A Megawatt Per Rack? | Deep Geeks ep01

Frontier lab deals are no longer announced in dollars or FLOPS. They are announced in gigawatts. This episode of Deep Geeks explores how we got here. https://www.weka.io/podcast/can-ai-survive-its-own-energy-appetite-deepgeeks-ep-1?utm_source=youtube&utm_medium=social CPU racks held thousands of cores. GPU racks hold millions. That jump is not just a parallelization problem for compute, networking, and storage software, it is an energy problem. A rack of GPUs draws hundreds of kilowatts today, and the latest generation is pushing toward a megawatt per rack. In this clip from DeepGeeks Episode 1, WEKA and Nebius unpack why the biggest announcements between frontier labs and GPU suppliers are now measured in gigawatts of capacity rather than performance benchmarks or deal value. For scale: one gigawatt is roughly what a single nuclear power plant generates. Multi-gigawatt AI deployments mean multiple power plants dedicated to a handful of data centers, largely to serve inference. The takeaway is not that AI is unsustainable. It is that the unit of account for infrastructure has changed, and the teams that plan for energy the way they used to plan for capacity will be the ones that scale. Learn more: Listen to the full episode: https://www.weka.io/podcast/can-ai-survive-its-own-energy-appetite-deepgeeks-ep-1? utm_source=youtube&utm_medium=social&utm_content=full-episode WEKA NeuralMesh: https://www.weka.io/product/neuralmesh/?utm_source=youtube&utm_medium=social&utm_content=product-page Subscribe for more AI infrastructure content: https://www.youtube.com/@WekaIO?sub_confirmation=1