Semiconductor chips are vital to modern existence: they allow us to drive to work, call our mothers, and much, much more. Unfortunately, demand has far outstripped supply. Manufacturers produced more than one trillion chips in 2021, and yet, shortages abounded. In Arizona, semiconductors have a surprisingly long history, dating back to the mid-1950s. As Steven Zylsta, president and CEO of the Arizona Technology Council aptly observed to a local NPR reporter, “We have no floods, hurricanes, tornadoes, earthquakes––all those things that could really upset semiconductor manufacturing.”
Most recently represented by Intel, the semiconductor industry’s presence in the Phoenix metropolitan area is only growing. Taiwan Semiconductor Manufacturing Company (“TSMC”), the world’s largest semiconductor manufacturer, broke ground on a new facility in North Phoenix in January 2022. Yet another semiconductor component manufacturer is set to build a factory in Chandler. Unfortunately, while a great boon for economic vigor, this explosion should give Arizonans pause: semiconductor manufacturing has a massive water footprint.
Of course, chips are not the only good in short supply. Arizona is undeniably running out of water. Thirty-six percent of the state’s surface water comes from the Colorado River, which reached Tier 1 water shortage levels in late 2021. Worse, Arizona’s groundwater, which provides roughly 40 percent of the water supply, is dependent on the Colorado River for recharge. But as Intel’s success in water management shows, the situation is not hopeless.
Sarah Porter, of the Morrison Institute’s Kyl Center, lives and drinks Arizona water issues and is deeply familiar with the compromises that will define the state’s future. When asked about TSMC and an Intel expansion, she cited data that shows “a million gallons of water can provide 200 high-paying semiconductor jobs, 30-40 lower-paying data center jobs, or about 50 still lower-paying golf course jobs.” As Porter acknowledges, “If you’re going to use your water for something, use it for lots of high paying jobs.” Further, semiconductor companies are able to anticipate and reduce water usage in a way that few water users can.
Enter TSMC. City of Phoenix (“City,” or “Phoenix”) officials anticipate that TSMC will use roughly 10,000 acre-feet each year to operate its plant with one 5-nanometer fabricator (“fab”). The plant could use up to 40,000 acre-feet if it adds five more fabs, as seems possible. Still, as Porter recognized, “A Phoenix that has only semiconductor plants and no craft beer factories and no golf courses is probably not a Phoenix that a lot of people would want to live in.” Surely, the solution is not to sacrifice Phoenix so the rest of the world has access to chips. (And making chips in Phoenix necessarily requires employees willing to live nearby.)
As is the case with semiconductor chip design and manufacture itself, the answer is balance. Semiconductor manufacturers will optimize water use for the benefit of their bottom line. This article argues in favor of crafting state and local policies that go further, requiring these companies to invest in water efficiency for the benefit of the local community. It further argues that mandatory water-neutral development policies do not run afoul of federal or state takings law. While policies that mandate a net-positive impact rise to a taking, state and local policies can—and should—facilitate water-positive development through voluntary incentives.
The TSMC plant is a massive endeavor. Each chip requires just shy of eight gallons of water; a cotton plant requires around ten gallons. To operate just one fab, TSMC will use 8.9 million gallons a day (MGD), or nearly 3 percent of the City’s current water production. For reference, residential use—including outside water uses, such as lawn watering—totals roughly 200 MGD. This water must come from somewhere, but where?
Broadly, the City of Phoenix has four primary water sources: surface- and groundwater from the Salt and Verde Rivers, delivered by the Salt River Project (SRP); Colorado River water, delivered via the Central Arizona Project (CAP); local groundwater; and reclaimed water. But 52 percent of the City’s water comes from SRP, and the TSMC plant is located outside of SRP boundaries (i.e., “off-project”).
Thus, TSMC will not be able to use SRP’s 350,000-plus acre-feet/year. To fulfill TSMC’s water demands, the City of Phoenix must use CAP water, local groundwater, reclaimed water, and non-SRP Salt and Verde River supplies. Phoenix plans on these non-SRP sources supplying as little as 140,000 acre-feet/year, in the case of “severe prolonged drought,” and as much as 180,000 acre-feet/year, in the case of “moderate drought.” This part will explore the contours of TSMC’s water demand and the water practicably and legally available to meet that need.
TSMC will use at least 10,000 acre-feet and as much as 40,000 acre-feet each year. A volumetric unit, an acre-foot is roughly equivalent to the amount of water needed to flood a football field one foot deep. But whereas water on a football field quickly disperses, the TSMC plant may be more aptly compared to a pool, which—once filled—only requires “topping off.”
