Big Tech’s AI Data Center Power Demands Trigger $15B PJM Auction and a Nuclear SMR Boom

TradingKey - The artificial intelligence buildout currently taking place across America is colliding with the physical infrastructure of the United States, which is not ready for this level of development, regardless of how advanced any of the chips used to build AI systems are. Furthermore, Washington has made a decision regarding who will pay for the development of AI and related technologies—hence this “reckoning” has forced those who wish to develop hyperscale sites to recognize their role in contributing to the infrastructure needs of the country. As a result, the structure of regional electricity markets, the financing of next-generation nuclear plants, and many other aspects of energy procurement have all been altered, causing energy taxes to become major expenses for the world’s largest tech companies.

The End of Hidden Subsidies

The flashpoint is the largest Regional Grid Operator in the US-PJM Interconnection Service Area stretches from Chicago to Washington D C. The combined impact of the boomlet in Data Center development and surging user demand for electricity over the last three years has caused an astounding 38% increase in Residential Energy Costs across the entire PJM footprint and a 13% increase just from 2020 to 2025 alone.

All of which created intense pressure from both the White House and State governments on PJM to conduct an Emergency Electricity Auction in January to compel the Big Technology Companies to underwrite the new capacity supply necessary to meet future Electricity Demand. This pressure culminated on March 4th with the signing of the Voluntary Tax Payer Protection Pledge by the Seven leading Hyperscale and Artificial Intelligence Companies (Amazon（AMZN）, Google（GOOGL）, Meta（META）, Microsoft（MSFT）, OpenAI, Oracle（ORCL） and xAI).

Although the Voluntary Taxpayer Pledge is not legally binding, it does create legally binding Financial Structures under U S Law, as all the Voluntary Taxpayer Pledge signatories agree to negotiate new Rate Structures dedicated to Data Center loads, where Data Center Operators will pay for all Costs related to Grid Upgrades and New Generation of Electricity supply to the Data Centers without passing those Costs on to ResidentalConsumers or non-tech Businesses.

Most importantly under the Voluntary Taxpayer Pledge, the Data Centers will pay for the capacity agreed to be purchased under Contractual Agreements to PJM , whether or not the Data Center Operator uses the Electricity that PJM must contractually reserve the capacity for. PJM anticipates that there will be a 5.2% Capacity Shortfall by 2027-2028, meaning that approximately $15 Billion of New Plant Investments will be required. The Emergency Electricity Auction is required to be completed prior to September 2026 with the intent that each Data Center Operator will sign Long-term Electricity Purchase Contracts that will effectively end the implicit Public Subsidy of the AI Infrastructure.

The Grid Bottleneck Forced Tech to Become Its Own Utility

PJM's electricity auction effectively changed power from a business-level detail and item into an item in a balance sheet that could delay project development. "As an example of this new burden," wrote Wedbush analysts in a market note, "Financing methodologies used to build data centres now may consider to be reflected in the additional costs associated with modifying the grid, such as on transmission lines, re-decking substations and installing new connections for data centre projects that would have historically been included in a utility's rate base." This means that datacenter electricity costs could increase 30% to 50% due to 15-year contracts, which will ultimately be passed down as higher prices for cloud and AI services.

There is little time to lose. At GTC 2026, Nvidia (NVDA) CEO Jensen Huang stated that the limiting factor regarding silicon is the ability to supply power to the chips, infrastructure, models or applications; all three of which, he claims, have scarce resources — as do land, electricity, and shell capacity. Consequently, the U.S. Department of Energy's SPARK program to expand transmission capacity by at least 50% (through reconductoring of existing lines and the addition of advanced technologies) will not be fully resolved until at least August 2026 when N/A funding will provide the final funding for this program starting late 2026 and early 2027. It will usually take 3–5 years to construct core transmission projects; additionally, if new dependable generation is not developed, the DOE has specified that the incidence of power outages could increase by 100 times by 2030.

With an inefficient grid, high costs to wait for a forced auction, and limited time to secure new sources of energy, hyperscalers have responded exactly as Huang predicted by building quickly and taking energy supply onto their own balance sheet. This strategic shift is the single strongest catalyst for the nuclear renaissance we are seeing today.

Big Tech Nuclear Deals Rewrite Reactor Finance

Instead of bidding on a small modular reactor (SMR) via auction and hoping for the best, the major players in the AI space are now writing direct checks to developers of reactors to help transform the SMR from being an idea that has not yet been regulated to being a fully funded commercial pipeline.

Some examples of this include: Meta has backed TerraPower, which is building a two-unit, 690 megawatt divided into two units nuclear unit (unit 1) at the site of one of its planned nuclear parks in Ohio and has entered into an MOU with Oklo to build an additional 1.2 gigawatt nuclear unit (unit 2) at the same site; Amazon is working with X-energy to deploy over 5 gigawatts of SMR at various locations throughout the U.S. by 2039; and Google is targeting to build a Kairos Power SMR in its designated area by 2030.

