Local AI? I’d love to have it. Just not yet.

I’m a geek. I always want one more computer.

I also know that wanting a machine and needing one are different things, so I try to keep my hardware habit under control. Usually.

And I’m firmly in the ownership camp. Given the choice, I’d rather buy than rent.

Even with those biases, I struggle to justify buying a dedicated machine for local AI.

The Macs I find myself considering cost thousands of pounds. Then the familiar negotiation starts: 128GB of memory looks tempting, but what about 256GB? If I’m spending that much already, should I stretch to 512GB? And surely I should upgrade the storage while I’m there.

The configurations I’m looking at range from around £5,100 to £23,000 or more. That’s an insane amount of money for a personal computer.

The question is whether the machine would give me the coding experience I’m paying for: capable models, enough context for substantial tasks, and a speed I’m happy to work at. More memory gives me room to experiment. It doesn’t settle all those questions.

I often see local AI advocates compare it to owning a car rather than renting one. As someone who prefers ownership, I understand the appeal.

But for the setup I want, aviation feels like a better analogy.

I can buy a ticket, fly economy or business, and let someone else operate the aircraft. Buying my own local AI machine right now feels like buying a private jet.

I don’t expect RAM prices to come down any time soon. But I’m hoping we’ll soon have models capable enough for the work I do, running smoothly with enough context on a £3,000–£4,000 machine.

That’s the computer I want.