Semiconductor sustainability: Chips take a smaller byte out of resources
Modern, new greenfield plants could help improve the industry scorecard, but manufacturing transformation can help both the greenfield plants and existing brownfield plants do better for energy, water, and process gas use.
Duncan Stewart
Karthik Ramachandran
Dr. Bobby Mitra
Christie Simons
The semiconductor manufacturing industry is notoriously volatile: As of fall 2023, it is in the middle of the seventh downturn since 1990.1 Measured in US dollars, chip industry revenue is expected to decline 10% in 2023 and is expected to rise 12% in 2024.2 Maybe unsurprisingly, when the chip industry shrinks, its use of energy, water, and process gases with high global warming potential (GWP), all go down. And when the industry grows, GWP goes up. Absolute measures of sustainability are often unhelpful, especially in an industry that—despite its volatility—is expected to grow, and is predicted to surpass US$1 trillion in 2030,3 almost double the anticipated industry revenues of US$515 billion in 2023.
Instead, a better yardstick may be resource intensity: For every dollar of revenue, how much energy, how much water, and how much high GWP process gases are going to be used next year, compared with this year? Deloitte predicts that there will be a year-over-year decline in average water intensity (figure 1) and energy intensity in 2024, as well as declines in energy intensity and growth in the percentage of energy used by leading chipmakers that is renewable.
Some of the improvement in resource use intensity is likely part of an ongoing, decade-long trend as the industry has recently been trying to get better at semiconductor sustainability. Some of it may come from a significant growth in brand new chip plants (greenfields). As might be expected, the various equipment, tools, and processes in a new plant (all other things being equal) are often more sustainable than equivalent technologies that might be five, 10, or even 20 years old. That said, newer plants using advanced node technologies pose a sustainability challenge for the industry: Moving from a mature technology such as 28 nm manufacturing to advanced node manufacturing at 2 nm needs 3.5 times as much energy, 2.3 times as much water, and emits 2.5 times as much greenhouse gases, and this trend is expected to continue as processes become ever more advanced.4 Interestingly, even more of industry progress towards greater sustainability will come from implementing manufacturing transformation on those older plants (brownfields): Deloitte predicts that a full manufacturing transformation project can significantly reduce the intensity of energy, water, and process gas use over a multiyear period.
But there’s still more to do
Looking at the lifetime energy and resource use of chips, manufacturing is only a part of the challenge. And the energy used by chips after they have been manufactured (for example, in power hungry data centers doing generative AI) can be a material factor. Equally, resource extraction, test and packaging, distribution, life cycle, and end of life are all important parts of the semiconductor sustainability equation.
While some chipmakers have set aggressive 2030 carbon zero and other sustainability targets, there is wide variation globally. In general, European Union-headquartered companies tend to have the most aggressive 2030 targets, and while some US-headquartered organizations have similarly ambitious targets, others are aiming for 2040 or later.5 Outside of Singapore, most Asian-headquartered chipmakers are setting targets for 2050 and beyond, or not at all.6 That said, in September of 2023, a leading Asian chipmaker moved up its commitment to use 100% renewable energy by 10 years, to 2040.7
The chip industry was responsible for approximately 0.2% of global carbon dioxide equivalent emissions in 2021.8 If it doesn’t want to double to 0.4% by 2030 as the industry doubles in size, it should improve both greenfield plants and transform brownfield plants.
Energy consumption
Making chips is energy intensive. Melting silicon, lithographing with high-powered lasers, creating and maintaining vacuums, and endless cleaning takes a lot of electricity. Semi fabs can use up to 100 megawatt-hours per hour,9 the equivalent of over 80,000 typical North American homes. With that said, there are only about 500 open fabs at present.10 Although there’s new plant construction between now and 2025, it’s only 41 fabs globally.11 Moreover, semiconductor companies have applied novel chip design techniques and advanced process technologies, for instance, using low-leakage transistors and low-power systems, and altering system power modes (such as off, idle, drowsy when select modules or IPs are not active). These have helped reduce energy needs across end-use devices and systems across industries, but with the expanding manufacturing footprint, the industry should look at other ways to optimize the use of resources and lower emissions.
