Will That Be Separate Checks?

FRIDAY, OCTOBER 9, 2026

We don’t need to settle the future of AI before asking who will pay for it.

Government support has helped an emerging industry expand across the country. Its networks are changing the way Americans do business. It seems larger consumers will enjoy more favorable treatment than smaller ones. Smaller consumers worry their lack of bargaining power will translate to higher service costs.

The emerging industry makes economic explanations that leave those smaller consumers less than impressed. Some state governments have tried to intervene, but regulating a business that crosses state lines is tricky and requires more than local action.

All the while, people worry that this emerging industry is changing the landscape and bringing a much faster pace to the world in general. Certainly, it brings its obvious benefits, but there are too many unknowns and “potentials.”

It’s 1887, and railroads were on many minds.

If you thought we were about to do a dive into AI, it wouldn’t be surprising. What was happening with railroads in the United States in 1887 bears some similarities to today's concerns about AI.

Before 1887, farmers and small businesses had spent years protesting railroad charges that favored big shippers. They received discounts and rebates, but smaller customers paid more for shorter routes, effectively underwriting the railroad’s investments in infrastructure and expansion. If a community was served by a single railroad, there were no options. Communities turned to their state and federal governments to create rules enabling a fair chance to compete and benefit from this new technology.

◆

1887

Congress agreed and, in 1887, decided railroads needed some supervision. The Interstate Commerce Act was the result. It required railroads to charge “just and reasonable” amounts for their services, blocked some discriminatory practices, and established the ICC (Interstate Commerce Commission).

The new laws were not instantly effective, and enforcing some provisions proved difficult. While Congress agreeing on “fairness” had been a significant accomplishment, getting the railroad industry to follow the new laws would require much more work.

Still, a precedent had been set: Farmers and small businesses had spent years demanding change, and now people who couldn’t bargain their way to better treatment had helped create a means of obtaining it. Their complaints forced the government to accept its responsibilities.

America's size demanded ambitious networks, but governing them required institutions that could adapt as they grew, building the plane while flying it, as it were. Congress could use its existing authority over interstate commerce to address the rapidly changing economy.

The next adaptation came almost 100 years later, with the Public Utility Regulatory Policies Act of 1978. It created electricity rate standards, including one that linked charges to the cost of serving different customers. The agencies and utilities covered by the law would now have to announce public hearings and give people a chance to weigh in on proposed electricity-rate standards. The utilities wouldn’t be required to adopt the standards, but they had to explain, in writing, what they decided and why.

Directly addressing the question established a framework that lets us examine what an industry is doing today without first deciding whether the industry itself is good or bad. Certainly, a public process doesn’t answer every question. It can require answers about who should pay for what. It should offer a chance to examine, and even challenge, the proposed arrangement.

Perhaps this is where the size of the AI discussion does us a disservice. Artificial intelligence brings practical and physical concerns, along with its usefulness. They are separate from the ethical and philosophical concerns about AI and its impact. Data centers are the ideal illustration. These facilities are packed with the computers necessary for operating the internet. They store information and run services like email, websites, and AI.

AI companies use them to process users’ requests and questions and to train AI systems. Operating so many computers demands a lot of electricity for both running and cooling them.

So, AI is known as a high consumer. Congress is attempting to address this by amending the existing laws. In the late 19th century, railroads sold transportation, but today data centers buy electricity, so the problems those laws address are different, but the public still has a vested interest in who pays the bills and how that is decided.

It’s impossible to view AI through utilities alone, since AI links to jobs, information, and control over the technology itself. Fortunately, history shows we don’t need to settle the future of AI before asking who will pay for it.

417–3

On September 16, the House passed the Ratepayer Protection Act nearly unanimously. A few weeks later, the Senate failed to advance it.

The bill would require state electricity regulators and some utilities to consider a rule requiring data centers to pay for the additional infrastructure they would need. But the bill would also leave decision-makers free to reject the rule.

In plain speak: Congress would demand they address who pays, without requiring data centers to cover the costs.

