An interview in Kurier's Lower Austria edition, printed alongside a report on a planned data center in Stetten. I argued that Austria does need local compute so that research, industry, and public services can process sensitive data at home, while being candid about the cost: a large AI data center draws hundreds of megawatts around the clock, including at times when regional demand already peaks. I pushed back on jobs as a selling point, since a typical facility employs around 50 people once the one-to-two-year construction peak is over, and noted that digital sovereignty depends on who operates a site and what it is used for. Since regions differ in resources, grid capacity, and priorities, siting decisions belong at the regional level, which makes a regional strategy essential.

Coverage by Austria's public broadcaster of our Joule paper on governing AI infrastructure. The piece reports the study's central argument that data centers should adapt to grid capacity rather than the reverse, through connection agreements that guarantee a base load and allow additional capacity to be curtailed when the grid is stressed. I am quoted on why operators site facilities where a grid connection can be obtained quickly rather than where electricity is cheapest, and on the need for better data on data center consumption to plan grid infrastructure.

Der Standard's report on our Joule paper. The article explains the study's finding that data centers must adapt to existing grids rather than the other way around, and places it against the backdrop of the AI boom reviving gas power plants in the United States, where by late 2025 a large share of planned gas capacity was earmarked for data centers. It also notes the growing public resistance to data center projects in Austria.

An explainer on the data center boom and what it means for the grid and Europe's digital sovereignty. Interviewed alongside Professor Ivona Brandic, I discussed why data centers are so much more present in public debate today than a few years ago, why much of the announced demand is speculative, and how flexible operation, such as scheduling training runs when renewable generation is high or the grid is lightly loaded, could ease the burden data centers place on the energy system.

A one-day workshop convening researchers, industry, policymakers, and system operators on how Norway can become a European AI data center hub without derailing its energy transition. Co-organized by TU Wien, NTNU, Imperial College London, the University of A Coruña, and DNV, the program covered global AI energy demand, grid integration, local impacts, and governance. I chaired the organizing committee and gave the morning wrap-up; the discussions fed directly into our Joule paper on governing AI infrastructure.