— Opinion column by manufacturing expert Buckley Brinkman
There’s a fatal flaw in artificial intelligence: almost no one makes money with it.
The technologists say AI will transform our jobs, our businesses, and the way we live. The impossible becomes possible, and new tools will ease – or eliminate – our toughest problems.
The fatalists say AI will destroy careers, eliminate jobs and doom society. At the extreme, this camp predicts the end of humanity as AI-powered robots, agents, and networks take over the world.
Both camps create lively debate, but the economics will affect most of us more than any laser-eyed robot. Investors committed more than $3 trillion to building AI models and the infrastructure that runs them. They expect a return.
The window for that return is starting to close. Anthropic, Google, OpenAI, and Microsoft need profitable business models built on wider adoption, growing revenue, and more efficient models. The limited numbers we can piece together say it isn’t happening.
The frontier companies spend billions to train their models and billions more on the computing power to run them. Right now, those costs exceed what customers pay. To close the gap, the companies need more customers paying higher prices. Two forces stand in the way.
First, cheaper Chinese and open-source models keep taking market share. By revenue, the U.S. frontier companies still dominate. By usage, the picture flips: Chinese models now handle roughly 60% of the tokens – the units of text AI models process – on OpenRouter, a leading platform that routes requests among models. When a cheaper alternative does most of the work, the frontier companies can’t raise prices.
Second, customers aren’t making money with AI either, and that caps what they’ll pay. McKinsey’s State of AI in 2026 survey paints a somber picture. Eighty-nine percent of companies use AI. Only 37% report any impact on earnings, and just 6% say AI lifted earnings by more than 5%. That’s weak performance for a transformational technology.
I’ve had the good fortune to join numerous AI conversations over the past five years. Most focus on what the technology can do. Very few focus on how companies and individuals make money with it.
Even when participants think they’re talking about money, they’re usually talking about efficiency – helping individuals finish tasks faster. That’s only half the story. Efficiency improves profits only when costs leave the organization or revenue grows. An employee who saves an hour on reports or email creates a return only when the company puts that hour to work cutting costs or winning business. Most companies never take that step with AI.
Wisconsin business leaders can close that gap. You can start with the next AI project on your desk. Before you approve it, ask one question: Where does the money show up? Name the cost that will leave the building or the revenue that will come in, put a number on it, and assign someone to deliver it.
If the project frees up time, decide in advance what that time will do – serve more customers, take on new work, or absorb growth without adding headcount. Then track the result on the P&L, not in a productivity survey. Check back in six months. If the financial return hasn’t materialized, change the plan or kill the project and find a better project.
The AI companies need their customers to make money. You do too. Stop asking what AI can do, and start asking what it will earn.
