Today, my latest research—published in Resources, Conservation & Recycling—addresses another, often overlooked consequence of the expanding AI infrastructure: electronic waste.

While this is not the first attempt to examine AI-related e-waste, it is the first estimate grounded in supply-side data, rather than inferred from demand-side assumptions.

The main findings are as follows:

• By 2030, AI servers could generate 131.0–224.8 kilotons of e-waste per year.
• AI systems may contribute less to global e-waste than previously anticipated.
• The gap highlights the need for supply-chain data and realistic AI server lifespans.
• 2030 AI e-waste could still match Denmark, Norway, or Austria’s 2022 e-waste.
• Substantial AI e-waste persists, underscoring the need for data center transparency.

The full article is available open access and can be accessed here: https://doi.org/10.1016/j.resconrec.2026.108872

AI systems may now have a carbon footprint equivalent to that of New York City in 2025, while their water footprint could be in the range of the global annual consumption of bottled water.

After previously estimating the global power demand of AI systems in 2023, 2024 and 2025, my latest research—published open access in the journal Patterns—now provides further insights into the carbon emissions and water consumption related to that power demand. Determining these metrics is a challenge, as “such estimates are complicated by the fact that data center operators do not publicly disclose the required inputs.” The new article highlights the shortcomings in the environmental disclosure of data center operators, in particular with regard to indirect water consumption of data centers and AI-specific metrics. Using the limited information that is available, the article finds that:

• AI systems could be responsible for 32.6–79.7 million tons of CO₂ emissions in 2025, comparable to the annual emissions of a major city.
• The water footprint of AI alone could reach 312.5–764.6 billion liters, potentially rivaling global annual bottled water consumption.
• Current corporate sustainability disclosures never distinguish between AI and non-AI workloads and rarely report indirect water consumption, creating significant uncertainty and likely underestimation of impacts—especially for water.

By highlighting both the scale of AI’s environmental impact and the gaps in available data, this research underscores the urgent need for improved disclosure to responsibly manage the growing footprint of artificial intelligence.

The full article can be accessed here: https://doi.org/10.1016/j.patter.2025.101430

Artificial intelligence (AI) is rapidly becoming the largest energy hog within worldwide digital infrastructure.

In my latest research, published in the academic journal Joule today, I show that AI systems were responsible for up to 20% of global data center power demand by the end of last year. Moreover, this share could approach half of data center power demand by the end of this year, as the power demand of AI systems could rise to 23 gigawatts. This exceeds the current power demand of cryptocurrency mining and is equivalent to twice the power needed to keep my home country the Netherlands running.

Big tech companies are well aware of this trend, as companies such as Google even mention having faced a “power capacity crisis” in their efforts to expand data center capacity. At the same time, these companies prefer not to talk about the numbers involved. Google was the only big tech company revealing that AI represented 10%–15% of their total energy use over the years 2019 – 2021, but since ChatGPT kicked-off the AI-hype we’ve never seen anything like this again. As a result, it remains virtually impossible to gain a good insight into the actual energy consumption of AI.

By diving into semiconductor manufacturing (concentrated in East Asia), a bunch of (Chinese) analyst reports and earnings call transcripts I found a way to shed some light on where the power demand of AI is heading – and it turns out this growing at lightning speed. This growth clashes with other social ambitions, such as achieving climate goals and reducing total energy consumption. However, effective policy responses will first require urgent transparency from an otherwise opaque industry.

The full article can be accessed for free during the next 50 days through the following link: https://authors.elsevier.com/a/1l8Ke925JEVNpG

DOI: https://doi.org/10.1016/j.joule.2025.101961

In recent years it has become well known that cryptocurrency networks such as the Bitcoin network are energy intensive systems. As of 2024, Bitcoin mining operates on more than 16 gigawatts of power demand, responsible for around 80 megatonnes of annual carbon emissions. Ethereum proved in 2022 that it is possible to replace this energy intensive (proof of work) mechanism with a more sustainable alternative known as proof of stake. As a result of this change, Ethereum reduced its total power demand by at least 99.85%.  However, while this was the most obvious impact of the change, it should be kept in mind that such a software change affected Ethereum in more ways that one. In particular, the Bitcoin community tends to argue that this software change has made Ethereum less decentralized and less secure.

Digiconomist has now added a new deep dive into the decentralization of cryptocurrencies to examine these claims in more detail. This new content explores how (de)centralization takes place in cryptocurrency networks when proof-of-work or proof-of-stake mechanisms are employed. Moreover, it also highlights other aspects of blockchain (de)centralization that aren’t exclusively related to the aforementioned mechanisms. By doing so, it is shown that centralization and decentralization in blockchain technology are not binary values, but rather a spectrum in which each software design will have its own unique landscape of risk factors that may impair decentralization.  A summarizing visual was added to the Bitcoin Energy Consumption Index page.

It has been nine years since Digiconomist was first launched with the objective of “exposing the unintended consequences of digital trends.” For a big part of these nine years, the sustainability of digital assets such as Bitcoin has been a key focus of the research by Digiconomist. However, in 2022 and 2023 a new digital trend has emerged that has an equal potential to rapidly grow in terms of electricity consumption: artificial intelligence (AI). If not managed properly, AI could be responsible for as much electricity consumption as Bitcoin is today in just a few years’ time. This is the conclusion of a new research by Digiconomist titled “The Growing Energy Footprint of Artificial Intelligence” that was published in the journal Joule today (October 10, 2023).

AI-servers are power-hungry devices. A single NVIDIA DGX A100 server can consume as much electricity as a handful of US households combined. Because of this, the electricity consumption of hundreds of thousands of these devices will start to add up quickly. While the supply chain of AI-servers is facing some bottlenecks in the immediate future that will hold back AI-related electricity consumption, it may not take long before these bottlenecks are resolved. By 2027 worldwide AI-related electricity consumption could increase by 85.4–134.0 TWh of annual electricity consumption from newly manufactured servers. This figure is comparable to the annual electricity consumption of countries such as the Netherlands, Argentina and Sweden. While this would represent half a percent of worldwide electricity consumption, it would also represent a potential significant increase in worldwide data center electricity consumption. The latter has been estimated to represent one percent of worldwide electricity consumption.

Given the potential growth of AI-related electricity consumption, the new research contains a call to action to be mindful about the use of AI. Emerging technologies such as AI and previously blockchain are accompanied by a lot of hype and fear of missing out. This often leads to the creation of applications that yield little to no benefit to the end-users. However, with AI being an energy-intensive technology, this can also result in a significant amount of wasted resources. A big part of this waste can be mitigated by taking a step back and attempting to build solutions that provide the best fit with the needs of the end-users (and avoid forcing the use of a specific technology). AI will not be a miracle cure for everything as it ultimately has various limitations. These limitations include factors such as hallucinations, discriminatory effects and privacy concerns. Environmental sustainability now represents another addition to this list of concerns.

For the first 50 days after the publication of the article it can be accessed for free using the following link: https://authors.elsevier.com/a/1huvY925JENm45 Copies can also be requested via email and through the contact form.

Also check out media headlines on this new release:

The New York Times: “A.I. Could Soon Need as Much Electricity as an Entire Country”
BBC News: “Warning AI industry could use as much energy as the Netherlands”
The Verge: “The environmental impact of the AI revolution is starting to come into focus”
NewScientist: “Should we be worried about AI’s growing energy use?”
Mirage News: “AI Power Demand May Equal Small Country’s Electricity Use“