AI crashes the power grid, triggers electricity bills? Trump calls on tech giants to build their own power plants to lower electricity prices

The American tech industry is staging a new spectacle: a large number of self-owned power plants are rising rapidly. Due to the surging demand for computing power and the strain on severely aging power grids, building independent power plants and self-generating and self-consuming will become standard in this round of AI investment boom in the U.S.

“Many Americans are worried that the energy demands of AI data centers could unfairly raise their electricity bills. I am pleased to announce that I have negotiated a new ‘Power Consumers’ Guarantee Commitment.’ I told major tech companies they are responsible for solving their own power needs. They can build their own power plants, so everyone’s electricity costs won’t be increased. In many cases, community electricity prices could even drop significantly—an unprecedented and unique strategy in the U.S.!” President Trump promoted this plan during his State of the Union address on the evening of February 24, claiming that the U.S. power grid is too outdated to handle such massive electricity consumption, “so I told tech companies they can build their own plants and generate their own power. This not only ensures that businesses have electricity but also can lower prices for others, with potentially very obvious effects.”

However, Trump and the White House have yet to provide specific details of the plan. Afterward, Energy Secretary Chris Wight told POLITICO that the government has reached agreements with all “notable” AI companies. Besides covering the costs of power generation for data centers out of their own pockets, these tech companies will also “prepay some funds for grid expansion.”

They overlook a key fact: the costs of building high-voltage transmission lines will ultimately be passed on to consumers’ electricity bills. Despite Trump’s vigorous promotion of his promises and tech companies’ willingness to bear more costs, ordinary Americans will still pay for the bubble of data center prosperity. Even if data centers can generate their own power, the costs of large-scale upgrades to the power system won’t just disappear.

In the AI wave, power shortages are more critical than chip shortages. As the battle for electricity intensifies, Americans’ concerns about energy affordability are growing, becoming a thorny issue for Trump since his second term.

One of the largest utility companies in the U.S., AEP, based in Ohio, recently stated that as power supply agreements with data centers doubled, the company is expanding its five-year capital expenditure plan to over $72 billion. Additionally, the company plans to invest billions of dollars in new transmission and generation projects.

The PJM Interconnection, which operates the grid covering 13 states in the Mid-Atlantic and Great Lakes regions, has approved $11.8 billion in new transmission projects, with data centers being the biggest beneficiaries. The added costs will be distributed among 67 million people in the PJM area, roughly double the previous two transmission budgets. Exelon, one of the major utilities in the PJM region, has seen electricity prices rise over 20% in some areas last year due to data center demand outpacing supply growth.

In December last year, PJM’s capacity auction for 2027/2028 set a record high price of $333.44 per megawatt-day, up 1.3% from 2026/2027, reaching the federal Energy Regulatory Commission (FERC) approved cap, reflecting the supply gap caused by the surge in data centers and the high costs of system reliability.

According to Fox News, tech giants like Amazon, Google, Meta, Microsoft, xAI, Oracle, and OpenAI will gather at the White House on March 4 to sign the aforementioned agreement. White House spokesperson Taylor Rogers claimed that participating tech companies will “build, introduce, or procure power supplies for new AI data centers,” ensuring that as AI computing demands grow, American residents’ electricity bills will not rise accordingly.

Faced with exponentially expanding computing power needs, America’s aging and overstressed power grid facilities are increasingly unable to keep up, and power reliability issues threaten the deployment speed of AI by tech giants. Unable to wait for grid expansion and the more than five-year grid connection waiting period, AI giants are bypassing the grid altogether by building their own gas-fired power plants, directly fueling a boom in gas turbine orders.

Tesla CEO Elon Musk once criticized the U.S. power infrastructure when xAI’s massive data center, Colossus 2, in Memphis, Tennessee, faced difficulties. Colossus 2 will feature the first gigawatt-scale training cluster, but the voltage is only 300 kV. “There are several high-voltage transmission lines next to this building, but connecting these lines takes a year. We had to piece together 1 GW of power using many gas turbines and large battery packs.”

One of the world’s top three gas turbine manufacturers, Siemens Energy, recently disclosed that its first fiscal quarter of 2026 (October-December 2025) continued strong momentum, with net profit soaring from €252 million last year to €746 million. Thanks to sustained demand for gas turbines and grid equipment, orders increased by 34% to €17.609 billion, with backlog reaching a record €146 billion. Profit (excluding special items) grew 141% year-over-year, with profit margins expanding from 5.4% to 12.0%.

Financial data shows that the U.S. contributed the most to Siemens Energy’s order and revenue growth this quarter, with increases of 59% and 25% respectively. In the first quarter, the U.S. accounted for 40% of Siemens Energy’s gas turbine orders and was the largest growth driver for its grid business. In early February, Siemens Energy announced a $1 billion investment to expand its manufacturing capacity for grid and gas turbine equipment in the U.S.

U.S.-based GE Vernova is also experiencing unprecedented capacity shortages. By the end of 2025, its backlog of gas turbines and capacity reservation agreements increased from 62 GW to 83 GW. CEO Scott Strazik said on January 28 that the company expects its gas turbine backlog to reach 100 GW by the end of this year, with capacity for 2029 and 2030 nearly sold out.

According to Global Energy Monitor, the U.S. currently leads the world in under-construction natural gas power capacity, with over one-third planned to directly supply data centers. By 2025, U.S. under-construction natural gas capacity will nearly triple to about 252 GW. If all these plants are completed, the U.S. natural gas capacity will grow by nearly 50%, with capital expenditures exceeding $416 billion.

Nuclear power is also experiencing a new boom amid the tech companies’ “self-owned power plant” craze. Since 2024, tech giants like Oracle, Google, Microsoft, Amazon, and Meta have been actively involved in developing small modular reactors (SMRs) through partnerships with energy companies, purchasing power from nuclear plants, and investing in nuclear energy firms to meet the surging electricity demand.

Many Democratic lawmakers and some clean energy organizations believe that Trump’s measures are insufficient to protect public interests. CNN reported at the end of last year that the rapid expansion of data centers in the U.S. requires massive amounts of electricity. Some Democrats have begun calling for comprehensive restrictions on new data center construction.

The risk remains that if Silicon Valley’s grand AI ambitions fail, ordinary electricity users could face potential economic burdens. “You spend $3 billion upgrading the grid, but if the data centers are not built or their power consumption doesn’t meet expectations, who will bear the losses?” questioned Abby Silverman, an energy researcher at Johns Hopkins University.