Who’s afraid of the big, bad GPU?

Who’s afraid of the big, bad GPU?by Justine CalmaJul 21, 2026, 11:00 AM GMT+1LinkShareGiftHow you feel about AI’s energy use might change how you feel about computers entirely.This is Microsoft’s Fairwater data center in Mount Pleasant, Wisconsin. Built on the site of the failed Foxconn LCD project, it is currently the most powerful AI data center in the world.Here’s a single Nvidia B200 Blackwell AI GPU, the kind used in data centers like Microsoft’s. It is enormously power-hungry: Configured at max capacity, it can draw 1,200W of power.Fairwater sites are full of racks stacked two high, each containing 72 B200 GPUs. Together, they form a giant supercomputer made up of thousands of chips. Add it all up, and data centers require an enormous amount of power. They drive up local energy prices but don't create a lot of jobs or economic opportunity to go with it.The AI data center boom has given GPU energy usage a moral and political dimension. But GPUs and data centers are everywhere, powering everything, and we don't have a good way of talking about what's worth it and what isn't.This is an Nvidia GeForce RTX 5090 gaming GPU, which is also based on the Blackwell architecture. It has a max TDP of 575W — energy that’s used for games, not AI. Is that a good use of energy?Sony has sold some 92 million PS5s, which are built around an AMD GPU. It’ll draw around 220W at maximum. Add it up, and that’s a lot of power use. Is that a good use of energy?What about thousands of people playing together online, with their individual GPUs all using energy?Here’s one of the several Amazon Web Services data centers around the world used to run Fortnite. How do you feel about these data centers? About Fortnite’s use of energy?Here’s Apple’s A19 chip, which has both an integrated GPU and a Neural Engine for AI workloads. You’ll never see it, but it shapes a lot of your daily life.It has a modest power draw. The iPhone Pro might use a couple watts for browsing, and 5W in a demanding game.But there are around 1.5 billion active iPhones in the world. Using their GPUs drains their batteries, which need to be recharged at higher rates, using more energy. Is running down an iPhone battery for AI a good use of energy? For games?In the end, almost every move you make with a computer involves energy — especially since GPUs and data centers are a part of nearly everything now.From electricity to water, this is a story about graphics cards eating our planet — and how to understand our role in that.HowHow does AI make you feel? Are you excited to “vibe-code” your smart home? Or anxious about all the added pollution and billions of gallons of water used by data centers? Dig a little deeper and you’ll start to question the actual value of the GPUs that underpin all the leaps and promises of generative AI.Sidebarby Sean HollisterGPU stands for graphics processing unit, and it originally had nothing to do with AI at all. GPUs were designed to be a companion to a computer's CPU — or central processing unit — to offload the hard work of rendering computer graphics. While all sorts of graphics accelerators predated “GPU” as a marketing term, modern GPUs excelled because they could process multiple graphical tasks in parallel across their multiple graphics pipelines. Roughly 20 years ago, Nvidia began to help programmers run non-graphical computing on those parallel computers too. GPUs eventually got hundreds, then thousands, and now tens of thousands of cores for massively parallel general purpose “GPGPU” computing, which sped up training of neural networks, image recognition, and now modern AI training and inference.Right now, GPUs, hundreds of thousands of them, are being crammed into data centers around the world to power the AI boom. These chips are also found in everything from smartphones to cars to gaming PCs. Nvidia — once a niche chipmaker that has become the world’s most valuable company — still brags about releasing what it calls “the world’s first GPU” an