What happened
The nonprofit Basel Action Network (BAN) published a report estimating that AI-related electronic waste will total between 395 and 617 million metric tons by 2050. This projection is substantially higher than prior studies because it includes the full data center infrastructure, such as cooling, power, and networking equipment, rather than just servers and accelerators.
Basel Action Network (BAN) released a new report warning that e-waste from the AI boom has been vastly underestimated. The organization projects that by 2050, AI-related e-waste could amount to 395 to 617 million metric tons, equivalent to filling 23 million shipping containers.
The report’s estimates are significantly higher than previous studies because BAN expanded its scope to include all infrastructure supporting data center servers. This includes power supply and distribution, cooling systems, backup power, and networking equipment, which previous analyses often omitted.
BAN also introduced the concept of 'AI Waste Contagion,' which accounts for telecommunications infrastructure and personal devices that may become obsolete earlier due to AI advancements. The organization estimates that 70,000 metric tons of e-waste are generated per gigawatt of data center capacity.
The report cites a McKinsey projection that total data center capacity could reach 219 gigawatts by 2030. Based on this, BAN calculates that AI-related equipment retired between 2025 and 2050 will result in 8.6 to 13.1 million metric tons of new e-waste annually.
Source details: theverge.com ↗
Why it matters
The report highlights a critical environmental blind spot in the AI industry's rapid expansion. By including 'AI Waste Contagion' and supporting infrastructure, the study reveals that current recycling systems are ill-equipped to handle the scale of upcoming waste. This has significant implications for regulatory policy and corporate responsibility, as less than a quarter of global e-waste is currently formally recycled, leading to toxic exposure in informal waste sectors.
Current global e-waste management is inadequate, with less than 25% of the 68.3 million tons generated annually being formally collected and recycled. The majority ends up in informal waste collection systems where burning or burying equipment exposes workers and environments to toxic materials like lead and chromium.
The United States, which hosts the most data centers, has not ratified the Basel Convention, a treaty designed to limit the international trade of hazardous wastes. Investigations have shown that US recyclers often ship e-waste abroad, where it frequently enters informal recycling streams that threaten the health of local communities, including children.
Previous estimates, such as a 2024 study predicting 1.2 to 5 million tons of AI e-waste by 2030, focused narrowly on servers and GPUs. BAN argues this approach misses approximately 87% of a data center's electro-mechanical infrastructure, leading to a significant underestimation of the environmental impact.
The report serves as a call to action for companies and governments to begin planning for this 'waste tsunami.' Without proactive measures, the current AI buildout could exacerbate an existing toxic waste crisis, creating long-term environmental and health liabilities.
What to watch next
Regulatory responses to the report, particularly regarding the Basel Convention and hazardous waste trade. Additionally, watch for industry initiatives to design data center hardware for easier recycling or to extend the lifespan of specialized AI components.
Legislative and regulatory actions in the US and other major AI hubs regarding hazardous waste management and the ratification of international treaties like the Basel Convention.
Corporate sustainability reports from major AI and cloud providers, specifically looking for new commitments to circular economy practices, hardware longevity, or take-back programs for data center infrastructure.
Further independent research or rebuttals to BAN’s methodology, particularly regarding the 'AI Waste Contagion' category and the projected growth rates of data center capacity.