The Smart Part Got Cheap. The Front Door Didn't.
Updated: 3 days ago
Who Gets Paid in AI | Part 1 of 3
By Hans Albrecht, Options With Hans
I've been pondering agentic and the direction of the AI value chain. Muse, the new AI agent from Meta Platforms (META), might be the most important AI development of the past year. Remember when everyone was getting down on Apple because their AI strategy was almost non-existent? I said nonsense, AI will come right to them. A mass of users and margin-protected loyalty. This is along those same lines.
Meta/Muse - the smartest AI on the planet? Nope. It doesn't need to be.
Over this three-part series I'm going to walk you through why. Part 1 is the foundation: where the money in AI is really heading. Part 2 is Meta. Part 3 is how to position for it.
Two years ago the whole industry obsessed over one question. Whose model is smartest? Every launch came with a scoreboard... crushed this benchmark, beat that one on math, remembers more than the last guy. That race is still running. It's just not the one that decides who gets paid anymore.
The reason is simple. The smart part is getting cheap. Really cheap.
Epoch AI, an independent research shop that tracks this closely, says the price of a given level of AI performance has fallen about 13x a year since 2023. Stack three years of that and the same brainpower costs something like 2,000 times less. Epoch calls it the fastest price drop of any world-changing technology in history. For perspective, computing got cheaper at about 1.5x a year from 1940 to 2001, and we thought that was fast.
You can watch it happen. Last month Anthropic released its new flagship model, Opus 5.5. It cut the price 20 percent, and Anthropic says it runs about 40 percent cheaper than the model it replaced because it finishes the same job with fewer tokens, the little chunks of text AI gets billed on. Smarter and cheaper, in one release. And that's before you count the open-source models anyone can download for free.
Remember Father of the Bride? Steve Martin gets sticker shock planning the wedding, Martin Short offers him steak or chicken, and he takes the chicken. That's where AI is going. When the near-as-good model costs a fraction of the price, most people take the chicken. Most businesses do too.
Yeah I know... Hans, the best models will always get paid. Some will. If you need the absolute best model in the world to crack a physics or chemistry problem, you want the steak, and somebody's going to sell it to you at steak prices. That's a real business. It's just not where most of the money gets spent.
And this should worry anyone selling smarts alone. There's no barrier to leaving. If a cheaper or smarter model shows up next month, you switch in a couple of taps. Nobody's loyal to a benchmark score. That's not a moat, right? It's a revolving door.
So if intelligence is turning into the cheaper chicken, where does the money go? Ask a simple question. Who spends the money in the end? Human beings.
The value goes to whoever owns the habit. The app you open first in the morning. The assistant that knows your files, your people and how you like things done, and quietly knocks out dozens of chores a day for you. Once that's baked into how you run your life or your business, leaving is a real pain. That's the moat.
Fifteen years as a market maker taught me a version of this. The sharpest pricing model on the floor didn't guarantee you a thing. What paid was order flow. If the customer came to you first, you ate. If they didn't, your brilliant model just sat there looking pretty. AI is heading the same way.
The models are turning into market makers. The front door is the order flow.
Which brings me back to Muse. Meta already has 3.6 billion sets of eyeballs on its apps every day. Last month I said Meta spent two decades building the world's biggest waiting room and Muse put a door in it. The bigger point is this. Meta doesn't have to win the IQ contest. It has to make life easier for the people already standing in its hallway. Save them time. Save them money. Help them make money. Do that well, and they don't leave.
Part 2 lands Tuesday. I'll show you why that's such a big deal, why Meta can afford to practically give the smart part away, and the browser war from the '90s that tells you how fights like this usually end.
Trade smart, stay hedged,
Hans
P.S. Drop a comment and tell me which AI app you open first every morning. I'm curious how many of you pick the smartest one... and how many pick the one that's already on your phone.
What to put on your sticky note
The smart part is getting cheap, fast. Raw intelligence alone is not a moat.
Most people and businesses take the chicken: near-as-good at a fraction of the price.
The money goes to whoever owns the habit, the front door people open first.
Models are the market makers. The front door is the order flow.
This post is for educational and informational purposes only and is not investment advice or a recommendation to buy or sell any security. Options trading involves substantial risk and is not suitable for all investors. Past performance does not guarantee future results. Do your own due diligence and consult a licensed professional before making investment decisions.
.png)



Comments