TL;DR
- AI adoption gets discussed as diffusion through industries. There is a second diffusion nobody plans for: how fast each layer of AI value turns into a commodity.
- The moat has moved twelve times in three years, from frontier models through inference, apps, workflows, verticals, data, synthetic data, evals, feedback loops, distribution and liability to judgment, and judgment hands it back to frontier models.
- Every rung was a real claim, defended by real people, funded at a real premium, and correct for somewhere between a few months and a few years. The pattern is not that they were wrong. It is that the half-life keeps shortening.
- Being in the business of AI is being in the business of selling a commodity. The winners play the arbitrage between the moment a piece of missing infrastructure appears and the moment it stops being scarce.
When founders and technologists deeply involved in AI meet the question we ask is "Where is this all headed?"
I am not talking about societal upheaval. Often the question's boundary is what will happen to AI businesses. When we talk about AI adoption, we talk about how long it takes for industries to adopt. That's diffusion. But there is another kind of diffusion. It's about how long it takes for various parts of AI value to commoditize.
Many moving parts exist, and each will become a commodity eventually. Along the way markets will be made, remade, and decimated.
I summoned the inner P.B.Shelley. It turns out I don't have one. So here is a high-school-level poem on how AI becomes a commodity and how moats will shift. I hope it is sufficiently Ozymandian.
The half-life of AI, in verse
Frontier models are commodity.
Inference is where the moat is.
Inference is commodity.
Apps are where the moat is.
Apps are commodity.
Workflows are where the moat is.
Workflows are commodity.
Verticals are where the moat is.
Verticals are commodity.
Data is where the moat is.
Data is commodity.
Synthetic data is where the moat is.
Synthetic data is commodity.
Evals are where the moat is.
Evals are commodity.
Feedback loops are where the moat is.
Feedback loops are commodity.
Distribution is where the moat is.
Distribution is commodity.
Liability is where the moat is.
Liability is commodity.
Judgment is where the moat is.
Judgment is commodity.
Frontier models are where the moat is.
There is no moat.
There is only cash flow.
The explanation, if you dislike poems:
Frontier models → Inference. For two years the entire industry agreed that whoever trained the biggest model would own the future. Then the capability gap narrowed to about six months and open weights arrived to finish the job. Attention moved to whoever could serve tokens cheapest, which sounded like a moat until we said the words "serve tokens cheapest" out loud.
Inference → Apps. Token prices fell roughly an order of magnitude a year, which is a wonderful thing for humanity and a terrible thing for the labs' gross margin. Serving inference turned out to be a utility business staffed by people who had recently raised at a software multiple. Everyone agreed the real value was closer to the user.
Apps → Workflows. Being closer to the user meant a page, a text box with a send button in the SaaS world. A competitor could ship the identical product over a long weekend and they did. The consensus updated: the durable thing was encoding how the work actually happens, with the approvals and the exceptions and the person who has to sign off.
Workflows → Verticals. Then everybody wrote a blog post about their agentic workflow architecture, complete with diagrams, and the workflows were copied within the quarter. Domain specificity became the answer. You cannot fake knowing what a clean claim is, went the reasoning, from people who had learned what a clean claim was eleven weeks earlier.
Verticals → Data. Domain expertise turned out to be four subject matter experts and a year of listening. Proprietary customer data was declared the true asset, on the theory that nobody else has it. This was the most emotionally satisfying rung and lasted the longest.
Data → Synthetic data. Enterprise data was smaller, dirtier, and more duplicated than the pitch decks implied, and models became sample-efficient enough that example number 10000 contributed roughly nothing. Better to generate exactly what you're missing. The moat became a prompt that says "produce two thousand variations of the following."
Synthetic data → Evals. Since anyone can write that prompt, the scarce skill became knowing whether the resulting model is worse than last Tuesday's. Evaluation harnesses were declared the real intellectual property, and for a brief period this was even true. Then people started publishing them.