Traditional industrial processes, such as semiconductor manufacture, require vast amounts of water. During the manufacturing process, chips must be repeatedly rinsed with ultra-pure water (UPW); one chip might require as much as 2,200 gallons of UPW. It takes 1.4 to 1.6 gallons of municipal water to produce one gallon of UPW, so the rinsing process can require as many as 3,520 gallons of tap water. In addition, large quantities of water go to cooling processes: wet-cooling is cheaper and more efficient than dry-cooling.
The energy demands of semiconductor manufacturing also tax water supplies. A 2016 study estimated that 90 percent of semiconductor plants’ water needs were for electricity. The plants require as much as 2,400 megawatts each day—as much as 50,000 homes and more than many other industrial uses. HVAC systems can be responsible for as much as 65 percent of energy consumption. In addition to requiring exceptionally clean air, cleanrooms require very particular—and consistent—humidity and temperature levels. Chillers operate year-round, providing cold air for AC units.
Climate impacts aside, this energy demand is costly. In theory, energy—and water—savings are mutually beneficial to the manufacturer and surrounding community. But energy efficiency sometimes comes at the cost of reliability (i.e., overkill in process specifications), and manufacturers can be hesitant. Still, a 2013 McKinsey & Company study identified process-cooling water system improvements as “always able to generate quick wins.”
TSMC has claimed that roughly 65 percent of the new fab’s water needs will come from “in-house water reclamation,” with the express purpose of reducing city water consumption. This number is conservative, as newer technologies allow semiconductor plants to achieve as much as 98 percent recycle. Notably, reuse has its limitations: water cannot be in two places at once, and thus, a plant’s water supply must accommodate the maximum amount needed at any given second, not the plant’s daily or annual use. Still, on-site reuse will be TSMC’s saving grace. Once the plant receives enough water to become operational, TSMC will need just 3,500 acre-feet/year, or 1 percent of the City’s current usage, to operate as a one-fab facility.
To meet this demand, the City will allocate unused CAP water, supplemented by the Salt and Verde Rivers. Because the TSMC project involves the purchase of state lands, Phoenix’s CAP allocation will increase. Colorado River shortages have not yet affected Phoenix, but cutbacks—voluntary and compulsory—are coming. Tribal water leases will be vital in bridging any gaps. With SRP water off limits, any Salt and Verde River rights must come from independent sources. Fortunately, the City began securing these rights as early as the 1950s through strategic investments in the Salt River’s Roosevelt Dam and the Verde River’s Horseshoe Reservoir. While totaling 57,300 acre-feet/year, the rights are vulnerable to shortages.
Phoenix may also supplement surface water with existing reclaimed water supplies. The City of Chandler has partnered with Intel to build a pre-treatment facility that will allow Chandler to supply Intel with reclaimed water rather than potable water. Phoenix and TSMC could do the same. Even after watering its beloved golf courses and other “turf-related facilities,” Phoenix commonly has surplus reclaimed water. Allocating this surplus to TSMC would maximize taxpayer dollars and save higher-quality water for non-industrial uses.
Of course, despite highly efficient recycle, TSMC will generate wastewater. The City of Phoenix expects to recoup upwards of 80 percent of TSMC water, which is similar to what Chandler recovers from Intel. Intel dedicates much of its wastewater (after appropriate levels of treatment) to aquifer recharge in Chandler, and TSMC could do the same in Phoenix. In addition to this indirect reuse, wastewater might also irrigate golf courses. Ultimately, when armed with the water treatment expertise and budget of a semiconductor manufacturer, 8,000 acre-feet of wastewater is nothing to sneeze at.
Two frameworks—water-neutral development and the mitigation hierarchy—are helpful in discussing water resource management practices and policies.
At their core, water-neutral development (WND) policies require water users to offset their water demands. Some programs aim to improve water efficiency amongst existing water users by incentivizing retrofits. Others target new development, such as subdivisions or industrial complexes, by conditioning permit approval on water neutrality. When applied to new development, water-neutral development has two stages. First, developers typically reduce on-site demand through design decisions. Second, they must offset any on-site use through offsite action. Because the second step is mandatory, developers can choose how to act but do not have a choice in whether to act.
