The AI Trap

Right now, companies are falling into a trap reorganising teams, products, and entire operating models around AI. Almost all of them are assuming the cost of using AI will stay roughly where it is today.

That assumption is wrong.

This isn’t about “tokens” or technical detail. It’s about basic business logic. AI is becoming a core business cost — like cloud hosting, energy, or logistics. And every serious signal says that cost is going up.

We’re already seeing it.

Google announced AI infrastructure price increases this week. AWS raised AI-related pricing earlier in January. These aren’t random tweaks — they’re early moves in a much longer shift.

Here’s the uncomfortable truth most companies are ignoring: 

If you replace a $100k developer with an AI model that costs a few hundred dollars a year, you should not expect that price to stay at a few hundred dollars.

Once your workflows, products, and teams depend on it, the rational price for that capability isn’t $100 — it’s closer to the $100k it replaced.

Otherwise........ why did all those investors put billions into these companies?

They didn’t do it to permanently save you money. They did it expecting pricing power once dependency is established. Dependency is growing fast.

Every time AI gets better, companies use more of it — not less. What starts as a pilot in customer support spreads to engineering, finance, sales, marketing, analytics, and product features. Its hard to say no when there are so many productivity gains/cost savings. Usage compounds. Costs follow.

At the same time, the infrastructure underneath AI is getting more constrained. The newest AI chips are dramatically more expensive than previous generations. Memory required for AI has already increased three fold in three months. Data centres are running into power and cooling limits. Governments are restricting where advanced hardware can be sold. This is no longer a cheap, frictionless global market.

Add one more factor: the AI investment bubble is cooling. Not collapsing — cooling. As capital becomes harder to come by, AI providers will be forced to do what cloud providers did before them: shift from growth-at-all-costs to profitability. When that happens, prices rise.

So what should you do?

Don't avoid AI, that’s like taking your watch off and expecting time to stop. But stop treating it like magic.

Design your business so you’re not locked into a single provider. Mix premium AI services with cheaper and opensource alternatives. Invest in software that lets you switch models without rebuilding the business. And don’t hollow out human capability so completely that a price rise becomes an existential threat.

The bigger you are, the harder this gets. Enterprise buys 'IBM', it rarely buys opensource.

The companies that win the next phase won’t be the ones that ran blindly into AI adoption, regardless of the short term gains. They’ll be the ones that assumed, correctly, that prices rise once you’re hooked.