Micron Is Building a $24 Billion Semiconductor Plant in Singapore — What Chip Manufacturing's Global Footprint Means for the Environment

In January 2026, Micron Technology announced it would invest approximately $24 billion to build a new advanced wafer fabrication facility in Singapore, one of the largest single semiconductor investments the company has ever made and a clear sign of how aggressively the chip industry is racing to keep pace with artificial intelligence's appetite for memory. The new plant will add 700,000 square feet of cleanroom space to Micron's existing NAND manufacturing complex, become Singapore's first double-storey wafer fabrication facility, and begin production in the second half of 2028. It is, by any measure, an enormous undertaking. It is also just one piece of a much larger global wave of semiconductor construction that includes TSMC in Arizona, Samsung in Texas, and dozens of other facilities being built or expanded simultaneously around the world, all chasing the same explosive AI-driven demand. That wave raises a question that gets far less attention than the impressive dollar figures and job creation numbers typically attached to these announcements: what does building and running this many chip factories actually cost the planet.

What a Modern Chip Fab Actually Consumes

Semiconductor fabrication is, at its core, an extraordinarily resource-intensive process, and understanding why requires appreciating just how exact and unforgiving chipmaking has become. Producing a modern microchip involves hundreds of individual processing steps, each requiring an environment virtually free of dust, contaminants and impurities, since even a single stray particle can ruin a wafer worth tens of thousands of dollars. Maintaining that environment, and the chemical processes layered on top of it, requires staggering volumes of water, energy and specialized chemicals.

Water sits at the center of this resource demand. According to researchers at Arizona State University, water is actually the largest-volume chemical used in semiconductor manufacturing, with prior research finding that production can require up to 10 million gallons of ultrapure water per day at a single facility, a grade of water purified to remove virtually all impurities before it ever touches a silicon wafer. Energy consumption follows a similarly extreme pattern. A single semiconductor fabrication plant can use as much electricity as a city with hundreds of thousands of residents, driven by the power demands of high-precision machinery, continuous air filtration systems, and intensive chemical processing that must run nearly around the clock. TSMC's planned Phoenix, Arizona complex alone is expected to use enough electricity to power roughly 300,000 homes. Beyond water and energy, fabs rely on an extensive list of specialized industrial chemicals, including highly corrosive acids and toxic gases used in etching and cleaning processes, materials that require careful handling, treatment and disposal to avoid contaminating local air, soil and water supplies.

A Global Buildout Driven by AI

What makes today's chip manufacturing expansion historically unusual is not simply its scale but its near-simultaneous occurrence across multiple continents, all driven by the same underlying force. Micron's Singapore investment builds on an earlier $7 billion commitment to a high-bandwidth memory packaging facility in the same country, and Singapore now produces 98 percent of Micron's flash memory chips, making the city-state one of the most concentrated semiconductor manufacturing hubs on Earth. Across the Pacific, TSMC has committed roughly $65 billion to its north Phoenix campus in Arizona, with reports suggesting the company may eventually build as many as twelve fabs and four packaging facilities at that single site. Samsung, meanwhile, has expanded its Taylor, Texas campus from an initial $17 billion commitment to as much as $44 billion, with a second fab reportedly being planned on the same 1,268-acre site to help meet demand for AI accelerators and automotive chips, including a deal to produce Tesla's next-generation AI6 chip.

The driver behind all of this construction is consistent and explicit across every company involved: artificial intelligence has created a level of demand for advanced memory and logic chips that existing global manufacturing capacity simply cannot meet. Micron's own leadership has stated the company can currently fulfill only roughly half to two-thirds of customer demand for its products in the medium term, a supply gap so severe that analysts have described the resulting capital expenditure race across Micron, Samsung, SK Hynix and TSMC as unlike anything the industry has experienced in recent memory. Every one of these companies is making the same calculation: the AI boom has created a once-in-a-generation opportunity to expand capacity and capture demand, and the environmental cost of doing so has, in nearly every public announcement, been treated as a secondary consideration to be managed rather than a central constraint on how fast and how large these projects can grow.

The Environmental Footprint at Scale

When this kind of construction happens simultaneously across multiple sites, the cumulative environmental footprint becomes considerably harder to wave away as a manageable, isolated cost. Research published in the journal PNAS examining climate-induced water stress found that if all of TSMC's newly announced Arizona facilities are eventually built out fully, they are estimated to require up to 40,000 acre-feet of water per year. Arizona's broader semiconductor cluster sits squarely within a region already experiencing severe, climate-driven water stress, with the US Bureau of Reclamation having declared its first-ever water shortage for the Colorado River basin in 2021. Research using a water-stress-weighted environmental impact metric found that Intel's Arizona fabs, despite consuming less total water than some of the company's other US facilities in Oregon, actually carry a much higher adjusted environmental impact precisely because of how severely water-stressed the surrounding region already is, a finding that underscores how the same volume of water can represent very different levels of environmental burden depending on where it is drawn from.

The scale of this challenge is also growing. Industry analysis from IDTechEx projects that water usage in semiconductor manufacturing will double by 2035, driven by the increasing complexity of advanced chip production, which requires more processing steps and increasingly favors single-wafer tools over the batch processing methods that were historically more water-efficient. Energy demand follows a comparable trajectory, compounding an AI industry that is already placing unprecedented strain on electrical grids through data center construction, meaning the chips powering AI and the factories that produce those chips are simultaneously driving two separate, enormous waves of new energy demand at the same moment in history.

