www.eetimes.eu, Jul. 25, 2024 –
As AI chip manufacturing grows to meet surging demand, its environmental impact becomes increasingly difficult to overlook.
AI chipmakers face a delicate balancing act, as how they proceed from here may determine whether AI helps combat climate change or intensifies it.
sustainable electronics
How AI improves global sustainability
With the development of AI, the world is gaining new and more powerful tools to support sustainability. Research suggests it could reduce global greenhouse gas emissions by 4% by 2030–equivalent to the combined annual total of Japan, Australia and Canada.
Much of this improvement stems from AI's impact on energy. Smart grids adapt to real-time changes in electrical supply and demand to prevent wasted power. Similarly, factories can use AI-enabled HVAC solutions to sustain production while using as little energy as possible.
AI's ability to highlight subtle trends in complex data and predict outcomes can also help. Businesses may use machine learning to find gaps in their sustainability initiatives and simulate changes to find effective solutions. For example, Nestlé has developed AI monitoring tools to assess the emissions of 34 farm groups to find opportunities for improvement. Amazon is using AI to redesign packages to use fewer raw materials. Similar solutions can apply to any business, and they're only possible if chipmakers provide sufficient supply.
How AI chip manufacturing threatens the environment
At the same time, AI takes a toll on the climate. While this technology can work wonders for some environmental initiatives, fabricating the necessary hardware requires considerable electricity.
Taiwan Semiconductor Manufacturing Company (TSMC), the world's largest semiconductor foundry, reported it had consumed a total of 22,400 GWh in energy, with purchased electricity accounting for about 94%, natural gases for 5.8% and diesel for 0.2% in 2022. TSMC is set to increase its electricity consumption by 267% and account for roughly one-fourth of Taiwan's population by 2030, according to Greenpeace's 2023 report titled "Invisible Emissions: A Forecast of Tech Supply Chain Emissions and Electricity Consumption by 2030." Not every foundry has an environmental toll so large, but this is just one example. The industry's projected growth also heightens these concerns.
The EU plans to account for 20% of global chip production by 2030. This shift would correlate with a 4× to 5× increase in the semiconductor sector's environmental impact. While transportation and end use account for some of that jump, most of the emissions linked to the lifecycle of electronic products come from the manufacturing phase.
The path to sustainable chip manufacturing
Given this impact, manufacturing must adopt new practices to become more sustainable. Here are a few strategies chipmakers can adopt to enable greener semiconductor production.
Understand existing and future energy needs
First, businesses must benchmark their current carbon footprints and estimate future demands. Optimization is only possible with a roadmap for improvement.
An energy consultant can help lower overall power costs by analyzing a fab's consumption patterns to highlight inefficiencies. These audits provide a baseline to set and measure future goals. Manufacturers can also use them to discover ways to reduce electricity consumption in the near term.
With this data in hand, chipmakers can model how much electricity their chip manufacturing processes will require in the future as production increases. These estimates are important because they show how much companies need to invest in efficiency-improving solutions to accommodate continued growth.
Embrace renewable energy
Chipmakers should also transition away from fossil fuels. Process improvements can reduce energy consumption, but rising AI's growing power needs are inevitable to an extent. Consequently, the only way to eliminate emissions is to use carbon-free electricity sources.
Manufacturers have several options when switching to renewable feedstocks. Solar and wind may be the most familiar examples, but solid biomass is the largest source, accounting for 40% of Europe's renewable energy in 2022. The most cost-effective solution depends on the facility's location, so companies must compare rates and availability to find the optimal path forward.
Some businesses may be unable to transition to 100% renewable energy at once, given the high upfront costs. In these cases, it's best to start with the most impactful use case and slowly expand from there. For example, a chip manufacturer may install enough solar panels to account for its most power-intensive process before investing in additional capacity.
Optimize chip manufacturing workflows
Renewables are a more viable option when chip manufacturers need less energy to run their facilities, as intermittency is less of a concern when consistent power needs are not as high. Reworking production steps to be more efficient will also help lower emissions before renewable capacity is sufficient for the entire facility.
Automating error-prone or time-consuming processes is a straightforward way to reduce consumption. However, not all robotic solutions are equally efficient. Chipmakers must ensure their automated equipment doesn't require more electricity than they save. This is often a matter of right-sizing tools and comparing multiple options.
Additive manufacturing deserves attention, too. It's now possible to 3D-print circuitry and other electronic components. This technology can reduce scrap by up to 90% and shorten the time it takes to create a part, resulting in less energy consumption per product.
Address supply chain concerns
Chipmakers must also consider their Scope 2 and 3 emissions. These secondary sources account for more than 70% of the total carbon footprint of many enterprises, so any sustainability measure that does not address them is incomplete.
Manufacturers can use technologies such as enterprise resource planning software and IoT tracking solutions to gain visibility into their supply chains. From there, they can work with suppliers and logistics providers to reduce all parties' emissions. Sourcing from other firms committed to long-term sustainability is vital.
Similarly, chipmakers should ensure ethical sourcing of materials such as rare Earth minerals. These resources often come from environmentally destructive processes, so they must take greater care in selecting their suppliers and holding them accountable.
Emphasize environmental AI use cases
Any environmental gains in production will be more impactful when AI's application promotes greater sustainability. Many leading chipmakers, such as Samsung and Intel, are prioritizing sustainability with plans to achieve net-zero emissions. By adding AI and automation to their operations, they can increase energy efficiency.
Chipmakers and OEMs can also put more pressure on their suppliers, using AI to improve traceability to reduce the environmental impacts of shipping.
They can also share emissions data with each other so all businesses can get a more holistic picture of emissions throughout the manufacturing and supply chain lifecycles. Combining this data with predictive AI can help chipmakers project future carbon emissions.
Increased collaboration will keep sustainability goals top-of-mind and help everyone stay on target.
AI chip manufacturing must adapt
The AI acceleration will impact the environment. Whether that effect is positive or negative depends on how chip manufacturers and their buyers proceed from here.
Sustainable manufacturing is possible, but it requires upfront investment and some initial disruption. While that may hinder some, organizations that embrace these changes can see significant improvements in the future.