I guess I just gave away the ending. I’m not in the mood for keeping you in suspense today.

But let’s unpack this, because the business world is going hog wild over the idea that AI not only gets things done more efficiently, but also much more cost-effectively.

You may have heard last week that Twitter founder and now Block CEO Jack Dorsey was laying off 4,000 of the company’s 10,000 employees because he simply realized there’s no way around the fact that AI can do the jobs of those 4,000 people.

Now, maybe it can, and I’m not here to minimize the personnel costs associated with 4,000 employees.

But I had a conversation recently with one of our vendors that shed light on at least one of the hidden costs associated with AI. He told me they are training white-hat hackers to expose vulnerabilities in client systems. This is hardly a new idea. Many companies use white-hat hackers to find where they’re vulnerable so they can be more strategic about patching.

Once they find what they need to learn from this process, they then train AI agents on what they’ve learned, so the AI agents have a better idea of what to look for in potential threats and action steps.

None of this raised an eyebrow for me, until he told me this: It’s costing him nearly $18,000 a month to do all this.

That much for white-hat hackers?

No. The money is for variables like GPU power, API calls and egress fees. These costs tend to be buried within Big Cloud and Big Colo providers, but that doesn’t mean you’re not paying them. People just don’t quite realize it yet.

And that’s what people miss. AI uses an enormous amount of power.

Even I could have missed this fairly easily. I live in a world of servers and data centers and all the accompanying infrastructure, and I’m used to sucking up lots of power. We consider it the cost of doing business.

But the more you rely on AI to handle functions your people used to handle, the more those power costs will rise. At the highest levels of the industry, it’s masked under your Big Cloud receipt.

And it’s not as if the users of this power are completely blind. I’m sure you’ve noticed all the public outrage that accompanies just about every announcement of a data center being built somewhere in America. These data centers are huge, so they’re usually looking for real estate in rural areas with lots of room – and that’s got the local residents worried about everything from sewage capacity to the safety of their water supply.

But what they should be worried about is the stability of the electrical grid.

The data centers will use so much power, it will completely alter the balance of how the grid services the public at large. And companies who make extensive use of AI for normal business functions are going to pay for the use of that power one way or another.

It’s the same thing we see with trendy innovations like crypto mining and electric cars. The power for all this doesn’t appear out of the ether. It has to be powered by the same fuel sources and the same power grid that keeps your lights on and your refrigerator cold.

Jack Dorsey’s payroll costs may be plummeting, but his jaw might drop when he sees the direction of his power and cloud bill.

Then again, he might also miss the institutional knowledge those people possessed. Consulting AI as a tool is resourceful. Relying on it blindly is asking for trouble.

The big new thing is never quite as simple, easy or free as it seems at first. The point of this is not to tell you not to use AI. We use it in all kinds of ways. It’s to help you reject the delusion that it’s free, or that its costs are merely nominal.

A few of us see this now. Soon everyone will.