The industry has two levers to pull in terms of reducing their energy use and associated carbon footprint. First, they’re trying to become more energy efficient, but progress in that is often slow, especially as chip companies resort to increasingly advanced manufacturing technologies to push the limits of semiconductor manufacturing.12
One thing they can consider is to make things (in addition to the actual chip manufacturing process) more energy efficient, faster: Leadership in Energy and Environmental Design buildings are mature technologies that the industry has relied on for a decade to improve its sustainability.13 They can also look to increase their use of renewable energy. As an example, 93% of a major America-headquartered chipmaker’s energy needs were met by renewables as of its 2022–23 fiscal year.14 However, across the three largest players, renewables were only 28% of the energy mix in 2022, up five percentage points over 2021.15
Combining both levers has led to the fab portion of the industry reducing its energy intensity (watt- hours/dollar) from almost 240 in 2020 to 219 in 2022 and a predicted 206 in 2024 (figure 2).16 Further, the percentage of energy used that was renewable is predicted to rise even faster than energy intensity falls: Renewable energy is predicted to be 28% of the mix by 2024, double the percentage from 2020.
Water “usage”
The global chip industry used 264 billion gallons (about 1 trillion liters) of water in 2019.17 But although some water is lost to evaporation or other causes, depending on geography and chipmaker, all of that water is not “used”: A major America-headquartered chipmaker used 16 billion gallons of water in 2021 but returned 13 billion gallons of it (over 80%) and doubled their water savings from two years earlier.18 Chipmakers across Taiwan averaged 85% water recycling between 2016 and 2020.19
Most water use in the semiconductor industry is for the manufacturing process (76%), but a lot of water is also used for cooling towers (9%) and scrubbers (11%).20 The biggest part of those scrubbers is in process gas abatement (see next section), which could present significant opportunities for water reduction: Switching the abatement systems to idle mode, when not actively processing, reduces water usage by 98%.21 Improvements could also be made in reducing both process water and water for cooling.
Process gases
The chip industry uses several gases, some of which have very high GWP. These are primarily fluorinated gases, namely perfluorocarbons (PFCs), hydrofluorocarbons, nitrogen trifluoride (NF3), and sulfur hexafluoride (SF6), used both in etching and cleaning,22 and nitrogen oxide gas that’s used in deposition and purging processes.23 As an example, SF6 has a GWP 23, 500 times higher than CO2.24 According to the US Environmental Protection Agency, historically, “anywhere between 10 and 80 percent of the fluorinated greenhouse gases pass through the manufacturing tool chambers unreacted and are released into the air.”25
There are three main avenues for reducing the impact of these gases: process improvements/source reduction, alternative chemicals, and destruction technologies (often referred to as ‘abatement’).26 In general, much of the low hanging fruit opportunities for improvement in the first category have likely been made, but progress continues to be made at the margins, and this is where manufacturing transformation can potentially have an impact. There have been some successes in finding alternatives: A number of PFCs were replaced by NF3, which, while still problematic, was an improvement.27 However, finding and qualifying for the manufacturing processes alternative gases can be a slow process, and only a few breakthroughs appear imminent, such as G1.28 Abatement remains the workhorse here. Capturing and destroying (usually by burning or conversion) as much of the high GWP process gases is the key. As an example, abating 99% of NF3 is likely both achievable and better than 95%.29 By and large, process gases are not often reused or recycled, due to issues around purity, cost, and ability to be integrated into the sub-fab physical footprint.30
PFAS
Although not the focus of this prediction, the chip industry also uses or produces many perfluoroalkyl and polyfluoroalkyl substances (PFAS). In 2023, a restriction proposal on their use was made by the chemical authorities in five EU countries, calling for their eventual elimination. This is expected to come into force in 2025–2026.31 The semiconductor industry association SEMI, in response to both the EU proposal and various existing or proposed US regulations, is looking at reducing the use of PFAS and finding alternatives.32
Manufacturing transformation
Today’s chip fabs are like a forest with trunks, leaf canopies, and vast roots. In addition to all the machines and clean rooms at surface level, there’s a canopy of pipes and ducts overhead, and an even more complex set of pumps, abatement systems, scrubbers, and transformers underneath the floor (the sub fab). This ecosystem has many parts that can be hard to access or monitor in real time, and by modeling, adding connected sensors everywhere, and continuously monitoring the use of energy, the use of water and process gases can be made more efficient. Enabling technologies include digital twins, generative AI, and private 5G networks. Leaks can be detected, and systems can be idled or powered down when not in use. It can cost hundreds of millions of dollars to take a decade-old chip plant and transform it, but the gains in sustainability, as well as lower costs and higher efficiency are likely worth it both for the bottom line and the planet.
