Is Net Positive Water a Thing?

Semiconductors are the foundational building blocks of all computing devices–whether that’s an AI model, a car, a phone, or a cheap greeting card that lights up and sings a song. And making semiconductors needs lots of water. Semiconductor manufacturers like Intel and TSMC are keenly aware of their water demands. Increasingly, so too are platforms like Meta/Facebook. Each of these companies report on their water demands in their sustainability reports, but it’s important to carefully consider how these companies report their water use.

In Intel’s case it is keen to offer impressive sounding numbers. For example, an Intel representative speaking to an industry trade journal claims that 9 million gallons of water are treated daily at the company’s Ocotillo campus in Chandler, Arizona. An additional 1.5 million gallons used by Intel’s facilities there are treated at an adjacent site. Intel also claims its New Mexico facility treats and returns 90 percent of that facility’s water needs (Dorsch 2021). These might be impressive figures, but in the absence of context they can be misleading.

Philosopher C. Thi Nguyen (2024) notes that there are always trade-offs to turning people, places, and things into data, to data-fying them. Quantified data, for example, permits useful operations such as aggregation and portability, but removes context and nuance that might nevertheless be deeply meaningful for particular people, places, and things and their persistence. Data-ification, including quantification, is always done in the service of someone’s interests. There is no disinterested data, but that doesn’t make data inherently ‘bad’. Nor does it make data ‘good’ or ‘neutral’. Instead, useful questions to ask about data include: who or what gets turned into data–whether quantitative or qualitative–by whom, for what purpose(s), when, where, with what effects for whom, on whom, and under what conditions (D’Ignazio and Klein, 2020)?

For example, data-ification (especially quantification) is useful for aggregating classifications of characteristics or variables into things like environment, social, and governance (ESG) reports. ESG reports are increasingly part of the gamut of information used by investors to make investment decisions. If you’re one of the principal institutional investors in companies like Intel, TSMC, or Meta, then those decisions are made at sites like 50 Hudson Yards in New York City (home of Blackrock), 45 Liberty Blvd in Malvern, Pensylvania (home of Vanguard), or 245 Summer St., Boston, Massachusetts (home of Fidelity), or perhaps at one of these company’s many other corporate sites. Wherever those decisions are being made, it’s rarely if ever in the watersheds where the activities associated with water consumption and/or restoration of companies like Intel, TSMC, or Meta/Facebook are happening. Data-ification makes those watersheds – and the people, places, and things particular to them – transportable via quantification to decision makers elsewhere. That travel strips those people, places, and things of the contexts important to their functioning, but grants them the ability to travel, to be compared, and to be assessed for their worth to others by others.

Geographer Morgan Robertson writes, “[t]he rise of ecosystem service markets allows us to observe – in much the same way that the rise of labour markets did – a transformation of the social world through creation of value-bearing abstractions from physical processes.” (Robertson 2012, 387). Net positive water is not about abstracting physical ecology for services per se. But it is about abstracting to circulate value of (ostensibly) functional hydrology. Perhaps it makes sense to think of the data-ification of hydrology-as-net-positive-water as extraction by abstraction—the translation of ecological processes of hydrology into data-as-numbers that can be aggregated and circulated beyond their watershed. It’s the data that circulate as something legible and, thus, valuable to capital. It’s true that the hydrology of net positive water is not being priced in the reports by Intel, TSMC or Meta/Facebook. But the data-ification of net positive water is part of the process by which real, ecological relationships are transformed into circulating references that centres of calculation (as Latour (1999, Chapter 2) would call them) use to make prices (e.g., for Intel’s, TSMC’s, and Meta’s publicly traded stocks) .

