The most dangerous AI edge is the one that works. It earns real money today, and it is commoditising faster than you can build the thing meant to defend it.
Nine in ten organisations now use AI in at least one business function. Fewer than four in ten can attribute any measurable impact on profit. Same tools, opposite results.
The variable is not the model. It is a rate.
The two rates
Dario Amodei named the two that matter: two exponentials. One for how fast models improve. One for how fast the economy can absorb them.
The first is the capability rate. It belongs to the field — you read it off a press release. The second is your absorption rate: how reliably you turn a new capability into a number your CFO or customer would recognise. When capability outruns absorption, you are stockpiling power you cannot use. A better model becomes the most expensive way to feel productive.
Your absorption rate has a ceiling, and the ceiling is a single layer: the thing a capability must pass through before it reaches your customer. For most teams it is mundane: the data nobody cleaned, the one engineer who understands the legacy system, the customer who will not change how they work, the sign-off that takes three weeks. Whoever owns that layer captures the value, because everything the model can do flows through it.
Where the money actually landed
Microsoft holds the largest AI distribution in enterprise. It built that lead by running other companies' models through the channel it already owned: Office, Teams, Azure, and the procurement relationship every large firm already had. ~420 million people use Copilot across Microsoft's products each month. The strongest models inside it are still OpenAI's and Anthropic's.
The model was rented. The distribution was owned. The margin followed the distribution.
Jasper shows the other failure: $125M raised as a writing tool on OpenAI's models. When ChatGPT arrived free, its product became something anyone could get for nothing overnight. It survived by climbing into the layer it had skipped — enterprise workflow and data.
Why Chegg lost everything in one day
Chegg owned its layer outright: a decade-deep library of step-by-step homework solutions and the student traffic to match, a moat no competitor could rebuild quickly. Then a general model could do everything for free. In May 2023 Chegg told investors students were leaving for ChatGPT. The stock fell by half. From its 2021 peak Chegg has since lost more than 95% of its value.
It owned the scarce layer. It owned the wrong one.
Put Venice (printing press era) and Chegg side by side and you see the variable. Same kind of edge — a scarce layer others had to pass through — and the clocks ran a hundredfold apart. Venice's lasted a century. Chegg's lasted months.
A scarce layer pays only if its clock is longer than the time it takes you to build on it. Clear that bar and you compound. Miss it and you have bought a melting asset at full price.
What sets the term
A layer's clock is short when a general model can absorb it: a clever technique, a fine-tune, generic data anyone can assemble. It is long when the scarcity rests on something a model cannot manufacture: a regulator's approval, a physical bottleneck like fabs or power, years of accumulated switching cost, a trust relationship.
Chegg's layer was content a model could regenerate — it had months. The chips an AI runs on are a physical bottleneck — that scarcity holds for years.
Even that one is eroding. Nvidia owns roughly 80% of the merchant AI-accelerator market, and its largest customers are designing their own silicon to route around it. Gross margin in the mid-70s will be the first thing to crack: I expect it below 70% by 2028. If it still holds in the high-70s by then, the thesis is more durable than my model predicts.
The diagnostic you can run today
The work fits on one page:
1. Name your absorbing layer. Use the doubling test: what, if it doubled tomorrow, would let you use twice as much model — while doubling the model itself bought you nothing more?
2. Test whether you own it. You own a layer only when a new capability cannot reach your customer without passing through something of yours that a rival cannot rebuild in a weekend. A model fine-tuned on your own documents does not pass.
3. Measure your absorption rate. Count the capabilities you seriously tried this year and the ones that moved a number your CFO or customer would recognise. Most teams get an honest reading that stings.
4. Read both clocks. Estimate how long your layer stays scarce. Estimate your payback — how long a build on that layer takes to earn back what it costs. If the clock is shorter than the payback, you are Chegg.
The concrete case
Take a forty-person logistics company with three years of messy carrier-integration data no rival has cleaned. Double the data and the model gets more useful. Double the model and nothing changes. Nobody rebuilds three years of dirty feeds in a weekend. Every test passes.
Then a frontier model ships that parses raw carrier feeds zero-shot. Three years of cleaning collapses into a prompt. The layer that looked like years of scarcity now has six months, against an eighteen-month build. The bet flips from compound to melting — while the data sat untouched.
Re-run the number the moment a model moves.
The full diagnostic runs in your browser here — name your layer, get your absorption rate, the layer's half-life, and the date to start building the next one.
This is part of the Durability Curve Analysis series — structural analysis of where durable value forms in AI infrastructure, through the Five Laws framework (Bottleneck Migration, Difficulty is Load-Bearing, Architecture Outlives Content).
Top comments (0)