The global expansion of data centers for artificial intelligence could lead to a significant increase in electronic waste, as equipment—including servers—will be upgraded more frequently. According to a new report by the environmental organization Basel Action Network (BAN), AI-managed infrastructure could triple the annual volume of e-waste worldwide over the next 25 years.

The organization predicts that the decommissioning of AI-related equipment could generate up to 13 million metric tons of electronic waste annually by 2030. These calculations include not only servers for AI models but also network equipment, power distribution systems, cooling devices, energy storage systems, cables, and other physical infrastructure components essential for large data centers.

The report notes that the amount of discarded equipment generated by AI in the coming decades could fill enough shipping containers to circle the globe six times. Such equipment contains hazardous materials, including lead, mercury, cadmium, and so-called “forever chemicals,” as well as valuable metals that are difficult to recover efficiently. According to BAN, currently only about one-fifth of electronic waste is properly recycled.

BAN’s estimates are partly based on forecasts for the construction of new data centers. Analysts factor in the addition of approximately 100 GW of computing capacity worldwide by 2030 and model the physical equipment required for this expansion. A single standard server rack can weigh around 1,360 kg, and network switches, cables, and auxiliary systems further increase material needs.

A key assumption in the forecast is the rapid replacement of equipment. The report states that most hardware associated with AI infrastructure will be replaced every two to five years, rather than being designed for repair, reuse, or long-term service. It also notes that growing AI computing demands may accelerate the upgrade cycle for consumer PCs, smartphones, and telecommunications equipment, as users seek devices capable of handling more complex tasks.

The organization stresses the need to develop plans for repair, reuse, recycling, and responsible end-of-life management of equipment to prevent the worsening of the waste problem as AI infrastructure grows. The figures provided are projections and depend on the pace of data center construction and the frequency of equipment replacement by operators.