Study finds X generates massive hidden CO2 costs from surveillance practices
X consumes 2x the network traffic of Mastodon for zero user benefit...
Get AI news that actually matters
One email a day. Zero fluff. Join 10,000+ professionals.
A new academic paper from researchers Nils Bonfils and Christoph Becker, submitted to the ICT4S 2026 conference and posted on arXiv (2605.26314), quantifies the environmental toll of surveillance capitalism. The study proposes a framework linking surveillance processes (telemetry, user tracking, data analytics, secondary data uses, AI model training, and massive data storage) to their material infrastructure and carbon footprint. Using a comparative case study of corporate platform X (formerly Twitter) and the decentralized, non-commercial alternative Mastodon, the researchers measured the proportion of network traffic driven by surveillance capitalism.
The key finding is the existence of 'corporate overhead' — excess resource consumption from X's for-profit data extraction practices that do not contribute to user experience. This overhead serves as a lower bound for CO2e emissions attributable solely to surveillance-driven activities. The study demonstrates that a significant portion of ICT's rising carbon impact can be traced back to the infrastructure required to harvest and monetize behavioral data. With 7 figures across 12 pages, the paper calls for greater accountability and offers a replicable method for assessing environmental costs of other platforms.
- X's surveillance capitalism practices (telemetry, ad targeting, data storage) create 'corporate overhead' — excess network traffic with zero user value.
- Mastodon, a decentralized non-commercial alternative, emits significantly less CO2e per user by avoiding profit-driven data extraction.
- The study establishes a lower bound for CO2e emissions attributable to for-profit surveillance activities that degrade user experience.
Why It Matters
Forces tech leaders to account for hidden environmental costs of data-driven business models beyond just compute.