Data Sweat is an interactive web-based art project that confronts users with the local environmental footprint of AI data centers.
This piece was created during the AI Ecologies: Ecological-Technological Entanglements workshop (13-14 and 20-21 September 2025 in Prague). It was developed by Jakub Zíka & Federica Dalla Costa under the mentorship of visual artist Ellen Pearlman, a Fulbright Specialist, visiting CETE-P.

Concept

The work employs a metaphorical visualization of the often invisible impact stemming from the abstract computational resources consumed by AI models, such as the tokens in mundane Large Language Model (LLM) queries. It relates these small actions to the immense impact of the AI data centers designed to meet their demand, translating the resources into visible metrics of ecological disruption: electricity use, water consumption, and noise pollution.

The data center shining, sweating, and humming

Users engage with a simulated AI system in an environment where their actions have visible consequences. To trigger the impact, a user selects a simple "digits task" to prompt the AI. The simulated LLM replies with gibberish, after which the screen displays the environmental footprint of that individual query. A single AI query may seem insignificant, but the combined consumption of a large-scale population, such as the 1.4 million residents of Prague, has a profound and often invisible impact.
This collective cost is visually represented by the presence of the data center, visualized as a sort of "divine entity" at the top of the screen, contrasting with a 3D model of Prague in the background. Following the prompt, the data center shakes, shines (with energy), sweats (with heat and water), and hums, representing the massive, disruptive energy required to satisfy AI's computational needs. The decision to set the piece in Prague underscores the fact that while AI's global footprint is often unseen and unheard, its environmental impact is intensely localized, directly affecting the communities where these facilities are built.

Environmental Impact

In reality, AI data centers can consume millions of gallons of water annually for cooling, putting a significant strain on local supplies, consume a large amount of electricity, and the powerful cooling equipment and on-site generators create a constant, low-frequency humming noise.
The Real-Time Visualization shows the consumption of electricity and water, along with a representation of noise pollution, putting into perspective the sheer resources needed for continuous AI operation.

Artistic Approach

Using a metaphorical approach informed by reports, such as from the International Energy Agency (IEA), Data Sweat is a art piece, it is explicitly not a scientific report, designed to visualize and personalize the immense consumption of large-scale AI infrastructure. It is important to note that precise data on AI consumption and data centers constantly changes and evolves as the technology advances.

Visit the interactive project: https://datacenter-sweats.jakubzika.com/