Chip recycling: Addressing e-waste in the AI hardware industry

Do recycling programs for AI chips exist?

With a dwindling resource pool and environmental concerns about throwing out AI chips, the industry needs recycling infrastructure at the individual level and for places like data centers that consume these products in mass. But do such recycling programs exist?

Take-back programs exist for electronics in certain states and municipalities, but there's nothing specific to AI chips. Retail tech stores, such as Best Buy, may offer recycling programs for consumers, but data centers don't have many options. They can send old GPUs back to their vendors -- if they offer such a program -- or send them to a recycling center. Still, these tend to be few and far between, and there's no guarantee they're able to harvest the materials or components successfully -- let alone refurbish them for reuse.

Recycling faces considerable obstacles. For example, because semiconductors and microchips are so small, extracting recycling materials can be difficult, often requiring processes that add to pollution and emissions. And the amount of material that can be extracted is usually negligible. When those products are not recycled, however, they may end up in landfills or third-world e-waste recycling facilities, which may lack environmental safety protocols and even exploit child labor, according to the International Roadmap for Devices and Systems.

The industry needs to develop better techniques for separating semiconductors from AI chips, more advanced recycling technologies, superior electronics design that prioritizes recyclability and a more vigorous circular economy that puts eco-friendliness at the forefront of manufacturing.

How will the fast pace of the chip-making industry affect the environment?

Refurbishing AI chips is ideal, but it's often not feasible. The components degrade through use and over time, and they can't be entirely restored. A refurbished chip cannot deliver the level of performance that modern AI chips demand.

Chip's life span vs. demand

The chip industry is moving quickly, compounding the situation. A chip's life span is usually three to five years. Every year, however, companies produce new and more powerful, efficient and advanced chips, often at an even higher frequency.

The time it takes to create the components for new chips is lengthy. The process is complex and intricate. Chip fabrication plants create AI chip semiconductors, and these facilities consume a significant amount of water to develop ultrapure water to rinse silicon wafers. This process is time-consuming, intensive and wasteful. The demand for these components is outpacing supply, which is only exacerbating environmental concerns.

Use of nonrenewable materials

The electronic components that make up an AI chip include transistors, semiconductors, circuits, insulators, wires and various electrical connections. To create these components, manufacturers use essential materials, such as copper, gallium, germanium and silicon, as well as various other rare earth elements and critical minerals.

Many of these materials are finite -- as companies manufacture more chips, they further deplete these finite resources. Also, many of these chemicals, such as gallium, arsenic and selenium, are hazardous substances that risk harming humans and the environment. The manufacturing process also consumes electricity, water and labor, and it generates waste that strains the environment without proper management.

Less than one-quarter of e-waste mass was properly recycled, and just 1% of rare earth element demand was met by e-waste recycling.

Increase e-waste

According to United Nations research, the world's e-waste generation is rising five times faster than documented e-cycling. A record 62 million tons of e-waste was produced in 2022, up 82% from 2010, and projections estimated this amount to rise another 32% by 2030, up to 82 million tons. Less than one-quarter of e-waste mass was properly recycled, and just 1% of rare earth element demand was met by e-waste recycling.

Without monitoring and change, e-waste will rise to new heights, and the supply chain will struggle even more to meet the demand for chip components. In turn, the negative impacts of chip production and manufacturing will only increase.

Are there green ways to construct AI chips?

Though difficult, there are paths ahead that can make AI chip manufacturing greener.

Develop automated robotic recycling technologies

AI chips are often manually recycled, which introduces the risk of exposure to toxic chemicals and human error. Robotics eliminates those risks and increases success rates of material harvesting.

Use sensors to gather data about component recycling

Sensors help compile recycling data and analyze it to discover which parts last the longest and which components are prone to failure. Results can inform future manufacturing.

Decarbonize AI chip manufacturing

Adopting renewables; transitioning from burning fossil fuels, like wood and coal; and using less water help lower the industry's greenhouse gas emissions.

Replace rare earth elements, minerals and chemicals with alternatives

Investing in R&D is vital to finding alternative materials for AI chips that are eco-friendly and less of a strain on the supply chain.

Create devices that can separate semiconductor packages from AI chips

Semiconductors, especially, are in rare supply and difficult to produce. By developing removable semiconductor packages, manufacturers could more easily recycle these components.

Sustainably source materials and alternatives

Countries like China are currently the dominant source of rare earth elements and other critical minerals found in AI chips. This reliance can bottleneck supply, so manufacturers should look to source materials locally and switch to sustainable alternatives where possible.

Adopt open-loop recycling and the circular economy at large

The industry must think of AI chip design and the production process with recycling in mind. This will lead to more thoughtful use of materials, an interconnected open-loop recycling system and a circular economy that reuses and recycles all parts.

The industry must carefully navigate the road ahead for AI chip development. Hardware companies must sustainably source materials to protect finite resources, as well as give more thought to manufacturing and retiring old chips in environmentally friendly ways. The industry must keep these considerations at the forefront to hold itself accountable and ensure a sustainable future until regulations catch up to the rapidly accelerating market.

Jacob Roundy is a freelance writer and editor with more than a decade of experience with specializing in a variety of technology topics, such as data centers, business intelligence, AI/ML, climate change and sustainability. His writing focuses on demystifying tech, tracking trends in the industry, and providing practical guidance to IT leaders and administrators.