Whereas water-neutral development is measured, if not resigned, in its goal—neutrality, other sustainability frameworks are slightly more ambitious, imagining ways that development can benefit sustainability goals. The mitigation hierarchy is one such framework. Developed by the International Finance Corporation to protect biodiversity amidst global development efforts, the mitigation hierarchy allows developers (or financiers) to set a goal of either “no net loss” or ”net gain” of biodiversity. Thus, the hierarchy offers a path toward water-positive development.
The hierarchy has four stages: (1) avoidance; (2) minimization; (3) remediation; and (4) offset. In the biodiversity context, the primary concern is a project’s impact on the immediately surrounding ecosystems. As such, the first three stages occur on-site. Here, TSMC needs off-site water, so off-site actions are relevant as early as the minimization stage. Through remediation and offset, TSMC can achieve water-positive development.
First is avoidance, which avoids impacts by taking affirmative, preventative actions. Choosing to build the TSMC plant in a water-abundant region would be the most avoidant option. But in lieu of that, many decisions made during plant design and construction fall into this category. Broadly, avoidance requires engineers and architects to optimize processes and the facility itself to require as little water as possible. Fortunately, improving water efficiency is often good for the bottom line. It also satisfies engineers’ itch to optimize each and every variable.
Next is minimization, which serves to minimize the “duration, intensity, and/or extent” of impacts that could not be practically avoided. In this context, water use is an unavoidable impact. Thus, minimization involves the decisions surrounding sourcing of water, first and foremost through reuse. Amidst the Colorado River’s mounting shortages, minimization may entail TSMC securing non-CAP water.
Then comes restoration, or remediation which restores or remediates on-site (or adjacent) impacts that could not be avoided or minimized. In the context of TSMC, remediation involves the quality and quantity of wastewater that leaves the plant. Once adequately treated and discharged, TSMC’s wastewater will be a boon to Phoenix rather than a blight. Thus, TSMC’s wastewater offers an opportunity for a net-positive impact.
Only after these stages comes offset, which compensates for “residual, adverse impacts” that could not be avoided, minimized, or remediated. Typically, these offsets occur off-site. For example, Intel offsets its water use by funding projects that will restore roughly one billion gallons to Arizona’s waterways each year. A mandatory offset might also take the form of an impact fee paid by developers, which some jurisdictions use to fund conservation programs. Offsets can exceed a project’s water demand, leading to a net-positive impact; Intel has committed to net-positive water use by 2030.
Approval of the TSMC project is subject to the City of Phoenix’s permitting decisions. In turn, takings law constrains what the state and local officials can require of developers.
An outgrowth of the police power, states have largely delegated their zoning authority to local governments. Phoenix’s City Council makes zoning decisions based on recommendations by the Planning Commission, which is an appointed body. The Phoenix General Plan, which is implemented and enforced through ordinances, broadly guides these zoning decisions. State law requires that the General Plan be updated or readopted every 10 years by a public vote; voters approved the most recent update in 2015.
The 2015 General Plan identified three “Community Benefits”: Prosperity, Health, and Environment. As part of the latter, it identified climate change as a “significant challenge,” and both water resource planning and ecosystem protection as “[o]pportunities.” TSMC’s development lies at the intersection of two core values: “strengthen[ing] our local economy” and “build[ing] the sustainable desert city.”
The General Plan also calls for the City to “[p]artner with the private sector to responsibly develop new infrastructure and water supplies that accommodates growth in a fiscally prudent and sensible manner.” As part of this, it should “[e]ncourage water efficient building and site design.” Examples of ordinances enforcing these principles include xeriscaping requirements for public rights-of-way and a requirement that City water customers keep customer-side water infrastructure in good repair.
The City’s Water Resources Acquisition Fee (WRAF) program implements these values during the permitting process. WRAF is an impact fee collected in certain growth areas to ensure that developers pay their proportionate share of their impact on Phoenix’s water infrastructure and supplies. Currently, only off-project development is subject to WRAF, reflecting the challenges associated with securing non-SRP water.
Alternatively, developers can show their project “will have features that provide permanent reduction in net annual water demand on the City.” Such features include “[w]ater-conserving plumbing fixtures” or “[n]on-City water resources procured by the developer which . . . partially or wholly satisfy the proposed development’s water resource needs.” The water savings from these “features” are credited against the project’s WRAF. Thus, akin to a WND scheme, a developer has two options: reduce or pay.
Unfortunately, because of statutory limitations on development fees, WRAF revenue can only be used for water rights acquisition, related infrastructure, and water efficiency programs; revenue is not available for conservation-driven offset actions. Notably, these statutory limits are largely crafted to ensure that development fees do not constitute a taking. Thus, the next part considers whether a more ambitious and clearly articulated WND mandate would be constitutionally viable.
