
Methodology: Every two weeks we collect most relevant posts on LinkedIn for selected topics and create an overall summary only based on these posts. If you´re interested in the single posts behind, you can find them here: https://linktr.ee/thomasallgeyer. Have a great read!
Thanks to Alyson Freeman, Arwel Owen, Navveen Balani, Simon R., ThankGod Obobo, Alex Orr, Andrew Sjoquist, Durval Jacintho, Joerg Wuttke, Mark Butcher, Matthias Haymoz, Michael Lesniak, Obinna Isiadinso, Petteri Nikki, Rachel Johnson Schiebout, Shaolei Ren, Steve Chavez, Tim Christophersen, Andy Davis, Carlos Sandoval Castro, Kristine Suber, Robert Keus, ADRIAN MUNDAY, Douglas Mouton, Jilali Mouale, Lexi Sydow, Lydia Sewell, Pascal Joly and Patrick Galey, and everyone else who contributed insights to this edition.
Find the full list of contributors and voices on LinkedIn: https://www.linkedin.com/pulse/best-linkedin-cw-32-33-green-ict-sustainable-ai-thomas-allgeyer-qwdhe/?trackingId=8EAJFdTL3A3BEN%2Bh%2B9U6wA%3D%3D
Not at the current pace. Global data center power demand will nearly double by 2030, and the US faces a projected 45 gigawatt shortfall by 2028. Amazon's GW Ranch gas plant in Texas is projected to emit 33 million tonnes of CO2 a year, nine times Amazon's global Scope 2 footprint. Texas is restricting new renewables while approving huge data center loads.
Because AI racks have outgrown air. AI clusters now exceed 100 kilowatts per rack, past air cooling's 10 to 15 kilowatt ceiling. Immersion cooling pushes PUE below 1.03, against 1.2 to 1.7 for air cooled sites. NVIDIA's liquid cooled Rubin architecture nearly eliminates on-site water use, saving an estimated 4 million dollars a year per 50 megawatt facility.
Because agents run long chains of model calls. Agentic coding tasks consume roughly 1,000 times more tokens than a standard chat exchange, per Microsoft Research and Stanford. Anthropic's Economic Index found 97% of its API usage now shows automation-dominant behaviour. Efficiency tools help: GreenPT's output compression cuts tokens by up to 70%.
How we solve real problems for real clients