What Chipmakers Are Actually Doing About It

To their credit, the major chipmakers building this new wave of capacity have not ignored the environmental dimension entirely, and several have made genuine, measurable investments in mitigation. TSMC has built an extensive water recycling program, achieving a water recycling rate of over 80 percent at some of its facilities, and the company is constructing a 15-acre Industrial Water Reclamation Plant at its Phoenix complex, scheduled to begin operations by 2028 and designed to recycle 90 percent or more of the water used on site. TSMC has also become the first semiconductor company to join RE100, a global corporate commitment to sourcing 100 percent renewable electricity, and the company has set a companywide goal of reaching net zero emissions by 2050, supported by measures including waste heat recovery, variable frequency drives, and temperature-optimized cooling systems designed to reduce the energy intensity of its manufacturing processes.

Other companies have pursued comparable strategies suited to their specific locations. Intel's facility in Chandler, Arizona has, according to water sustainability researchers, actually managed to purify and restore more water to the surrounding community than it consumes on site, partnering closely with local water utilities and exceeding the restoration standards required by local sewer ordinances. Tower Semiconductor has experimented with capturing condensate from the dehumidification systems required to maintain dry indoor manufacturing environments, while Taiwan-based companies including SMIC have combined rainwater collection with air conditioning condensate recovery to reduce reliance on municipal water supplies. Some companies located near coastlines, including TSMC's facility in Hsinchu, Taiwan, have turned to seawater desalination as an alternative water source, though this approach trades reduced freshwater consumption for substantially higher energy use, illustrating how thoroughly water and energy challenges in this industry are intertwined rather than separable problems with independent solutions.

Can the Industry Decarbonize Fast Enough to Meet Climate Targets

This is the question that ultimately matters most, and the honest answer requires holding two things in tension rather than collapsing the issue into easy optimism or easy condemnation. On one hand, the mitigation efforts described above are genuine, technically sophisticated, and represent real reductions relative to what an unmanaged buildout of this scale would otherwise produce. Companies committing to net zero targets, joining renewable energy procurement initiatives, and building dedicated, expensive water reclamation infrastructure are not making symbolic gestures. These are substantial capital investments that meaningfully reduce the per-chip environmental footprint of modern semiconductor manufacturing compared to a decade ago.

On the other hand, efficiency improvements per chip do not automatically translate into a smaller total environmental footprint when the sheer volume of chip production is expanding as rapidly as it currently is. A facility that recycles 90 percent of its water is still drawing an enormous absolute volume of fresh water if the facility itself is large enough, and a fab built with industry-leading energy efficiency standards still represents a substantial new addition to regional electricity demand if it is one of a dozen similar facilities being constructed simultaneously across the same water-stressed region. Civil society groups, including the Coalition for Clean Air in Arizona, have explicitly called on government regulators to conduct full environmental impact statements for projects like TSMC's and Intel's Arizona expansions, arguing that claims these projects will have no significant environmental impact do not hold up to scrutiny once their full water, energy and chemical release implications are accounted for at scale.

The most realistic assessment is that the semiconductor industry is decarbonizing and improving its resource efficiency in real, measurable ways, while simultaneously expanding production at a pace fast enough that total resource consumption and emissions are likely still rising in absolute terms, even as per-unit efficiency improves. Whether that combination is compatible with global climate targets depends heavily on factors outside any single chipmaker's control, including how quickly regional electricity grids can decarbonize their own power generation, how effectively water-stressed regions like Arizona and Taiwan can manage increasingly competing demands from agriculture, residential use and industrial expansion simultaneously, and whether the extraordinary pace of AI-driven chip demand growth itself proves durable or eventually moderates as the technology matures.

The Bottom Line

Micron's $24 billion Singapore expansion is, on its own, a single data point in a much larger story about how thoroughly artificial intelligence has reshaped the economics, geography and environmental calculus of global chip manufacturing. Multiply that single project by the comparable, simultaneous expansions underway at TSMC in Arizona, Samsung in Texas, and dozens of other facilities worldwide, and the cumulative environmental footprint of this buildout becomes a genuinely significant factor in the broader climate equation, not a rounding error attached to an otherwise purely economic and technological success story. The chipmakers driving this expansion have made real commitments to water recycling, renewable energy and emissions reduction, and those commitments deserve genuine credit rather than reflexive skepticism. But credit for effort is not the same as confirmation that the math actually works out in the planet's favor once the sheer scale of simultaneous global expansion is accounted for. Whether the world can build the chips artificial intelligence demands without meaningfully undermining its broader climate commitments remains, at this stage, an open question that the industry's next several years of construction and operation will answer far more definitively than any single sustainability report or groundbreaking ceremony can on its own.

*This article is for informational purposes only. Data and research findings are sourced from CNBC, Mothership.sg, CRN Asia, the journal PNAS, IDTechEx, Manufacturing Dive, ScienceDirect, and TSMC's and Samsung's public sustainability and corporate disclosures as of June 2026.*