Net positive water is about crediting water ‘restored’ in one place against water debits in another place. In its 2019 sustainability report Intel, for example, discloses that it used 12.6 billion gallons across its operations. If 90 percent of this total water used is treated and returned for other uses, that means 1.26 billion gallons are released untreated into whatever hydrological systems receive them, whether those are municipal water systems, lakes or rivers, the atmosphere, etc. Waste water from semiconductor manufacturing contains a number of chemicals known to be toxic, including PFAS (‘forever chemicals’). Even when that waste water is treated, such chemicals are not entirely eliminated.

Intel’s same sustainability report discloses that between 2015 and 2019, its total demand for water rose 40 percent, up from 9 billion gallons in 2015 to 12.6 billion gallons in 2019 (Intel Corporation 2020, 22). It’s telling that although the company discloses these figures, it is keen to talk about the volume and rates of water conserved which suggests its demand is in check. It isn’t. A 40 percent increase in water demand over four years is a good demonstration of the rebound effect (or Jevons Paradox)—even as the company treats and reuses higher percentages of the water it needs, its aggregate demand for water goes up (Jevons Paradox basically describes the relationship between efficiency gains and changes in aggregate demand. Gains in efficiency typically result in increases, not decreases, in aggregate demand. Why? Because as efficiencies are gained, less of what was previously used is used; that leaves more of X–whatever ‘X’ is–available than before, thus increasing supply of X relative to demand for it. When supply exceeds demand, prices tend to fall making X available to a larger number of ‘demanders’ than before and, thus, aggregate demand for X increases).

A similar pattern is evident at Taiwan Semiconductor Manufacturing Corporation (TSMC). Its 2022 sustainability report discloses that TSMC’s water consumption increased by over 35  percent between 2020 and 2022 (TSMC 2022, 229). Whereas in 2020 TSMC needed 77.3 million metric tonnes of water, those needs increased to 82.8 million in 2021, and to 104.6 million metric tonnes in 2022. These increases in demand for water happened while the company’s claimed water recycling rates hovered between about 85 and 86 percent. The rebound effect bites again. What’s more, in TSMC’s case, more than 90 percent of its water demand occurs in Taiwan, where the bulk of the company’s facilities are—and Taiwan has been under drought conditions for a number of years (drought conditions are bad enough that Taiwan’s farmers are being paid not to grow food so that water can be diverted to chip manufacturing; Fitch Ratings 2021; Narvaez et al. 2021; Reuters 2021).

Under these conditions companies like Intel and TSMC have started talking about achieving ‘net positive water’ in their manufacturing operations. And, it’s not just semiconductor manufacturers getting in on net positive water. So is Meta/Facebook, which set itself the goal of being net positive on water by 2030. Like Intel, Meta claims they are going to do this through water restoration projects in the watersheds that their infrastructure operates using a methodology developed by the World Resources Institute (Reig et al. 2019).

At a basic level, net positive water means putting more water back into a system than is taken out for use in a company’s process (e.g., for manufacturing; for hosting a Meta profile; for serving an ad). The ‘net’ part of ‘net positive’ is, however, the rub. Here’s why: there’s some basic conservation of mass issues going on. At a global level, no new water is being added to the Earth system regardless of what semiconductor companies (or any companies) are doing (Kral et al. 2024 point out there are hot debates over the origins of water on Earth, but those aren’t relevant for the similarly hot debates about net positive water). 100 percent of water already is accounted for somewhere on Earth, be it the oceans, lakes and rivers, ground water, the atmosphere, etc. This means, that in order to achieve net positive water companies such as Intel, TSMC, or Meta can only do so by accounting for debits of water (withdrawls) in one place or other with credits (deposits) somewhere else.

Where Intel funds water restoration projects it counts surpluses of water in a given location against deficits in other locations. The company claims, for example, that it achieves net positive water of 346 percent in India and counts that surplus against the deficits that it carries in all other regions of operations such as China, Costa Rica, Ireland, Israel, Malaysia, Mexico, Poland, and Vietnam (Intel Corporation 2023, 107–8; see also Vigliarolo 2022). Assuming Intel’s stated figure for net positive water in India is accurate, their claim is arithmetically correct but from a real use of water point of view not very meaningful. No amount of surplus water in India is going to help offset water withdrawals from facilities in Arizona or other parts of the desert southwest (a major hub of Intel’s operations) where a drought has been ongoing since 1994 (Arizona Department of Water Resources, n.d.).

