Compute is becoming electricity

The four biggest cloud providers plan to spend on the order of $725 billion on AI infrastructure this year, with well over $130 billion of capex in a single quarter, most of it pouring into AI data centers. That number is staggering on its own, but the more useful thing it tells you is about everyone who is not a hyperscaler.

It means compute is becoming a commodity. When the world's cheapest, fastest, most reliable model is one API call away from any team, "we use AI" stops being a differentiator. Anyone can ship that. Anyone is shipping that. The capability that felt like an edge in 2023 is now infrastructure, the same way a database or a CDN is infrastructure.

Where the moat actually moves

When the model layer commoditizes, defensibility moves up the stack. The durable advantages are the ones a competitor cannot replicate just by buying the same API access:

Proprietary data your product creates and learns from. A model anyone can call, fed by data only you have, is a combination nobody can copy off the shelf.
Workflows you have embedded AI into that customers cannot easily rebuild. The deeper AI sits inside a real process your users rely on, the higher the switching cost.
Trust to handle the messy parts of a real business. The hard, regulated, high-stakes corners of an industry are not solved by a smarter model. They are earned through reliability over time.
Integrations that turn a model into a useful coworker. A raw model is generic. A model wired correctly into the tools and context of a specific job is valuable.

Founders still pitching "we have the latest model in our app" are pitching from a couple of years ago. The teams winning now treat models like electricity: necessary, generic, replaceable. They spend their energy on the layer above, where the model stops being the product and becomes a component of it.

The question worth asking this week

If every competitor had the same model access tomorrow, what would still be hard to copy about your product?

Whatever answers that question is your actual moat. If the honest answer is "nothing," you do not have a strategy problem you can solve with a model upgrade. You have a product problem.

This is not an argument against using AI aggressively. It is an argument for being clear-eyed about where the value sits. Use the best model for the job, treat it as the swappable commodity it is becoming, and pour your real effort into the data, workflows, trust, and integration that make the whole thing defensible.

The commoditization of compute is genuinely good for builders. It means a small team with the right data and the right workflow can compete on the only layer that still matters, instead of trying to out-spend hyperscalers on the layer that does not.

We're here to help founders and teams design and build digital products that are built to scale with you, not slow you down. If you're looking to build something, get in contact with us today!

Models are becoming a utility. Your moat is everything you build on top of the socket.