- According to Bank of America, utility prices are falling after years of hikes.
- However, the company said a sudden increase in demand for electricity had led to a sharp increase in electricity prices.
- BofA noted that AI could boost electricity demand for years or even decades.
Lower utility rates have given consumers some breathing room, but that relief could be negated by artificial intelligence's massive demand for electricity, Bank of America reported Tuesday.
The bank said in a new report that median payments fell 1.4% from a year earlier in the three months to May, a welcome decline for consumers who have faced rising costs over the past two years.
While prices are still well above 2019 levels, analysts said the negative price growth cannot be ignored. On average, households earning less than $4,000 a month spend about 8% of their income on utility bills, according to internal data.
But there are already signs that the grace period may not last long.
Bank of America noted that while payments are down, electricity prices, a major driver of utility bills, are rising sharply: Through May, electricity prices were up 5.9% year-over-year, up from 3.8% in January.
This rise is due to the increasing demand for electricity from various sources.
Warmer weather is also driving up electricity consumption as people turn up their air conditioners and use heat pumps more frequently at home.
Legislation such as the Contract with Inflation Act and the CHIPS Act also contributed, as they not only helped boost the adoption of electric vehicles but also led to a surge in electricity-hungry domestic manufacturing, the bank said.
“Since it takes a long time to build large-scale power plants, we may see many more being built in the future,” he predicts.
However, Bank of America noted that AI is the single most important driver of rising utility costs over the next few decades.
While the technology's capabilities have fascinated Wall Street in recent years, the bank stressed that using AI comes with incredible power consumption.
“For example, Bank of America Global Research shows that AI-based text classification takes about 2 kilowatt-hours per 1,000 tasks, while AI-based image generation takes about 3,000 kilowatt-hours for the same number of tasks,” the analysts wrote. “For comparison, the average U.S. household consumes about 886 kilowatt-hours (kWh) per month.”
Moreover, the semiconductor chips that run AI software run much hotter than traditional chips and require liquid cooling, which is also a power-hungry system.Additionally, there is the expansion of cloud servers that also help process AI services.
Bank of America projects that an additional 18 to 28 gigawatts of power capacity will be needed by 2026. Rising demand for cloud servers is already driving up commercial electricity demand in states like Virginia and Texas.
Both states are major hubs for data centers operating this technology.
“Voltage in global commodity prices, particularly fossil fuels, is likely to continue to influence the prices of energy paid by consumers, while the need for additional capacity in the generating system could act as a headwind to longer-term declines in utility prices,” the bank said.
For investors, the growing demand for electricity is drawing attention to utility stocks, which are expected to be the real winners from the AI boom.