Additionally, in its pursuit of diversifying its energy supply with nuclear, Microsoft has added to its current efforts in the nuclear arena by layering nuclear agreements on top of existing agreements it has for combined cycle gas turbine generation and renewables. Together, the four companies have secured or will secure more than 10 gigawatts of nuclear capacity to support renewable energy by 2035 and more new opportunities are still being announced.

While the amount of megawatts associated with energy alternatives is important, much more extends beyond that. "Bringing a top tier corporate balance sheet into an industry that has historically been funded through a regulated, ratepayer backed revenue stream means a lot for the overall credit profile of that sector," said Shioly Dong, a senior analyst at BMI.

That credit profile will be required by commercial lenders and institutional investors to provide the financing required for construction. "The modular design concepts have a smaller scale and shorter timeframe for construction; therefore, they have a significantly lower level of upfront capital risk," said Tim Winter, portfolio manager for the Gabelli Utilities Fund（GABUX）. In addition, HSBC analysts stated that the technology sector's willingness to accept cost overruns or schedule delays will be a key factor in determining how much these agreements will actually facilitate the deployment of advanced reactors, as well as the expansion of fuel supply chains and advanced reactor technology parks, as these projects begin to attract direct investment, which has historically been thought of as unattainable from the standpoint of conventional nuclear financing.

Energy Storage Becomes the Non-Negotiable Buffer

Although nuclear generation and solar energy generation cannot meet the continuous service requirement that artificial intelligence (AI) workloads want due to the inability of a nuclear plant’s output capacity to ramp up quickly, long-duration energy storage systems have become an essential part of AI data center energy supplies as a result of this fact. Specifically, liquid-flow battery systems will likely be used together with both renewable energy generation and nuclear plants because of their favorable safety record, constant energy output, and long cycle life.

Given that as of September 2025, over 100 gigawatts of new-type energy storage systems had already been constructed in China, where the National Energy Administration has encouraged their development as a result of both government mandates for ‘synergy between computing and electricity’ and direct green-connection commitments from power generation sources.

Data center backup power needs have also helped to contribute to developers of storage systems in the United States repairing their valuation models during 2026. The result of load growth and instability in electrical grids is that batteries represent an important level of risk management and protection against electrical grid volatility.

The Nuclear Execution Gap and Regulatory Friction

The economics of small modular reactors (SMRs) are unproven on a large scale as evidenced by the Vogtle expansion in Georgia, which was seven years behind schedule and $18 billion overbudget (about $15,000 per kilowatt — nearly five times the cost of similar Korean projects). According to Edwin Lyman (a physicist with the Union of Concerned Scientists), there has been no significant improvement in the basic economics of nuclear energy and previous chairs of the Nuclear Regulatory Commission (NRC) have criticized efforts by the Trump administration to deregulate; they believe that a rushed growth process will lead to an increase in potential safety problems. A recent $80 billion partnership between Brookfield Asset Management and Westinghouse to design and build eight large nuclear reactors must overcome multiple hurdles related to financing and permitting. For investors, there is no expectation of short-term returns, as the current SMR order book is growing rapidly, but the cash payoffs will still likely take decades to materialize.

A Hard Infrastructure Bet in a Software Cycle

Investment activity around AI and clean energy is transforming the capital markets as we know them today. The initial investment of $12.5 billion made by BlackRock’s Global AI Infrastructure Investment Partnership in partnership with Microsoft (totaling $30 billion) will create the potential for $100 billion of projects to be financed through this newly formed fund once leverage is included. In addition to their investment in the AI Infrastructure Investment Partnership, BlackRock’s Infrastructure team made a $33 billion offer to acquire AES Corporation （AES）and provided $1.7 billion of minority equity to TC Energy, making them a direct investor in grid modernization and generation assets.

As an example, the iShares Global Infrastructure ETF (IGF), which provides broad-market exposure to infrastructure assets (including the aforementioned funds), has returned greater than 10% during the first two months of 2026, significantly outperforming the S&P 500 and turning a historically defensive investment into an aggressive thematic investment.

It is becoming increasingly apparent that the development of AI energy infrastructure has formed into a new investment area, which includes all of the following types of products; grid equipment, gas turbines, advanced nuclear reactors, and energy storage. Reservoirs are being filled with endless compute demand, but supply is limited due to permit timeframes, financing issues and physical limitations. The PMJ (Power Purchase Agreement) auction and DOE SPARK program (Department of Energy Smart Power Advancement Program) are some of the many tools to support this trend; however, it ultimately leads down to one understanding: large technology companies realize the grid will not solve their needs; therefore, they are going to solve their needs independently of the grid. So, the big question now is how many companies can afford a $15 Billion auction commitment, along with SMR (Small Modular Reactor) contracts and 10 Gigawatt projects before all profits are negatively impacted?