Yesterday an 18-year-old Anthropic intern made $2.2 million with one prompt. Today the same company is said to be exploring a $2 trillion IPO. The distance between those two numbers is not a coincidence. It is the market’s precise measurement of the bet that AI has become infrastructure, not product.
The model layer is racing toward commodity at an accelerating pace. OpenAI’s latest release trades raw parameter count for agentic speed. DeepSeek’s new harness optimizes for tool chaining. X has open-sourced its recommendation algorithm. Each move treats intelligence as a utility, and utilities get priced accordingly. When the underlying capability becomes cheap and ubiquitous, value migrates to whoever controls the interface, the distribution, and the trust users place in the system. This is not speculation. We have watched it happen with compute itself; the same deflation curve now applies to the reasoning wrapped around it. The model is becoming the engine, not the car. X’s decision to open-source its recommendation algorithm is not an act of generosity. It is a recognition that the algorithm itself is no longer the moat; the data and the engagement it generates are. When the code is public, the advantage shifts to whoever can iterate faster on the feedback loop.
That is exactly where the competence gap opens. Roadmaps without dates are not strategy; they are wish lists dressed in corporate language. Teams deploy AI in isolated experiments and call it transformation. The tool answers the email. The tool drafts the report. The tool sits in the corner while the actual workflow keeps humming the way it always has. The stall is not technical. It is the gap between what the model can do and what the organization knows it should ask for. We see this in the surveys showing that most knowledge workers have tried an AI tool at work but only a fraction use it daily. The gap between trial and adoption is not a training problem. It is a design problem. The tools are built to demonstrate capability, not to fit into the messy middle of actual work where permissions, legacy systems, and ambiguous goals live. Judgment is the scarce resource now, and it is not something you can download or fine-tune into existence. You have to build it the old-fashioned way: by making decisions, watching them fail, and adjusting. The companies that figure this out will not be the ones with the best models. They will be the ones with the clearest sense of what is worth building.
The $2 trillion figure, if it holds, would make Anthropic’s offering roughly four times the size of the largest public offering in history. That is not a valuation of a company. It is a wager that every enterprise will eventually pay a toll to the layer that sits between raw capability and actual decision-making. The interns and indie hackers making millions are the canary. They prove the model layer is accessible. The IPO proves the enterprise layer is not. Coatue’s early exit, four years ahead of its planned 2030 timeline, is the template: identify the infrastructure layer, buy in before the tollbooth is built, sell when the tolls become obvious to everyone else. The remaining investors are not betting on Claude. They are betting on the decades of lock-in that follow when a capability becomes essential.
This pattern repeats every time a new technology drops to utility status. The foundations of the modern world were built in concentrated fifty-year windows when something new became cheap enough to deploy widely and expensive enough to control. The fifty-year windows were not accidents. They were the time it took for a new capability to diffuse from hobbyist to utility to monopoly. We are in the middle of that window for AI. The companies that survived those transitions were not always the ones with the best technology. They were the ones that owned the interface to it, the ones that turned capability into habit, and the ones that understood trust is the only moat that does not get replicated in a training run. Standard Oil did not own the best refining process. It owned the railroad rates. AT&T did not invent the telephone switch. It owned the copper pair running to your house.
Payments infrastructure still moves at the speed of trust, not the speed of code. Cashflows’ strategic investment in Tap & Go is a reminder that even in an AI-native world, the money-moving layer still needs consolidation. The rails underneath the intelligence still need to be reliable, regulated, and quietly profitable. Tempo Earn unlocking Stripe integration is another signal: the infrastructure bets are not just about the model. They are about who sits between the model and the transaction, who collects the fractional cent on every intelligent interaction that moves value. The payments layer is where AI meets the real world, and the real world still requires settlement. Someone has to be the counterparty when the algorithm decides to move money.
The intern’s $2.2 million is the proof that execution has never been cheaper. A single prompt, a capable model, a clear objective. That is the new minimum viable product, and it is accessible to anyone with a browser and a good idea. The $2 trillion IPO is the proof that the market believes most organizations will never develop the taste to build those things themselves. They will rent the interface instead, pay the toll, and call it transformation.
The scarce thing is not capability. It never was. It is the judgment to know which problem is worth solving, the patience to integrate it into systems that already work, and the honesty to admit when the bottleneck is not the tool but the person holding it. That gap between the intern and the IPO? That is not wealth inequality. That is the spread between raw possibility and the taste to direct it. The market is betting the spread stays wide. History suggests it usually does. The question is whether you are on the right side of the tollbooth.
Leave a Reply