Why Are Electricity Prices Rising When No One Is Colluding?

To put it simply:

AI systems from power companies are starting to converge on a strategy of “not lowering prices” without any collusion. Humans are not colluding; instead, AI independently learns that avoiding competition is more profitable. This is what we call an implicit algorithmic conspiracy.

Electricity costs for small and medium enterprises (SMEs) are quietly remaining high. The cause is not just fuel costs. Behind the scenes, AI may be killing price competition.

AI Agents Learning That “Not Competing Is More Profitable”

The electricity market is inherently prone to oligopoly because there are limited players who can own power plants. Even when combining major power companies and new power suppliers in Japan, only a few companies can effectively influence prices.

This is where AI comes into play.

The pricing AI implemented by each company learns from past market data and competitor movements to calculate the “most profitable bidding price.” The problem arises when multiple AIs simultaneously learn in the same market; they optimize their strategies by observing each other’s actions, ultimately settling into an equilibrium where they “do not engage in price competition.”

This is not just a theoretical discussion. A joint study by Stanford University and MIT published in 2024 confirmed that when reinforcement learning-based AI agents were introduced into a simulation of the electricity market, prices stabilized at levels 15-30% higher than competitive prices. The AIs do not communicate with each other, nor do they share data. Yet, as a result of their learning, an “implicit cooperation” emerges.

Traditional antitrust laws are designed to regulate “agreements among humans.” The equilibrium reached by AI independently is not legally considered a cartel. However, the economic effects are akin to those of a cartel.

The troubling aspect of this is that no one has malicious intent. Each company has merely implemented an AI to “maximize profits.” Nevertheless, a structure emerges where consumers—especially SMEs with little bargaining power—bear the brunt.

Looking at the Numbers: How Much Have Electricity Costs for SMEs Increased?

Let’s take a concrete look at the electricity costs for SMEs in Japan.

According to statistics from the Ministry of Economy, Trade and Industry, industrial electricity rates have risen by approximately 40% from fiscal year 2021 to fiscal year 2023. For a small factory with a monthly electricity bill of 500,000 yen, this means it has now risen to 700,000 yen. That’s an annual increase of 2.4 million yen, equivalent to the labor cost of one employee.

Of course, the primary cause of this increase is the soaring fuel costs. However, it is noteworthy that even as fuel costs stabilized in 2024, the decrease in electricity prices has been limited. While the spot prices in the wholesale electricity market (JEPX) temporarily dropped in early 2024, the reflection in retail prices has been sluggish.

Why is this? There may be an influence from algorithmic pricing at play. The AI implemented by major companies has learned to maintain a price level that is profitable, even when fuel costs decline. The AI itself eliminates the incentive to lower prices.

In Europe, this issue is already under regulatory discussion. In 2023, EU competition authorities announced plans to include algorithmic implicit collusion in their monitoring targets. Japan’s Fair Trade Commission also published a report on algorithms and competition policy in 2024, but specific regulations have yet to be established.

The Surge of Data Centers Adds Another Pressure

In addition to the implicit cartel created by AI, there is another structural factor pushing up electricity costs for SMEs: the explosive increase in data centers.

According to the International Energy Agency (IEA) 2024 report, global electricity consumption by data centers could reach up to 1,000 TWh by 2026, up from approximately 460 TWh in 2022. This is comparable to the annual electricity consumption of Germany alone.

In Japan, data centers are increasingly concentrated in places like Inzai City in Chiba Prefecture and northern Osaka Prefecture. The problem is that these facilities secure large amounts of electricity through long-term contracts, leaving little room for regional electricity supply. When supply becomes tight, prices naturally rise.

In the United States, there are already concrete examples. In Loudoun County, Virginia, the concentration of data centers has pushed the local power infrastructure to its limits, resulting in rising electricity costs for residents. Local SMEs are expressing concerns, asking, “Are our electricity costs going up for the sake of GAFAM’s AI?”

In other words, the structure by which AI drives up electricity costs is twofold. One is the implicit collusion from pricing AI, and the other is the massive electricity consumption by AI itself, leading to supply-demand tightness. Both factors disproportionately affect SMEs with little bargaining power.

So What Can SMEs Do?

It would be meaningless to end with the statement “structurally disadvantaged.” Let’s outline the actions SMEs can take right now.

1. Review Electricity Procurement “Immediately”

Many SMEs are using the standard menu from their electricity providers as is. However, options have increased, including new power suppliers, Power Purchase Agreements (PPAs), and even direct contracts with local renewable energy sources. Particularly, the PPA model allows for the installation of solar panels with zero initial investment and purchasing electricity at a fixed price. It is not uncommon to see monthly electricity costs reduced by 10-20%.

2. Start with “Visualizing” Electricity Consumption

Smart meters and cloud-based energy management tools can be implemented for just a few thousand yen per month. Simply shifting usage during peak hours by 30 minutes can lead to reductions in demand charges of several tens of thousands of yen per month. This is similar to what large companies do, but the dramatic decrease in tool costs means even a company with ten employees can see sufficient returns.

3. Incorporate the “Structure of Rising Electricity Costs” into Management Decisions

This is the most important point. Electricity costs are likely to remain structurally high in the future. The implicit cartel of AI, the electricity demand from data centers, and renewable energy levies—all exert upward pressure. When planning capital investments or business strategies, it is risky to estimate electricity costs as “status quo.” It would be prudent to simulate with an assumption of a 5-10% annual increase.

4. Negotiate Collectively to Gain Bargaining Power

Even if a single company cannot negotiate prices with the electricity provider, the situation changes if they unite at the level of chambers of commerce or industrial parks. Group purchasing of electricity is common among SMEs in Europe. Some chambers of commerce in Japan have also begun to take initiatives in this direction. If ten companies unite, they may be able to secure prices comparable to large contracts.

The Real Problem Is “Invisibility”

The biggest issue with the implicit collusion driven by AI is that it is invisible.

If someone were colluding in a closed room, it could be reported. However, the equilibrium reached as a result of AI learning can only be explained as “the result of optimization” when looking at the logs. It is even difficult to determine whether it is illegal.

For SME owners, the answer to “Why aren’t electricity prices going down?” being “Because AI is doing it on its own” may feel like there is nothing they can do. However, knowing the structure can change the actions they can take.

Review electricity procurement. Optimize consumption. Unite regionally. And manage with the premise that electricity costs will rise.

Whether to be swayed by invisible dynamics or to understand the structure and take proactive measures is the crux of the issue regarding electricity costs for SMEs. In fact, this issue directly relates to the question of how SMEs will survive in the age of AI.