Conclusion

AI’s electricity costs will ultimately impact your company’s utility bills.

The power consumed by AI data centers is no longer just an “IT industry issue.” It strains local power infrastructure, drives up electricity prices, and ultimately hits fixed costs for small and medium-sized enterprises in rural areas. Let’s trace this pathway with specific numbers.

The catalyst was Amazon CEO Matt Garman’s announcement of the “abolition of NDAs (non-disclosure agreements) regarding data center construction.” At first glance, this move appears to enhance transparency, but it is rooted in the growing backlash from residents against data centers worldwide. At the core of this backlash is the “competition for power.”

The Real Reason Amazon Abolished NDAs

Until now, Amazon has concealed its data center construction plans from local residents under NDAs. The location, scale, and power consumption — all kept secret until a massive facility suddenly appears next door. It’s no surprise that residents are angry.

In fact, moratoriums on data center construction are spreading across the United States. In Loudoun County, Virginia, Georgia, and Indiana, construction has been halted due to referendums and municipal decisions. Similar movements are occurring in Europe, particularly in Ireland and the Netherlands.

Amazon’s decision to abolish NDAs is not an act of goodwill. It is a strategic move to prevent construction from being halted altogether. By increasing transparency, they aim to alleviate backlash and continue with construction — this is the business judgment at play.

However, we must consider, “Why are residents so angry?” While there are issues of noise and aesthetics, the primary concern is “electricity.”

The Power Consumption of AI Data Centers is on a Different Scale

The power consumption of traditional cloud data centers and AI-focused data centers is on a different level.

• Traditional data centers: tens of MW (megawatts)
• AI-specific data centers: hundreds of MW to 1 GW (gigawatt)

According to the IEA (International Energy Agency) 2024 report, global data center electricity consumption was approximately 460 TWh in 2022. This is projected to reach up to 1,050 TWh by 2026, nearly equivalent to Japan’s annual electricity consumption (around 900 TWh).

The primary driver of this rapid increase is AI. A single search on ChatGPT is said to consume about ten times the electricity of a Google search. GPU clusters used for AI training consume 5 to 10 times more power per rack than traditional servers. Cooling also requires enormous amounts of electricity and water.

The question is, where does this electricity come from?

The Mechanism by Which “Competition for Power” Raises Electricity Prices

For small and medium-sized business owners in rural areas, AI data centers are neither direct clients nor competitors. However, they become “invisible competitors” fighting for the same power within the same transmission grid.

The mechanism works as follows:

1. Sudden Increase in Demand → Rise in Wholesale Electricity Prices

Data centers secure large amounts of electricity through long-term contracts. Local electricity demand surges, causing wholesale electricity market prices to rise. Reports indicate that electricity demand in Georgia increased 17-fold after attracting data centers.

2. Transfer of Infrastructure Investment Costs → Increase in Transmission Fees

To meet the surging demand, enhancements to transmission lines, the establishment of new substations, and the expansion of power plants are necessary. Dominion Energy, a power company in the U.S., is planning to invest billions of dollars in transmission infrastructure in Virginia alone. These costs are distributed as transmission fees among all local electricity consumers. Not just data centers, but also the town’s bakeries and factories pay the same rates.

3. Competition for Renewable Energy → Rise in Green Power Prices

Amazon, Google, and Microsoft advocate for “100% renewable energy.” If they monopolize local renewable energy, the costs for small and medium-sized enterprises to procure renewable energy will also rise.

Through these three stages, the electricity consumption of AI data centers structurally drives up local electricity prices.

How Much Will It Actually Increase?

So, how much will this impact the wallets of small and medium-sized enterprises?

While this is an example from the U.S., the structure applies to Japan as well.

• New York State: Industrial electricity rates have risen by about 20% over the past three years, attributed in part to increased demand from data centers.
• Virginia: Due to the transfer of infrastructure investment costs, residential electricity rates are expected to rise by $100 to $200 annually.
• Georgia: Power companies are applying for rate increases, citing infrastructure investments to accommodate data center demand as one reason.

In Japan, data center construction is accelerating in areas like Hokkaido and Inzai City in Chiba Prefecture. The Ministry of Economy, Trade and Industry estimates that domestic electricity consumption by data centers will double by 2030. Japan’s electricity prices are already on an upward trend due to rising fuel costs and renewable energy surcharges, and the addition of data center demand will further exacerbate this trend.

For a small factory with an annual electricity bill of 3 million yen, a 20% increase would mean an additional burden of 600,000 yen. For a company with a 5% operating profit margin, this equates to sales of 12 million yen. This is not just an “AI issue”; it is a matter of “fixed costs.”

What Should Local SMEs in Japan Do?

You might think, “Big companies consume electricity, causing prices to rise. But there’s nothing we can do about it.” However, there are actions you can take.

1. Visualize Electricity Costs and Review Contracts

First, review your electricity contracts. Since the liberalization of the electricity market, there are more options for new power suppliers and pricing plans. Time-of-use rates, demand control, and optimizing contracted power can often yield cost reductions of 5-15%. For an annual bill of 3 million yen, this could mean savings of 150,000 to 450,000 yen. Start here.

2. Introduce Self-Consumption Solar Power

Install solar panels on your roof for self-consumption. This is about self-consumption, not selling electricity. With current panel prices, the return on investment is 7 to 10 years. As electricity prices rise, the payback period will shorten. Ironically, the more data centers raise electricity prices, the more economically viable self-generation becomes.

3. Shift to Being a “User” of AI

There’s no point in lamenting the electricity consumption of data centers. Instead, it makes more sense to leverage the AI powered by that electricity as a tool for your business. With AI tools costing a few thousand yen per month, tasks that previously cost 300,000 yen per month to outsource can be handled in-house. If you are on the paying side of AI’s electricity costs, you should at least position yourself to maximize the benefits of AI.

4. Raise Your Voice in Local Power Policies

In the U.S., residents are imposing moratoriums on data center construction. In Japan, it is necessary to question local government policies regarding invitations to data centers: “What will happen to the costs of power infrastructure?” and “Have you calculated the impact on local electricity prices?” While attracting data centers creates jobs, the actual number of jobs is relatively small compared to the size of the facilities. We should calmly assess the balance between electricity consumption and employment.

“AI’s Electricity Costs” Are Not Someone Else’s Problem

Amazon’s abolition of NDAs is because they can no longer keep it hidden. The power consumption of data centers is now at a scale that cannot be concealed.

And that electricity cost will eventually reflect on the fixed costs of small and medium-sized enterprises in rural areas. The transmission grid is interconnected. The electricity market is interconnected. Whether you use AI or not, electricity prices will rise.

Therefore, we must act on the assumption that prices will rise. We need to implement measures to reduce electricity costs, use AI to enhance productivity, and absorb the increase in fixed costs.

“AI’s electricity costs” are not just a concern for tech companies; they are a matter that will affect your company’s electricity bill next month.