I believe the benefits AI can bring to humanity are immense and still largely unknown. However, AI expansion without regulatory structures and resource reallocation is a massive problem. The systematic removal of existing structures that protect both humans and nature amounts to corporate welfare dressed as innovation policy. This is the first in a series of articles about that problem.
For the sake of argument, let’s temporarily assume that AI is valuable. Before we can evaluate whether that is true, the assumption raises a cascade of questions that must be answered first.
What is value?
Valuable to whom?
At what cost, and paid for by whom?
What kind of value is being created?
From a generative frame, Value is a customer's assessment that an offer from a performer provides more satisfaction than any alternative, including doing nothing.
Let’s work with that for a moment.
Each of the current AI companies would be a performer. They are the ones offering a suite of services. They have customers who accept that offer by paying for AI services. They also have upstream customers (investors) hoping for a substantial return on their investment. We can safely assume that AI creates value for its subscribing customers, else they would not pay for the service. And for investors to get value, AI needs to generate more revenue from subscribers (and other sources) than what the investors contributed. The jury is still out on that last part, but assuming it comes to pass, how would that value be categorized?
Creative vs. Extractive Value
For Creative Value, the end state is more desirable to all stakeholders than the starting state. There is a performer responsible for creating this value, a customer assessing satisfaction, and all participants in the exchange are better off in a moderately closed system.
For example, as a Notion consultant, I might redesign a client's workflow and introduce new capabilities that the client did not know to ask for. If I exceed the client’s expectations, I will be rewarded with an improved reputation and additional business. Everybody wins, and this process can be repeated without negative consequences or side effects.
In the case of Extractive Value, the short-term end state is more desirable to some, but that comes at the expense of long-term value, or others who may not have participated in the agreement.
For example, payday lending apps are extractive. The borrower needs short-term cash and ostensibly benefits from the loan, but at a substantial and unsustainable cost. The borrower sacrifices their long-term financial health for their short-term needs, often falling into a vicious cycle. Also, when the borrower is part of a family system, the entire family is affected by predatory loan costs, not just the individual borrower, and interest paid to the lender seldom goes back into the local economy.So far, we’ve distinguished two factors that separate creative from extractive value. The first is customer satisfaction over a longer time horizon. The second is external players who bear the cost of the exchange when they were not part of the conversation leading up to the transaction.
Rather than looking at small-scale transactions between a consultant and a client, or a lender and a borrower, what happens when we widen our lens to larger-scale exchanges? In these cases, the customers and performers are part of a system of exchanges that cannot be evaluated at the individual level.
Utilities as Creative Value
The success of a public utility is not measured by profit. Instead, it is measured by how efficiently it provides a public good while still charging a fair rate to its customers. Examples include:
The water company
The power company
Public education
Fire departments
Power and water are paid services. The more a customer uses, the more they pay. They are also necessary services that are almost impossible for individuals to create. People living entirely off-grid are the exception here, but it takes a utility company to scale value by distributing to a larger customer base.
Education and fire departments are paid for, too, but through taxes rather than directly by customers. Further, we pay for these services whether we use them or not because society recognizes we are all better off with an educated public that does not burn to the ground.
Data Centers as Extractive Value
AI offers customers easy answers, deep analysis, automation, and speed. It offers people and businesses the opportunity to do more with less. It offers an opportunity to significantly cut costs, reduce overhead (staff), and synthesize vast quantities of information in ways that are difficult (if not impossible) for people to do on their own. That’s a compelling offer. Compared with the cost of a subscription, even at organizational levels, the value is clear. I am a customer myself, and part of my offering includes helping organizations get more value from their use of AI in their businesses.
But who is paying the long-term costs of what AI offers? Organizations no longer need to pay entry-level employees for work that can be automated. Organizations are now like the borrower in the payday-loan example, because those former employees no longer mature into seasoned professionals. Organizations are saving in the present by borrowing from their future.
Additionally, if we look at the real cost of offering AI services, none of the major AI providers (Anthropic, OpenAI, Google DeepMind, Meta AI, MS Copilot, Grok/xAI) are even close to profitable as of this writing.
The implication is that customers assess the value of the service only because they are not actually paying the full cost of operating it. That means that somebody else is paying that cost, and it’s only a matter of time before the piper wants his due.
In every single community where data centers are built, there is a pitched battle between citizens' desires and the demands of capital investment and opportunity. More often than not, capital wins, and it only does so based on the false promise that it makes genuine economic sense. So far, it does not. Instead, data centers represent the payday loans that communities are forced to bear against their will.
The entire business model of AI is inherently extractive and built on the hope that, at some point in the future, it will all pencil out for investors. Under the best of circumstances, investors will make massive gains that were largely paid for by the public, who will not have evaluated the true cost until it is way too late to change course.
This is not the first time this model has played out. In the next article, we’ll look at what we can learn from historical examples and the wolves of Yellowstone.
This is part 1 of a three-part series
1. AI Is a Payday Lender
2. Wolves of Yellowstone
3. Reentry: Lessons from Apollo