While some in the Senate argued for advancing the House-approved measure, others demanded binding requirements. Several introduced a broader permitting proposal that would require data centers to cover their own transmission costs, not simply consider the idea. Another bipartisan group of senators proposed a separate bill requiring data centers to pay the bills for the upgraded utilities they’ll require. The initial bill has stalled, but it wasn’t the only proposal. Congress is running the hypotheticals.

Suppose an electric provider builds the substation necessary for a new data center only to have the project canceled. Unless the data center company is required by law to foot the construction bill, some of the cost could end up on other consumers’ electric bills. It is a potential risk, not something we can assume will happen.

But what we can do is ask data center companies to take the financial gamble, rather than household consumers and smaller businesses. They may be asked to pay before construction, set aside funds for changes in the plans, and even agree to keep funding their share should they abandon a data center. These are simple solutions that can be examined and agreed upon without also deciding whether or not AI will ruin things or make everyone wealthy.

One crypto expert, Arthur Hayes, said he believes firms are already overspending on AI, which he sees as a buildup to an investment crash and a subsequent government bailout.

Bill Gates, on the other hand, asked what happens to the taxes workers pay when robots replace them during his interview with The New York Times on September 29. In it, he pointed out how workers and employers pay taxes that support today’s Social Security recipients. He then pointed out that replacing people with robots means reduced wages and less funding for retirees. His solution was to suggest firms using robots pay something like payroll taxes for their labor. In this approach, replacing a worker wouldn’t eliminate the contribution to Social Security.

Clearly, retirement, electricity, and government rescue are not the same and require different answers. The commonality is that we must find solutions before any funding shortage arises. We don’t need to argue against AI to make decisions about who pays for it. We also need to know what AI companies plan. This, too, is something Congress can address, and is already exploring.

On September 24, a bipartisan group introduced the No Secrets for Data Centers Act. This legislation would block secrecy agreements hiding project details, including any agreements involving public money and officials. In 1887, knowing the terms before challenging them was useful, and the same process underpins this bill.

Amazon has recently indicated it will no longer allow NDAs in its agreements with utilities. They describe this step as a company promise, but the proposed law would make such disclosures mandatory.

On October 7, Duke Energy announced an agreement with North Carolina’s state office representing utility customers, Google, Meta, Microsoft, and Amazon. The larger customers would now pay upfront for any electrical equipment built for them, with no refunds possible. They would also have to guarantee and deposit funds for infrastructure improvements that the energy company would need, even if shared with other customers. The deal is not yet approved, but the Southern Environmental Law Center has examined it and described it as too weak to protect other customers.

Ohio has already acted on data centers, and in July 2025, following rounds of hearings and public proposals, approved an agreement supported by AEP Ohio and the office representing household electrical consumers. The agreement keeps other customers from paying for any equipment built for data centers, as well as allowing AEP Ohio to resume new data center connections.

The issue of paying for electricity does not answer consumer questions about the water, noise, heat, or whether data centers even belong in specific locations. A hearing doesn’t guarantee fairness. Instead, it gives people a place to question plans, present and receive evidence, and demand or ask for changes before construction begins. Managing data centers today means pressing for better rules; it doesn’t mean we have to know how to control everything.

During a 1977 speech on energy, President Billy Carter said: “By acting now we can control our future instead of letting the future control us.” He was responding to a problem of his time, not predicting AI. He was telling us to recognize choices we still have. We can’t settle the future of AI in utility hearings, but we can’t ignore the importance of the hearings.

By acting now we can control our future instead of letting the future control us.

PRESIDENT JIMMY CARTER, APRIL 18, 1977

America is big and always has room for big ideas, but a neighborhood or community isn’t just a space on a map ripe for data center development. Making way for something new means looking at the lives and landscapes it changes.

Farmers challenged railroads and their rates over a century ago, helping us create ways to challenge enormous and powerful companies and demand fairness today. The Carter administration recognized that policy should plan for the future. History has shown us that efforts to turn principles like fairness into rules can make those rules something people eventually use. That’s where today's work is; history tells us we are not finished just because a law passes. We don’t know what will happen with AI, but policymakers have started to think about tomorrow. We’re all invited to pull up a chair, and we’re still allowed to ask for separate checks.

Faithfully,