TSMC is in a similar predicament. It’s latest sustainability report refers to the goal of being water positive. However, it’s own data on water consumption, water recycling and usage efficiency, and wastewater discharge tell a different story. By its own claims, the company’s water recycling and usage efficiency has improved from 86.7 percent in 2019 to 90.3 percent in 2023 (TSMC 2023, 114). Over the same time period, total water consumption increased between 2019 and 2021, then plateaued with a slight drop between 2022-2023,. However, the amount of water needed per semiconductor wafer manufactured increased from 133.4 L per 12 inch equivalent wafer to to 176.4 L per 12 inch equivalent wafer (an increase of 32 percent; TSMC 2023, 114; although TSMC’s report does not get into the details, the increase in water needed per wafer manufactured likely arises from the increased degree of rinsing with high purity water to achieve semiconductors with smaller and smaller etching patterns). Thus, even higher rates of water recycling and usage efficiency (90.3 percent) do not stop aggregate demand for water increasing. The increases seen in TSMC’s 2023 report do not include any of the water demand that will come online soon as a consequence of TSMC’s new facilities in Arizona. Rebound bites TSMC, just as it does Intel.

Meta/Facebook’s data published in its latest sustainability report suggests that its claims to net positive water are problematic in similar ways to Intel and TSMC. The company breaks down its measures of water into three categories: water withdrawal, water consumption, and water restoration (Meta 2024, 49). Water withdrawals or what is taken out of “local water utilities local aquifers” (Meta 2024, 93 of pdf). Water consumption is calculated as the difference between water withdrawal and waste water discharge at the company’s data centres. Meanwhile, water restoration refers to a variety of projects funded by the company that are claimed to increase water availability in given locations or regions (examples are offered in Meta 2024, 50–51). Before getting into debate about whether those restoration projects actually bring ‘new’ (or surplus) water back to the places where they operate, there’s a more fundamental aspect of Meta/Facebook’s own data to recognize: over the period between 2019 and 2023, water withdrawals, water consumption, and water restoration have all increased, but note that volume of water claimed to be restored remains substantially less than water withdrawn and consumed. The company’s own data show that water withdrawls and water consumption total 8,352 megalitres in 2023, whereas water claimed to be restored total 5,889 megalitres.

Meta claims of its water restoration projects that “[o]nce all projects are online and fully implemented, they will restore 1.9 billion gallons of water annually. In 2022, the operational restoration projects returned 621 million gallons (2,351,562 cubic meters) of water to high and medium water stressed regions.” (Meta 2023, 38). But the way Meta reports its data is super squishy. The company claims it “returned” X million gallons of water, but it’s own water consumption calculations say that the company defines water consumption as “[c]alculating the difference between water withdrawal and wastewater discharge” [p. 59] and that, “[a]ll of our wastewater is discharged to local wastewater facilities.” [p 59]. So is Meta counting its wastewater as ‘returned’ water?

Now, it could be that adding water withdrawals and water consumption is a double counting. It’s actually hard to know if that is the case from Meta/Facebook’s report. But, it is also hard to know whether the kinds of water restoration products the company highlights can or do lead to additional water entering the hydrological systems in which they are located. For example, one Meta project funds removing “invasive plants, conducting prescribed fires and planting 100,000 seedlings over the course of five to ten years” (Meta 2024, 50) and doing so, the report claims, means “increased water filtration is projected to provide over 200 million gallons per year.” (Meta 2024, 50). This is an ambiguous statement. Even if water filtration is improved, filtration is not the addition of ‘new’ water moving into and or through the system. So how to characterize that claim of 200 million gallons per year? Prior to the project were those 200 million gallons already moving into and through the watershed system in question? And, if so, even assuming they are better filtered now thanks to the project, that is not an addition of 200 million new gallons of water that weren’t there before the project. It’s also impossible to know from the report if Meta is counting these millions of gallons of water against water deficits elsewhere in its operations (i.e., in a different watershed from which the restoration project is located). If indeed the company is creating water surpluses in some locations but which are different from where its operations generate water deficits, creating those surpluses in other places doesn’t do anything for the places that go into water deficits due to Meta’s operations. The result of net positive water under that scenario is just accounting arithmetic, not hydrology.