Evals → Feedback loops. An eval describes a moment that has already passed. What compounds, we were told, is production signal: users correcting outputs, corrections improving the system, forever. This is a genuinely good mechanism, though it does require users.
Feedback loops → Distribution. It requires quite a lot of users, in fact, and the loop tooling got packaged and sold by six vendors. Whoever has the customers has the loop. Everyone rediscovered that enterprise software has always been a sales business and pretended to be surprised.
Distribution → Liability. Channels are rentable. You can buy the same conference booth, the same partner program, the same outbound sequence, and your competitor will. What isn't purchasable is agreeing to be the one who pays when the agent is wrong.
Liability → Judgment. Liability is a thing insurers price, and once it has a price it has a purchase order. It also only covers the small fraction of work that fails loudly enough to be actionable. The rest of the job (the overwhelming majority of it) is calls where there is no correct answer, no clean error, and nobody to sue. Which invoice to chase first. Which exception to escalate and which to let ride. Whether this customer is annoyed or leaving. So the retreat went there: not to who pays when it breaks, but to who decides when nothing is obviously broken.
Judgment → Frontier models. Good judgment, written down, is a checklist. Checklists are training data. Every quarter another category of expensive human discernment quietly ships as a default behavior in a model release. Congratulations: you have arrived back at rung one, where a lab you don't control eats the layer you just spent eighteen months defending.
What the pattern actually says
There is no moat. There is only cash flow. Every rung was a real claim, defended by real people, funded at a real premium, and correct for a few months to a few years. The pattern isn't that they were wrong. It's that the half-life keeps shortening. Which is why speed is the only position that survives the cycle, a point I make at more length in With AI, the only moat is speed.
I wrote a poem and a long-winding explanation to make a simple point: Being in the business of AI is being in the business of selling a commodity. The ones that win play the arbitrage between the inception of a piece of the missing infra and its ultimate loss of value in the chain as it gets prevalent.
Where Moative fits
Moative works as an AI operating partner with mid-market companies, which means we are usually on the buying side of this cycle rather than the selling side, and that changes the advice. If the moat moves every few quarters, an enterprise should not be trying to buy one. It should be buying the specific economics of a specific process, measured, and it should be keeping the judgment and the operating data on its own side of the contract. In healthcare revenue cycle management, for instance, the durable asset is not the model deciding which claims to work. It is the accumulated sense of which denials are worth appealing and which are not, and that belongs to the operator. If you are evaluating an AI vendor on the strength of its moat, it is worth asking what you own when that moat closes.
Frequently asked questions
What does the half-life of AI mean?
It describes how long a given layer of the AI value chain stays scarce before it becomes a commodity. Frontier models, inference, apps, workflows, verticals, data, evals, feedback loops and distribution have each held the position of "where the moat is" and each has lost it, in successively shorter periods.
Why is there no moat in AI?
Because every candidate layer has been copied, published, packaged or absorbed into a model release within months of being declared defensible. Each claim was real while it lasted. The problem is that the interval keeps shrinking, so no layer holds long enough to compound into a durable advantage.
Is proprietary data a moat in AI?
It was the most durable of the rungs, and it still helps, but it is weaker than the pitch decks suggest. Enterprise datasets tend to be smaller, dirtier and more duplicated than expected, and modern models are sample-efficient enough that the ten thousandth example adds very little.
Can AI replace human judgment?
Progressively, yes, in the sense that judgment written down becomes a checklist and a checklist becomes training data. Each quarter, another category of discernment that used to be expensive ships as default model behavior. The judgment that survives longest is the kind nobody has written down.
What should a company do if AI has no moat?
Stop buying moats and start buying economics. Work out which process the system improves, by how much, and what you still own if the capability becomes free next year. The answer should be your operating data, your customer relationships and your accumulated judgment about the work.
What is the difference between AI diffusion and AI commoditization?
Diffusion describes how long an industry takes to adopt a technology. Commoditization describes how long each part of that technology stays scarce enough to charge a premium for. Adoption timelines get most of the attention, but the commoditization clock is what decides where the profit ends up.
