Don't Worry About the Tokens
The focus on token economics and business models is completely misplaced. Investors debating them, bullish or bearish, are missing the big picture.
The Token Debate
There's a bunch of theses out there about tokens and token costs as the driving economics of the AI industry. Some of them are bullish. Some of them are bearish.
One camp believes that token costs are going to be dropping. Therefore that'll impact margins negatively, and you're going to see some domino effect from the customer-facing AI providers all the way through the value chain, all the way back to the cards and the chips.
Another camp argues the opposite. Maybe token costs will go up in the near term, because of geopolitical reasons, or supply, or whatever your thesis is.
But the reality is that this whole thesis, whether you're bullish or bearish, and whatever the argument is around token costs, is the wrong focus. It's pretty nonsensical for investors today to be focused on tokens and token costs as the driving economics of the AI industry. You're missing the big picture.
Tokens Are Irrelevant
Tokens are an input. They're just a raw material, not the product.
Customers don't want tokens. They want work done. And the way that work is going to get delivered is changing the economics entirely.
AI Providers Are Turning Into Fully Formed Software Platforms
What's relevant is how these AI providers are turning into fully formed software platforms.
They're becoming operating systems. They're becoming the apps running on the operating system. And they're becoming the AI running the apps.
So, pretty soon, users will just be using these AI operating systems. They won't be thinking about models, APIs, or tokens. They'll be thinking about getting things done inside an environment.
The Next Step: AI Employees
And it's only one step further for these AI operating systems to become AI employees, AI staff, AI divisions.
At that point, a business isn't buying compute. It isn't even buying software. It's hiring. It's bringing on an AI employee, or an AI team, or an entire AI division, to do a job.
And once you get to that point, tokens don't matter. A company doesn't care what the token bill is behind its AI employee any more than it cares about the electricity bill behind its accounting software. It cares whether the work gets done.
Who Owns the End User
So the real question isn't what tokens will cost. It's who owns the end user. Who owns the user interface. Who owns that AI employee that you're hiring. Who is providing that to you?
Is it a cloud provider? Is it one of these AI model vendors? Is it some enterprise software company like Microsoft? Can they do it? Is it Anthropic who creates their Claude employees? Or is it someone else entirely?
That's where the value in this industry is going to land. Whoever owns the relationship with the customer, with the AI employee doing the work, captures the economics. Everyone behind them in the value chain becomes a supplier.
What Investors Should Focus On Instead
Stop focusing on tokens. Focus on:
- Who owns the user interface. The AI operating system that users live in every day.
- Who owns the end user. The depth of the relationship, and how hard it is to switch.
- Who can deliver AI employees. Not just models, but fully formed AI staff and divisions that do real work.
- Who has the distribution and trust to get those AI employees hired inside enterprises.
At some point, all of these questions become the same question.
Eventually
Whether you're bullish or bearish on token costs, the token thesis is the wrong frame. Tokens are irrelevant. The winners in AI will be the companies that own the AI operating system, the apps, the AI running the apps, and ultimately the AI employees that businesses hire.