] Texan By Nature, “Texas Instruments,” txn20.org, 2020 water reuse disclosure; Texas Instruments, 2024 Corporate Citizenship Report, cited via MarketScreener, June 2025 (2023 water reuse figures).
[45] The Register, “Mystery German chip fab sips on Gradiant’s ultrapure water,” January 22, 2024; Manufacturing Dive / ESG Dive, “Semiconductor industry faces water, sustainability challenges,” August 2025.
[46] Heatmap News, “The Pentagon’s Rare Earths Deal Is Making Former Biden Officials Jealous,” July 2025.
[47] The White House (Biden-Harris Administration Archives), “Fact Sheet: Biden-Harris Administration Takes Further Action to Strengthen and Secure Critical Mineral Supply Chains,” September 2024; Global Policy Watch, “Made in America: The Outlook for Critical Minerals,” October 2025; Vulcan Elements, “Money Finally Flowing to US Rare Earths Can’t Come Fast Enough,” August 2025.
FORTRESS AMERICA SERIES
Paper I: Fortress America — Hemispheric Consolidation, Infrastructure Cascades, and the Regional Transformation of the American Economy
Paper II: If It Can Work, It Can Fight — The Case for Domestic Humanoid Robotics Manufacturing
Paper III: Where Fortress America Lands — The Southern Spine and the Race to Build It
Paper IV: The Government Stake — Equity, Speed, and the Limits of Disclosure
Paper V: The Atom and the Chip
Paper V.5: Forthcoming
This paper represents independent analytical and systems research and is the third in the Blue Collar Analytics Fortress America series. Nothing contained herein constitutes financial advice, investment recommendations, or legal counsel. Readers should conduct their own due diligence and consult qualified professionals before making investment decisions.
