Your Brain Sips 20 Watts. AI Orders a Power Plant.

I noticed something odd while scrolling through crypto Twitter last week. Someone had posted a chart comparing the human brain's power draw to a supercomputer's, and the comments were all variations of "see, AI is a disaster." The numbers look damning. Your brain runs on roughly 20 watts. LineShine, the world's fastest supercomputer in Shenzhen, pulls 42.2 million watts. That's about two million times more. It's a great meme. It's also mostly wrong.

The 20-watt myth

Here's the thing nobody sharing that graphic mentions. The comparison traces back to a single academic paper, and the most striking statistic has been misattributed for three years. I'm not entirely convinced anyone actually read the source. They just saw a big number, felt a dopamine hit, and hit retweet.

Granted, the raw comparison holds up. The brain does run on 20 watts, and it performs incredible feats of pattern recognition, language processing, and motor control on that budget. LineShine eats 42.2 million watts to do similar things, and it still can't understand context the way a five-year-old can. To be fair, though, the brain is a narrow, specialized system honed by millions of years of evolution. It's not running a general-purpose training algorithm in the cloud. It's a custom chip for survival, not for summarizing your email inbox.

So when people say AI is inefficient because it uses more power than a brain, they're comparing an F-16 to a bicycle and complaining about the fuel bill. Any car is less efficient than a bicycle. That doesn't make the car useless.

Electricity is the real contest

Pull the camera back and the real fight becomes clear. It was always about electricity. AI companies are hoovering up power wherever they can find it, and now they're going head to head with the one energy-hungry industry that knows exactly how to feel that pressure: crypto mining.

Miners have spent years chasing cheap gigawatts, negotiating with grid operators, building next to hydro dams and flare gas sites. Now the hyperscalers have shown up with deeper pockets. In places like Texas, grid operators are already watching data centers and mining facilities compete for the same electrons. The question worth asking: what happens when an AI data center can outbid a Bitcoin miner for a block of baseload power? The miner can switch off when prices spike. The data center can't.

And there's a parallel industry emerging where this all gets stranger. Some projects are using AI to make mining more efficient, which sounds clever until you remember that running the AI itself takes power. It's the heat pump problem, where the cure requires a bigger plug than the disease.

My honest take

I'll say it plainly: we need to stop clutching pearls over the brain comparison. It's a feel-good story that doesn't survive contact with reality. The human brain is incredible, but it didn't build ChatGPT while watching reruns of The Office. AI runs on electricity because that's what it costs to do this work in a general, scalable way. Age isn't necessarily better.

For regular people and crypto folks alike, the practical takeaway isn't about watts. It's about the grid. The next major bottleneck for AI and mining isn't a shortage of GPUs, it's a shortage of power. If you want to know which AI projects and miners will survive, watch their power purchase agreements, not their tech demos. History suggests otherwise for most startups, but the ones that lock down cheap, reliable electricity will be the ones still standing.

So sure, your brain runs on 20 watts. But it's the same brain that invented the power plant, and it's using that power to build something it doesn't fully understand yet. Time will tell if that's a good investment of electrons.