The federal Takings Clause succinctly states, “[N]or shall private property be taken for public use, without just compensation.” Historically, it applied to per se, physical takings of land. More recently, a doctrine has emerged for regulations that “go ‘too far’” and “restrict[] a property owner’s ability to use his own property.” The U.S. Supreme Court’s most recent takings decision, Cedar Point Nursery v. Hassid, purportedly clarified this distinction but arguably muddied it. This article sidesteps these new issues by considering water-management policies lacking in any physical appropriation, however temporary.
Two cases are hallmarks of the U.S. Supreme Court’s approach to zoning exactions: Nollan v. California Coastal Commission and Dolan v. City of Tigard. Nollan introduced the requirement that a regulation bears a “rational nexus” to a legitimate state interest, and Dolan added a proportionality standard. In both, a project’s anticipated burden, or impact, is the relevant baseline.
Here, there is a clear nexus between mandated water-neutral development and TSMC’s impact on the City of Phoenix’s water infrastructure and water supply. Unlike the “psychological barrier” to beach access in Nollan, TSMC’s water demand is quantifiable and tangible. And by nature, WND requirements are proportional. Conversely, a net-positive mandate would face both nexus and proportionality problems: once a project achieves water neutrality, it cannot pose a burden to the government. With no burden, there can be no nexus, and no exaction can be proportional. Thus, mandatory water-positive development programs are likely to be unconstitutional takings without just compensation.
Perhaps surprisingly, public use is not a condition precedent for a lawful taking. In Kelo v. City of New London, the Court approved the use of eminent domain to seize private land to then sell to private developers. It reasoned that the economic benefits of private development were a “public use” within the meaning of the Takings Clause. Notably, the Fifth Amendment Takings Clause does not require literal public use, but rather a “public purpose.”
While “public purpose” is not the sole trigger for a lawful taking, and Kelo squarely involved traditional condemnation, it is helpful to contemplate Kelo’s potential reach. Here, at a minimum, WND policies further a public purpose, and some policies may directly further a public use. For example, a law dedicating impact fee revenue to forest health furthers a public purpose, i.e., long-term water management; a law dedicating the same revenue to necessary infrastructure improvements furthers a public use. The same can be said of water-positive development policies: a law requiring developers to invest, above and beyond their own impact, in long-term watershed health certainly furthers a public purpose.
Thus, Kelo spurred national concern that private land could be condemned for the vaguest of public purposes. Perhaps forecasting the likes of water-positive development regulations, property rights advocates went a step further, arguing that Kelo showed the Court’s willingness to allow regulatory takings to obliterate property rights as we knew them. In response, they turned to ballot measures. Notably, only Arizona’s Proposition 207 passed.
Arizona has enshrined additional protections against eminent domain power in its constitution and by statute. Article 2, section 17 of the Arizona Constitution states that “[n]o private property shall be taken or damaged for public or private use without just compensation.” It further notes that “whether the contemplated use be really public shall be a judicial question,” decided without deference to the legislature. As with the Fifth Amendment Takings Clause, diminution in value alone is not a taking; a regulation must prevent a property from being used for any “economically viable” purpose to which the land is “reasonably adapted.” WND requirements—which oftentimes align with internal corporate interests—are unlikely to prevent semiconductor manufacturers like TSMC from using the property as intended.
Proposition 207, now the “Private Property Rights Protection Act,” aims to protect Arizona’s private property owners from partial regulatory takings, i.e., state and local regulations that diminish property value but fall short of outright condemnation. Under section 12-1134, a landowner is owed just compensation if three conditions are met. First, the regulation must be a “land use law” that was “enacted after the date the property is transferred to the owner.” Second, the law must reduce “the existing rights to use, divide, sell or possess private real property.” Third, and crucially, the law must reduce the property’s fair market value.
Section 12-1134 includes several exemptions, a few of which are relevant. It exempts laws that “[l]imit or prohibit” uses for public health and safety, including “health and sanitation” and “pollution control.” For example, laws requiring dustproof paving of certain parking lots and driveways are exempt as pollution control measures. Still, protecting health and safety must be the law’s “principal purpose”; “[m]ere declaration” is insufficient. It also exempts regulations that “[d]o not directly regulate an owner’s land.” In addition, section 12-1134 allows cities to request that landowners voluntarily waive any “Prop 207” rights, and developers that decline to do so may receive a negative recommendation from the city’s planning department.