At least three things are important about all this. First, we see the rebound effect at work with water demand for semiconductor manufacturing. At the same time, climate change is having a greater impact on water stress in the watersheds where semiconductor manufacturers are located (Lepawsky 2024). TSMC is explicit about this, stating in its latest sustainability report that the, “environmental impacts caused by climate change are becoming increasingly severe, and the demand for the cleanliness of water used in the advanced processes continues to rise.” (TSMC 2023, 111). Rebound in demand for water and increased water stress risk associated with climate change are on a collision course for semiconductor manufacturers.

Second, Meta’s reporting on its ostensible water positive projects tell us that it is not just hardware companies that are looking at the climate emergency and water issues associated with it with respect to their operations. Meta isn’t a hardware company, but Meta needs chips. Chips are the infrastructure for any and all software-based companies. This is one reason I find it clarifying to focus on what’s happening with semiconductor manufacturers. Even though that sector is massively complex, all things ‘tech’ otherwise run on that hardware. That makes semiconductors an obligatory passage point for ALL players in the tech sector.

Third, net positive water might be becoming an obligatory passage point for ESG reports, but in practice there are real questions about what net positive water can provide to actual ecologies. No amount of surplus water in a place disconnected from other places in water deficit will matter to the people, places, and things experiencing that water deficit. In some instances, disconnections can be overcome with infrastructure. For example, as Taiwan has experienced its drought TSMC and other semiconductor manufacturers resorted to trucking water from watersheds in surplus to those in deficit. Other forms of infrastructure such as aqueducts can be built to overcome disconnects between water surpluses and deficits (this has, of course, already been done in many cases). But, building that infrastructure is neither cheap, nor free of political struggle. Arizona, for example, is contemplating contracting for a desalination plant on the coast of the Sea of Cortez and a pipeline to bring the desalinated water to the state (Rose 2022; Water Infrastructure Finance Authority of Arizona 2022). At this point, the project is estimated to cost $5 billion. It is also imagined as a public – private partnership in which investment companies will own a century long contract on the project. Should that come to pass, that will mean that private equity will be able to extract rent from the provision of a basic human need–water–and, consequently, come to wield an important form of power over the people places and things that need that water.

References

Arizona Department of Water Resources. n.d. “Drought Frequently Asked Questions | Arizona Department of Water Resources.” Accessed June 20, 2023. https://new.azwater.gov/drought/faq#:~:text=A%3A%20Arizona%20has%20been%20in,according%20to%20statewide%20precipitation%20patterns.&text=Unlike%20hurricanes%20and%20tornadoes%2C%20drought,and%20long%2Dterm%20drought%20conditions.

D’Ignazio, Catherine, and Lauren F. Klein. 2020. Data Feminism. MIT Press. https://data-feminism.mitpress.mit.edu/.

Dorsch, Jeff. 2021. “Gigafabs: Why the Industry Is Turning to Fab Clusters | Electronics360.” August 2, 2021. https://electronics360.globalspec.com/article/17009/gigafabs-why-the-industry-is-turning-to-fab-clusters.

Fitch Ratings. 2021. “Taiwan Drought Highlights Water Stress as Growing Environmental Risk.” May 4, 2021. https://www.fitchratings.com/research/corporate-finance/taiwan-drought-highlights-water-stress-as-growing-environmental-risk-04-05-2021.

