Fixing chipmaking’s dirty climate secret 

The Chiayi-Tainan Plain, traditionally one of Taiwan’s key agricultural regions, remained largely unplanted this spring, marking the region’s third consecutive year without a full harvest. After another season of low rainfall, local authorities offered farmers subsidies to hold off on planting their rice. With water in scarce supply, Taiwan prioritised its prized semiconductor plants — the country’s so-called ‘silicon shield‘ against the threat of an invasion by Beijing. Nearby, once-abundant rice paddies looked barren: cracked brown earth interspersed with a few wilting blades of grass.

Semiconductor manufacturing, or chipmaking, is an extremely thirsty business. According to the company’s latest sustainability report, Taiwan’s semiconductor vanguard, TSMC, used 82.8 million metric tonnes of water to make 14.2 million semiconductor wafers in 2021. That’s around 5,840 litres — or nineteen fully-brimming bathtubs — per chip, although the water used to produce some of the most-advanced chips at TSMC’s future plant in Arizona is estimated to reach up to three times that figure. This immense sum puts huge pressure on drought-struck regions in Taiwan, which produces more than 60% of the world’s semiconductors, including 90% of the most-advanced ones.

Semiconductors power everything from iPhones to EVs. They’ll be essential for the green transition, enabling renewable energy technologies and optimising energy efficiency. Even so, their production poses ecological obstacles of its own, extending far beyond its immense thirst. Semiconductor manufacturing is projected to consume 237 terawatt hours of electricity globally by 2030, according to Greenpeace East Asia. That’s roughly equivalent to Australia’s total electricity consumption in 2021. The sector’s also on track to emit 86 million tons of CO2 equivalent in 2030 — more than twice Portugal’s total emissions in 2021.

As geopolitical tensions push nations to reconsider their offshore semiconductor supply chains, there are also mounting sustainability pressures facing the industry. Some of the sector’s biggest clients — including giants like Apple, Microsoft, and Google — have set ambitious net-zero targets that, according to analysis by McKinsey, will require major reform in chipmaking, which remains one of the biggest contributors to their corporate carbon footprints. There’s broad recognition that production needs to go green, but less consensus on what — and how much — manufacturing companies should be doing. Are they really reaching for the (sustainability) stars?

Cutting chipmaking’s carbon costs

Some chipmakers have answered the call for a green revolution by setting their own targets. German manufacturer Infineon plans to reduce greenhouse-gas emissions by 70 percent by 2025, compared with its 2019 baseline, and aspires to reach carbon neutrality for emissions directly under its control by the end of 2030. US chipmaking giant Intel has committed, as of 2022, to net-zero greenhouse gas emissions in its global operations by 2040 and has targeted achieving 100% use of renewable electricity as an interim milestone in 2030. TSMC, the world’s largest chipmaker, has, meanwhile, pledged to reach net-zero emissions by 2050. In 2020, the company signed the world’s largest renewable corporate power deal, agreeing to buy up the full output of a 920 megawatt offshore wind farm in the Taiwan Strait.

Gas abatement can solve some, if not all, of the chipmaking climate dilemma. Common semiconductor manufacturing processes, including wafer etching and chamber cleaning, rely on highly toxic greenhouse gases, including nitrogen, argon, and perfluorocarbons (PFCs) — greenhouse gases with a considerably higher global warming potential than CO₂. These were Intel’s primary source of direct emissions in 2021 and are a key focus of the company’s sustainability strategy. Through gas abatement — which involves changing the chemical compositions of toxic gases before they’re released into the environment — manufacturers hope to minimise, if not entirely eliminate, the noxious warming effects.

But more ambitious than abatement, some industry players, including Intel, are trying to find ways to switch out these toxic gases entirely. This, says Ondrej Burkacky, a semiconductor analyst at McKinsey, “is often technically tricky because we are operating at very limits of physics.” Nevertheless, he says he’s cheered by collaborative efforts, like the SEMI Sustainability Initiative, which are uniting leading industrial figures, chemists, and gas suppliers to fund research into alternative chemical compounds for chipmaking.

TSMC, meanwhile, has been working to serve its plants’ insatiable thirst without totally depleting Taiwan’s resources. Reclaimed water is expected to replace over 60% of the total water consumption at TSMC’s fabs in Taiwan by 2030, according to Global Water Intelligence. Moreover, TSMC’s first water-recycling plant in the Southern Taiwan Science Park went into operation in September 2022. It is currently capable of supplying the chip plant’s daily water requirements of 10,000 tonnes, and capacity could grow to 36,000 tonnes daily by 2026. “Circularity can be, and should be, improved,” says Burkacky. That’s not just a question of sustainability, but also sheer practicality. If you don’t have enough water at any given time, you’ll have to shut the plant down, he says. “If you shut down a chemical plant, that means two-to-three weeks to bring it back up.”

Beyond silicon

Nevertheless, there’s still a long way to go to achieve a green chip supply chain, and sustainable developments might be offset by increased demand for more sophisticated semiconductors. Indeed, as it ramped up chip output to meet rising demand in 2020, TSMC missed key sustainability targets for water usage and waste generation.

There’s still an inherent tension in the sector between technical progress and environmental sustainability, according to Scott White, who founded chipmaking company Pragmatic Semiconductor in Cambridge in 2010. “Most of the industry is focused on innovating in the direction of better performance [and] more complexity,” he says. “It’s actually been going the wrong way from an environmental perspective because they’ve been trying to continually enable more functionality.”

And the biggest environmental problem might actually be silicon itself, says White. His company, Pragmatic, aims to eliminate silicon as the basic material for semiconductor devices, instead employing flexible integrated circuits (FlexICs) based on thin-film transistors. These materials, explains White, aren’t comparable to high-end semiconductors (“It definitely will not replace all silicon,” he says), but could replicate the performance of mature-node or legacy-node chips.

Pragmatic’s chips might not be sophisticated enough to completely power our iPhones or EVs, but White argues that cutting the emissions of the basic tools could have a big impact. “The vast majority of chips actually do much simpler functions,” says White. Modern-day cars tend to have upwards of 1,400 chips inside them. “Only a couple of those are the engine controller and things that need to be high-performance,” says White. “Most of the chips are doing very simple tasks, like making the windscreen wipers work.”

‘By using flexible semiconductors for the myriad tasks where ‘just enough’ performance is more than enough, we can free silicon – with its weightier environmental overheads – to tackle the big-ticket items,’ argues Pragmatic’s latest sustainability report. Nevertheless, moving beyond silicon on a broader scale will still take time, says Burkacky.  “This is something which not only [needs to] work in a lab, it needs to work in a highly-industrialised environment at a certain cost level.”

We might, however, have reached a pivotal moment for industrial awareness and development, says White. Eco-friendly improvements are being driven not just by consumer and government pressure, but also straightforward practicalities, like the limited water supplies in the Chiayi-Tainan Plain.

“As always, it could go faster and it could go better,” says Burkacky. “But I think what makes me positive is that there is really this demand from customers.” He’s also optimistic that future chip iterations will keep facilitating more energy-efficient machines — even if the manufacturing industry has a long way to go to clean up its act. “I’m not worried that more semiconductors means more problems,” says Burkacky.