Ultimately, section 12-1134 is unlikely to prevent WND requirements. Regulations that require semiconductor plants to optimize their on-site water use, i.e., avoidance, almost certainly increase the plant’s property value. In addition, regulations that address the water quality of plant discharge are safely within the public health and safety requirement. And because a land use law must “directly regulate” land, offsite measures, i.e., offset requirements, are also likely exempt.
There is a clear legal basis for requiring semiconductor manufacturers, such as TSMC, to engage in water-neutral development. Most obviously, there are numerous requirements at the state and local levels in Arizona that already require developers to reduce their water impact. A scheme requiring new semiconductor plants to be water-neutral further survives any federal or state takings issues, including that of “Prop 207.”
Still, Arizona’s current statutory limitations on development fees prevent water-neutral development, and the WRAF program in particular, from reaching its full potential. Legislative change is necessary to allow WRAF revenue to finance nontraditional water resource projects. At the very least, the City of Phoenix should modify the WRAF credit program to allow developers to privately offset impact through investment in conservation projects. On the other hand, mandating water-positive development likely fails as an unconstitutional exaction that lacks an appropriate nexus. Still, state and local actors can enact policies that incentivize semiconductor companies to voluntarily pursue water-positive development.
The City of Phoenix should modify its existing WRAF program to better reflect a mitigation hierarchy approach to water management. First and foremost, it must allow for offsets that go beyond short-term augmentation. Unfortunately, statutory limits on development fees hamstring progress in this area. To be a true WND policy, WRAF needs to expand to fund and facilitate rehabilitation, infrastructure improvement, and conservation efforts.
Current takings law does not prevent regulations that mandate water-neutral development. As long as the fee remains proportional to impact, WRAF can—and should—be dedicated to offset strategies. If prior appropriation can be updated to recognize in-stream flows as a beneficial use not subject to forfeiture, then conservation efforts should be fairly seen, per statute, as a “beneficial use” that is “associated with providing necessary public services” to new, water-intensive development. It is merely a matter of being willing to consider long-term impact versus short-term impact.
The state legislature can make this change by expanding WRAF’s definition of “necessary public services” to include “water offset projects.” These projects should be defined broadly, allowing in-state investments in watershed and forest health, tribal infrastructure improvements, or reservoir storage. This change can be made while still requiring the City to partially shoulder infrastructure construction costs and fully pay operating costs. The City already has authority to distinguish between types of industrial uses, and thus, it could limit this broader reading to semiconductor manufacturers.
A more measured change would be expanding Phoenix City Ordinance section 30-5 to expressly credit developers, including semiconductor manufacturers, for in-state or in-basin offsets. This route sacrifices the benefits of a coordinated city-level approach to financing water management projects. Regardless, requiring semiconductor manufacturers to offset their water impact will provide a novel funding source for projects that are vital to Arizona’s long-term success but notoriously difficult to finance.
The credit system is ripe for change in other ways. Not all water is created equally, and credits for procuring water should reflect that. Chip manufacturing is important but does not warrant depleting Arizona’s groundwater. Thus, the City should offer a much-reduced groundwater credit that disincentivizes use of groundwater rights. Such a scheme will better internalize the long-term impact of groundwater depletion.
Last, the City needs to greatly increase transparency in its use of WRAF funds. Even without the modifications suggested here, WRAF reflects WND principles and should be advertised this way. Still, to best invest in long-term water security, the City must look beyond short-term augmentation. If the state legislature is unwilling to adapt, the City should encourage semiconductor manufacturers, which reside at the forefront of technology and efficiency, to do so instead.
While state or local governments cannot mandate water-positive development, they can readily develop incentive programs. For example, the state legislature could develop a property tax scheme that grants a credit or reduced tax rate to projects that demonstrate water-positive impact. Approved projects would be subject to an annual audit that would confirm ongoing compliance. At the local level, the City of Phoenix and other municipalities should consider developing an expedited permitting process for water-positive designs. In addition, the City of Phoenix should add water-positive development as a goal in its 2025 General Plan.
Together, these changes can make water-neutral development the new baseline for semiconductor manufacturers. Codifying best practices has the added benefit of heading off public concerns; in theory, an approved project will necessarily be a responsible water user. As the City explores more ambitious water-management strategies, semiconductor manufacturers are ideal test cases and partners. Armed with resources that most can only dream of, semiconductor manufacturers can help Phoenix become “the sustainable desert city.”
