short4:33
How much electricity does AI actually use?
A concise explainer from the International Energy Agency on how AI changes electricity demand and why the numbers matter.
IEAElectricityAI Demand
AI, HPC, distributed systems, and energy intelligence.
This page brings together selected videos and technical notes on how AI workloads, data streams, and large-scale compute systems affect power demand, efficiency, and infrastructure decisions.
deep-dive16:17
How Google, Microsoft And Amazon Are Racing To Solve The AI Energy Crisis
A current overview of how hyperscalers are responding to the power, cooling, and infrastructure strain created by fast-growing AI workloads.
Useful as a high-level briefing on the real operational side of AI adoption: electricity supply, data-centre expansion, and the pressure to make compute growth sustainable.
AI InfrastructureData CentersEnergy Demand
Source and references
Recent talks and explainers
A short selection of recent public material on AI electricity demand, data-center expansion, power-grid pressure, and more efficient model design.
talk1:27:40
Energy Optimal AI
A deeper academic talk on designing AI systems that are not only accurate and fast, but explicitly energy efficient.
Efficient AIResearchSustainability
deep-dive5:26
Why America’s Power Grid Will Be Able To Withstand The $2.5 Trillion A.I. Datacenter Building Boom
A short, current look at how grid operators and infrastructure planners are thinking about AI-driven data-centre growth.
Power GridData CentersPlanning
Current writing themes
Short essays and research notes for readers who want measurable systems insight rather than generic trend commentary.
AI & Energy
2026-03-12·7 min
AI demand is now an infrastructure and energy question
A practical look at why model growth now has to be discussed in terms of power budgets, cooling limits, and grid-facing capacity, not model quality alone.
AIEnergy DemandInfrastructure
Distributed Systems
2026-02-18·6 min
Measuring distributed systems in watts, not just latency
Why throughput, stale reads, and response time tell only part of the story when workload placement and node behaviour also shape energy efficiency.
Distributed SystemsEnergyCassandra
HPC Method
2026-01-09·9 min
From profiling to control in HPC performance work
How to design experiments that reveal causal system behaviour across CPUs, memory, and workload phases before optimisation starts.
HPCPerformanceMethodology