The utility company serving TSMC Arizona has industrial electricity rates 50% higher than those for Samsung Austin.

News of building a factory in the U.S. is often accompanied by figures for investment and job creation, but the electricity cost for that factory rarely follows. We obtained the coordinates of 29 U.S. semiconductor factories from regulatory records and identified the utility company responsible for each location. We then divided the actual revenue earned by each company from all industrial customers by the volume of sales to calculate the unit price. Differences that were not apparent at the state average level became clear when we examined each factory individually.

The electricity cost for factories cannot be generalized to the weekly average.

When discussing average electricity rates in the United States on a per-state basis, the figure is an aggregate of all utilities within a state, making it far removed from the actual cost a single factory must pay. Electricity sales in the U.S. are regional monopolies, so the choice of utility company is determined together with the location of the site. As a result, even factories built in the same state but a few kilometers apart may face different costs.

This difference is particularly significant in semiconductor factories where the processes of slicing, cleaning, and baking wafers operate continuously. Electricity is practically a raw material, so a 10% difference in unit price results in a corresponding annual cost difference.

Coordinates were obtained from regulatory agencies, and unit prices were derived from reported performance data.

When the location of a factory is taken from company press releases, it often only provides the name of the city. If the coordinates of the city center and the actual site differ by just two kilometers, the responsible utility company changes. Therefore, the coordinates of facilities that require discharge permits were taken from the U.S. Environmental Protection Agency's compliance monitoring records, which are registered by regulatory agencies.

Similarly, unit prices were not estimated but derived from the annual performance reports submitted by utility companies. The sales performance reports received annually by the U.S. Energy Information Administration from utility companies include both the revenue received from industrial customers and the amount of electricity sold. Dividing the revenue by the amount of electricity sold yields the actual unit price received by the company. This is not the value listed in the tariff but the actual value recorded in the company's annual accounts.

Fabs clustered in Texas produce almost the same value

When 29 locations are listed in order of cheapest unit price, Texas is concentrated at the top. Samsung Austin, NXP Austin, and Texas Instruments Dallas factory fall into the same utility company territory and are grouped under the same unit price of $0.0683. Texas itself has an industrial average rate that is cheaper than any other state listed in this table at 7.07 cents.

Even within Texas, there are divisions. Texas Instruments' Richardson plant is in a different city and utility company, placing it in a higher-cost area at 0.0770 dollars, compared to the company's Dallas plant. Even if a company operates multiple factories in a single week, electricity prices are set separately for each factory and are not grouped together.

Only three locations have higher-than-average prices, and TSMC Arizona is among them.

Out of 29 locations, 26 have unit prices from their utility companies that are lower than the average for their respective states. This is interpreted as factories that consume large amounts of electricity selecting cheaper areas. To make such a definitive conclusion, records must exist showing what factors were considered when choosing a site. Only three locations exceed the average: Texas Instruments Richardson, 3M Columbia, and TSMC Arizona.

TSMC Arizona's unit price of 0.1021 dollars is 9% higher than Arizona's average of 9.35 cents and 52% higher than the lowest price in the table, Texas Instruments Stafford. Considering it is the most prominent semiconductor investment in recent U.S. history, this price is notable.

This record does not answer why, but the table itself partially answers that question.

In selecting a state, they choose a cheaper region, but not when selecting the state itself.

Given such differences in electricity costs, it is a natural question to ask why they were built there. The fact that 26 locations have a unit price lower than the state average suggests that after selecting the state, they moved toward the cheaper utility company regions. This is a significant skew, unlikely to be coincidental, but there are no records explaining the reason for the site selection.

The remaining half points in the exact opposite direction. The picture changes when considering the stage of selecting the state. Since both the cheapest Texas and the most expensive Arizona have large fabrication plants, at the stage of selecting the state, electricity costs were not a leading factor. The table aligns better with this interpretation. Electricity costs appear to have become a decisive factor only at the stage of narrowing down the site after the state was chosen.

How much this difference amounts to annually depends on how much electricity the factory consumes. Assuming a large fabrication plant uses 1 terawatt-hour per year, the unit price difference between the two regions corresponds to $33,800,000 annually. Since the amount of electricity used is not disclosed, this number is an estimate to gauge the scale. While it may not be a large sum compared to the cost of building a single factory, it is not insignificant when repeated every year.

To go a step further, records containing the company's stated reasons are needed. We looked for the reason Arizona was chosen in TSMC's annual report, but found no such statements, and could not confirm the amount of funding allocated to this factory in federal grant records either. Therefore, this article does not speculate on reasons such as water supply, labor, or tax incentives.

The most significant reduction was at the solar factory in Georgia.

When 29 locations are rebuilt based on the state average, Hanwha Q Cells Dalton plant has the lowest at 0.72. While the industrial average in Georgia is 10.03 cents, the factory's utility company receives 7.23 cents. Both the Micron Manassas at 0.74 of Virginia's average are in high-cost states but selected low-cost areas.

GlobalFoundries in Vermont has an absolute value of $0.1084, which is on the higher side in the table, but compared to Vermont's state average of 13.07 cents, it becomes 0.83. The fact that the absolute value and the state average ratio tell different stories is the most important point to be cautious about when reading this table.

23 locations were excluded because they could not be counted.

Some U.S. states allow competition in electricity sales, so utility companies only handle transmission and distribution, while factories purchase electricity directly from selected providers. In this case, only the distribution fee remains on the utility company's books, and the unit price is recorded as a fraction of the actual cost. One utility company in Ohio reported 1.8 cents for industrial use, but the state average is 10.95 cents.

The actual cost paid by the factory lies in the contracts made with competing businesses, and these contracts are not disclosed anywhere. Therefore, this article excludes such cases, and 23 factories, including Intel Ocotillo and Onsemi Eastfishkill, are listed outside the table. Only factories where generation and sales are tied to a single company, revealing the cost in a single ledger, remain in the table.

What this table does not say

Before comparing this table with other data, there are three points to clarify. First, the unit price listed here is the average received by the utility from all industrial customers, and thus differs from the bills of large factories negotiated separately.

Second, the industry classification is based on self-reported information from the facilities, so large fabs and small electronic component factories are mixed under the same code. The six facilities concentrated in California are manufacturing sites for printed circuit boards and adhesive films. Third, this table only covers the United States, so the company's factories in South Korea or Taiwan cannot be counted in this chain.

What readers should verify next

Sales performance reports are issued annually, so by repeating the same calculation with the following year's data, it is possible to directly compare how the values have changed. Whether the utility company raised industrial rates or whether factories changed their contracts becomes clear in the numbers a year later.

First, look at TSMC Arizona, where expansion is ongoing. By checking whether the industrial unit price from the same utility company remains the same or changes due to an increase in major customers, you can see the trend. Next, look at factories in the retail competition zone, where unit prices can be calculated from the time the rate structure changes.