Everyone keeps yelling about AI being a bubble.

[atomicpoet@atomicpoet.org]

Everyone keeps yelling about AI being a bubble. “No one will adopt it.” “It will collapse under its own hype.” All the usual doomer talking points.

They are staring in the wrong direction.

The real risk of AI is not lack of adoption. The real risk is what happens to the U.S. electrical grid if the country gets slammed with extreme winter conditions while data centres are pulling record power.

This is the part almost no one talks about. We have compute demand rising fast. Data centres are adding massive load. Natural gas supply is fragile during deep freezes. Renewables drop when weather turns ugly. And the grid margins in several regions are already thin.

If a cold snap hits the wrong areas at the wrong time, the pressure lands squarely on the grid. Not on adoption curves. Not on VC funding. Not on whether ChatGPT writes good emails.

On the grid.

So here is the uncomfortable scenario. Rolling blackouts in peak winter. Homes freezing. Hospitals stressed. Millions of people furious. Politicians scrambling for someone to blame.

And who is an easier villain than “those giant AI data centres sucking up all the power”?

It would not surprise me if lawmakers snap and demand that the biggest clusters get kicked off the grid entirely. Not forever. Maybe not even for long. But long enough to make everyone rethink what “scalability” actually means when it runs headfirst into physical infrastructure.

AI will not fail because people refuse to use it. It will fail if the grid buckles under the weight of the future we are building.

And if that happens this winter, the debate will change overnight.

https://finance.yahoo.com/news/us-faces-winter-blackout-risks-201312926.html

[mike@thecanadian.social]

@atomicpoet Everytime I see these massive DC projects I always wonder, Where's the power coming from? And to a lesser extent the water. They're literally building huge DC's in the desert.