Chip Manufacturing: Grants and Tax Credits, Not Government Equity

Why the underlying bet is working

The technical and commercial signals through 2026 are strong, not just aspirational: TSMC's Arizona operation posted a $514 million profit in a single quarter, evidence the facility is commercially viable, not merely subsidized, and its second Arizona fab began installing 3-nanometer/2-nanometer equipment ahead of a 2027 production target. Intel's Chandler, Arizona fab reached high-volume manufacturing on its 18A process, the first US-based facility to cross the 2-nanometer-class threshold, a milestone for a company that entered this period in financial distress. That distress is also why the picture is uneven rather than uniformly positive: Intel's Ohio "Silicon Heartland" megasite has slipped to 2030, years behind its original target, and Samsung's Texas fab pushed its 2-nanometer production timeline to late 2026, both slippages traceable to company-specific financial and demand conditions, not to a failure of the underlying federal policy.

The equity-stake precedent, and why this issue rejects it regardless of who proposes it

In August 2025, following a meeting between Intel's CEO and the President, the administration converted $5.7 billion in unpaid CHIPS grants plus $3.2 billion in separate Secure Enclave program funds into an approximately 10% federal equity stake in Intel. The objections to this came from different directions, which is itself informative: Senate conservatives (Rand Paul: "wouldn't the government owning part of Intel be a step toward socialism?") framed it as a philosophical break from free-market principle; Cato Institute analysts called it "a dangerous turn in American industrial policy"; and Senate Democrats (Elizabeth Warren) objected on more concrete terms: the shares were purchased at a discount to the prevailing market price, approved without a shareholder vote, and came with none of the labor, union, or childcare conditions the original Biden-era CHIPS grants had attached. A shareholder derivative lawsuit is now testing the governance implications directly.

This issue's position isn't that the government shouldn't get anything for its money: the original CHIPS Act's conditional-grant model already did that, attaching strings (workforce commitments, cost-sharing, profit-sharing above a threshold) without making the government a part-owner of a publicly traded company whose stock price the same government can move by policy announcement. An equity stake creates a structural conflict of interest a grant with conditions does not: the government both regulates and part-owns the same company, and gains a direct financial stake in decisions (layoffs, stock buybacks, foreign sales) that its policy arm is supposed to evaluate neutrally. That conflict is the specific, mechanism-level objection here, not a blanket claim that public money should come with no strings attached.

The H-1B fee working against the onshoring goal it's supposedly unrelated to

The $100,000 supplemental fee on new H-1B petitions filed for workers currently abroad applies directly to the pipeline TSMC and Samsung use to bring process engineers and equipment specialists to staff their US fabs during the highest-skill startup phase. TSMC's Arizona site alone filed dozens of H-1B petitions in the relevant categories in a single fiscal year, and the company has said roughly half its Phoenix-area workforce has held some form of Taiwanese work authorization. A "national interest" exception exists on paper for strategically sensitive sectors including semiconductors, but the relevant agency has signaled such exceptions will be granted only rarely. The result is a tension inside the administration's own policy: an immigration fee designed to encourage domestic hiring directly raises the cost of the specific technical expertise transfer the same administration's chip policy depends on in its startup years, before a fully domestic workforce pipeline exists to replace it.

The workforce gap and the hiring-discrimination question, read together

The Semiconductor Industry Association projects roughly 67,000 unfilled semiconductor-sector jobs by 2030, about 22% of the industry's needed workforce, against workforce-training commitments (CHIPS's own ~$200 million earmark, plus company-specific commitments from TSMC, Intel, and GlobalFoundries) that exist but aren't close to closing that gap at current scale. At the same time, a class-action lawsuit alleges TSMC's Arizona operation systematically favored Taiwanese hires over qualified American applicants and conducted internal communications primarily in Mandarin, a direct tension with the domestic-workforce commitments the company made in exchange for its $6.6 billion CHIPS award. Both problems pull in the same direction rather than opposite ones: underfunded domestic training plus unenforced hiring commitments together produce exactly the pattern critics of the whole program point to, even though the underlying manufacturing investment itself is working.

Water use: substantial, but disproportionate only against the wrong baseline

A modern leading-edge fab needs millions of gallons of ultra-pure water daily for wafer cleaning and cooling. TSMC's Arizona operation alone is projected to need roughly 17 million gallons a day across its planned fabs, and Intel's Chandler campus around 14 million more, in a state where Colorado River allocation cuts are already deepening and where Phoenix-area groundwater has lost tens of millions of acre-feet over the past two decades. That's a non-trivial new demand on an already-stressed water system, and this issue doesn't minimize it. But it's also worth stating precisely rather than let a vaguer "water-guzzling factory" framing stand unchallenged: Arizona's agricultural sector consumes roughly 70% of the state's water, against all industrial use combined, semiconductor fabs included, at around 5%. Both companies have made infrastructure investments in response (TSMC's on-site reclamation plant targeting a high recycling rate, Intel's reclamation facility and aquifer-restoration partnerships), though independent reporting has found their self-reported recycling percentages inconsistent across different accounting methods and has flagged that "net positive" framing can obscure large absolute volumes and uneven performance across a company's different global sites. This issue's answer is neither to trust the self-reported numbers nor to dismiss the underlying concern: it requires the kind of independent verification that would settle which characterization is accurate.

The contamination legacy this industry hasn't fully reckoned with

Water volume isn't the only environmental question, and arguably isn't the most serious one: water contamination is the semiconductor industry's documented legacy, not a hypothetical risk. The original Silicon Valley, Santa Clara County, California, carries one of the highest concentrations of EPA Superfund sites in the country, a direct result of 1980s- and-90s-era semiconductor manufacturers dumping industrial solvents (trichloroethylene and similar chemicals) into unlined pits and letting them reach groundwater, contamination that in several cases remains under active remediation decades later. Celebrating a new round of CHIPS-funded fab expansion without engaging that specific history, not just today's water- volume numbers but yesterday's contamination record from the same industry, would be a gap in this issue's own accounting, not a manufactured one. This issue's answer is the remediation-bonding requirement in Proposal 4: not a claim that today's fabs will necessarily repeat that history (modern facilities operate under materially stricter environmental regulation than 1980s ones did), but an acknowledgment that "modern facilities are safer" was also true, relatively speaking, at the time, and that the honest response to a historical pattern is to require financial assurance against it, not just an assumption that this time will be different.