Software is getting cheaper. The AI bill is not disappearing.
The barrier to building software is collapsing, but a viable product still needs margins that survive the cost of running it.
Expanded from a LinkedIn post originally published on 2026-10-07.
There’s a weird irony happening in tech right now.
AI is making it faster and cheaper than ever to build digital products. Apps that would have taken months can now be thrown together in days. Design is cheaper. Development is cheaper. Content is cheaper.
Which means the price people are willing to pay for those products is probably going to keep falling too.
But the AI bill?
That keeps turning up.
So eventually we may reach the perfect circular economy:
Build an AI product.
Sell it for £10.
Spend £8 on tokens getting someone to use it.
Use the remaining £2 to ask AI how to increase your margins.
Repeat until homeless.
The barrier to building software is collapsing, which is brilliant. But I think the interesting question over the next few years won’t be “Can you build it?”
Almost everyone will be able to.
It’ll be:
Can you build something valuable enough that the economics still make sense after the AI bill arrives?
Anyway, if anyone needs me, I’ll be outside the data centre reducing the cost of transferring energy between me and it.
#AI #SaaS #SoftwareDevelopment #IndieHacker #ArtificialIntelligence #Startups #Tech #BuildInPublic
Building is only one part of the economics
There is a difference between making a product cheaper to build and making it cheaper to operate. AI can help with the first while adding a new, recurring cost to the second.
The numbers in my original post are deliberately comic, not a forecast. The underlying question is serious: how much of the money a customer pays is left after serving that customer?
Value has to survive the invoice
A useful product is not automatically a viable business. If every interaction carries a cost, pricing and usage need to be thought about together rather than treated as separate problems.
That does not make the falling barrier to entry a bad thing. It means the next challenge is less about proving that software can be built and more about building something people value enough to keep paying for.