Intel Corporation. 2020. “Corporate Responsiblity at Intel | 2019-2020 Report.” https://csrreportbuilder.intel.com/PDFfiles/archived_reports/Intel-2019-20-CSR-Report.pdf.

———. 2023. “Intel 2022-23 Corporate Responsibility Report.” https://csrreportbuilder.intel.com/pdfbuilder/pdfs/CSR-2022-23-Full-Report.pdf.

Kral, Quentin, Paul Huet, Camille Bergez-Casalou, Philippe Thébault, Sébastien Charnoz, and Sonia Fornasier. 2024. “An Impact-Free Mechanism to Deliver Water to Terrestrial Planets and Exoplanets.” Astronomy & Astrophysics 692 (December):A70. https://doi.org/10.1051/0004-6361/202451263.

Latour, Bruno. 1999. Pandora’s Hope. Cambridge, Massachusetts: Harvard.

Lepawsky, Josh. 2024. “Climate Change Induced Water Stress and Future Semiconductor Supply Chain Risk.” iScience, January, 108791. https://doi.org/10.1016/j.isci.2024.108791.

Meta. 2023. “Meta 2023-Sustainability Report.” Meta/Facebook. https://sustainability.fb.com/wp-content/uploads/2023/07/Meta-2023-Sustainability-Report.pdf.

———. 2024. “Meta 2024 Sustainability Report.” https://sustainability.atmeta.com/wp-content/uploads/2024/08/Meta-2024-Sustainability-Report.pdf.

Narvaez, Liliana, Sally Janzen, Caitlyn Eberle, and Zita Sebesvari. 2021. “Taiwan Drought.” Technical Report. United Nations University. https://s3.eu-central-1.amazonaws.com/interconnectedrisks/reports/2022/Case-studies/TR_220830_TaiwanDrought.pdf.

Nguyen, C. Thi. 2024. “The Limits of Data.” Issues in Science and Technology 40 (2): 94–101. https://doi.org/10.58875/LUXD6515.

Reig, Paul, Wendy Larson, Samuel Vionnet, and Jean-Baptiste Bayart. 2019. “Volumetric Water Benefit Accounting (Vwba): A Method for Implementing and Valuing Water Stewardship Activities.” World Resources Institute. https://files.wri.org/d8/s3fs-public/volumetric-water-benefit-accounting.pdf.

Reuters. 2021. “Chipmakers in Drought-Hit Taiwan Order Water Trucks to Prepare for ‘the Worst.’” Reuters, February 24, 2021, sec. Autos. https://www.reuters.com/article/us-taiwan-drought-semiconductors-idUSKBN2AO0G3.

Robertson, Morgan. 2012. “Measurement and Alienation: Making a World of Ecosystem Services.” Transactions of the Institute of British Geographers 37 (3): 386–401. https://doi.org/10.1111/j.1475-5661.2011.00476.x.

Rose, Jordan R. 2022. “The Arizona Water Project Solution Proposal.” https://www.azwifa.gov/component/edocman/id20221211-technologies-wifa-final-proposal-presented/download?Itemid=0.

TSMC. 2022. “TSMC 2022 Sustainability Report.” https://esg.tsmc.com/download/file/2022_sustainabilityReport/english/e-all.pdf.

———. 2023. “TSMC Sustainability Report | 2023.” https://esg.tsmc.com/en-US/file/public/e-all_2023.pdf.

Vigliarolo, Brandon. 2022. “Intel’s Net Positive Water Use Only Tells Part of the Story.” The Register. July 13, 2022. https://www.theregister.com/2022/07/13/intels_net_positive_water_use/.

Water Infrastructure Finance Authority of Arizona. 2022. “WIFA Resolution 2023-027 – IDE Technologies’ Proposal.” https://www.azwifa.gov/component/edocman/approved-unsigned-wifa-resolution-2023-027/download?Itemid=0.