In general, this idea is integral to the 6D building information modeling (BIM) concept, which encompasses the inclusion of a sixth dimension, specifically the performance and sustainability of the building. BIM is the process of creating and managing information on a construction process, and by incorporating this additional dimension, the virtual model is designed to depict not just the physical, temporal, and cost-related aspects of the building, but also its environmental and social implications. Consequently, this could become a valuable resource for evaluating how the building affects its surroundings and community, and for recognizing opportunities for enhancement.33
Gross versus net: Chips are greener than you think
Making a trillion dollars’ worth of chips in 2030 is going to have an environmental impact in terms of energy, water, and greenhouse gases. The industry is working to make that impact as low as possible, but it may be worth noting that chips could help enable sustainability gains. Road and air travel can be avoided by using video conference technology enabled by chips, computers are the standard tool for discovering and designing drugs,34 hyperscalers are investing in more renewable energy sources to power chips in data centers,35 and food can be grown more sustainably (see our AgTech prediction). Some might argue that the environmental benefits of chips are greater than the environmental costs of making semiconductors. On the other hand, others believe that: 1) This might be true of chipmaking itself, but factoring in resource extraction, test and packaging, distribution, life cycle and end-of-life considerations paints a less beneficial picture;36 and 2) The Jevons paradox (aka rebound effect) could lead to worse sustainability outcomes. If chips get made more efficiently, we will just make and use more of them, potentially leading to a worse overall outcome.37
The bottom line
For semiconductor companies, environmental awareness is its own reward, being more sustainable is good, and is increasingly being required by what is sometimes called the 5Cs framework: capital (investors,) compliance (regulators,) constituents (such as employees), communities, and creativity (innovation). But being more sustainable is often also better for reducing costs, can help in the competition for semiconductor talent, and can reduce semi supply chain vulnerabilities.
Dedicated environmental, social, and governance (ESG) funds are US$8 trillion today and are predicted to be as much as US$30 trillion by 2030.38 Even outside of dedicated funds, asset managers are increasingly using ESG screening tools when building their portfolios, and this includes chipmakers. One additional challenge is from regulators. Currently, most public companies report scope 1 and scope 2 emissions (direct and indirect energy use) but not scope 3 (the supply chain, both upstream and downstream). It’s possible that regulators, both in Europe and the United States, may require scope 3 disclosures. As is well known after the pandemic chip shortage, almost every industry has chips in their supply chain, and chipmakers would likely be asked by the customers to have the best sustainability profile possible.
Energy, gases, and (usually to a lesser extent) water are expensive, and becoming more so. Reducing these input costs can have a positive effect on the bottom line.
Equally, semiconductor companies are attempting to build greenfield plants both in traditional strongholds such as Asia, but increasingly in the United States and Europe.39 And there is a global competition for talent: The semi industry is competing for scarce technically skilled talent with multiple other industries and should have a positive environmental track record.40 Workers, especially younger workers, prefer working for companies with the best sustainability track records. As a 2023 Deloitte survey of Gen Z and Millennials found, “one in six have already changed jobs or industries due to climate concerns, with another quarter planning to do so in the future.”41
Finally, reducing reliance on energy and water significantly expands the areas where chip plants can be located. Droughts have recently affected chipmaking in multiple regions, and as one headline put it, “No water, no microchips.”42 Equally, chipmakers in Asia and the United States were subject to power outages due to climate change, and chip plants are exceptionally reliant on uninterrupted power.43 Raw material dependencies (see prediction on raw material and supply chains) and transport disruptions and investing in semiconductor sustainability can offer material supply chain resilience benefits.