The largest electric utility in the United States, NextEra (NEE.US), is collaborating with the North American asset management giant Brookfield Asset Management (BAM.US). The project partners intend to invest over $100 billion to transform a decommissioned Cold War-era uranium enrichment facility in Kentucky into a massive data center campus, complete with a supporting power plant.
This project's blueprint, combined with the latest data center expansion plans from tech giants like Facebook's parent company Meta and Google's parent company Alphabet, provides a significant positive fundamental catalyst for the globally battered AI computing power theme. It also offers hard evidence from the industrial chain to empirically refute the "AI computing power glut" theory.
The U.S. Department of Energy announced on Wednesday that this fully privately funded project will include the construction of 2 gigawatts of natural gas power generation facilities and up to 2.6 gigawatts of battery energy storage capacity at or near the Paducah site in western Kentucky. For context, 1 gigawatt of installed capacity is roughly equivalent to the total output of a traditional large-scale nuclear power plant, capable of powering about 750,000 homes at any given time.
A joint statement from the parties indicates that the U.S. power giant NextEra and Brookfield Asset Management are cooperating to develop the Kentucky data center and supporting energy project, with total private investment exceeding $100 billion. This figure refers to the overall private investment scale of the project, not necessarily that the two companies will share the cost equally or jointly fund the entire amount.
The launch of this large-scale data center project comes as the Trump administration is determined to address the surging electricity demand from the data center construction boom, which is generating significant political opposition. The inability to add new power capacity poses a major constraint, threatening a key priority for Trump—winning the artificial intelligence race against China—and could exacerbate already high U.S. electricity prices, becoming a political liability ahead of the November midterm elections.
A recent research report from Wall Street giant Citigroup indicates that the AI arms race is shifting from "whose model is the smartest" to "who can produce intelligence at the lowest cost and highest efficiency under physical constraints." The rapid convergence of open-weight models like Kimi K3 towards closed-source frontiers suggests the commoditization of model capabilities is accelerating. However, the simultaneous expansion of parameter scale, long context, and multi-step agent reasoning is shifting the bottleneck from pure FLOPs to HBM capacity and bandwidth, GPU high-speed interconnects, cluster scheduling, and power access.
Nvidia's research also indicates that HBM often becomes the primary scaling constraint as model size, sequence length, and batch size increase. The IEA's forecast shows that electricity consumption growth for AI data centers significantly outpaces overall electricity demand, while grid construction cycles are generally longer than data center deployment cycles.
A research team led by senior analyst Brian Nowak from Wall Street financial giant Morgan Stanley recently released a report significantly raising the 2027/2028 capital expenditure forecasts for the world's five largest hyperscale cloud and computing companies (Meta, Amazon, Microsoft, Google, and SpaceX) to approximately $1.2 trillion and $1.4 trillion, respectively. The institution's forecast for the capital expenditure of major U.S. tech giants in 2026 has been substantially revised upward from $433 billion a year ago to $805 billion.
From Nuclear Fuel Site to Data Center Supercity
U.S. Secretary of Energy Chris Wright stated that this massive private investment of up to $100 billion is one of the key measures the U.S. government is taking to increase power generation, create jobs, and ensure American victory in the artificial intelligence race. "The U.S. government is using its own assets—such as our federal lands—to increase power generation capacity, create jobs, and ensure America wins the AI race," Wright said in a statement.
The Energy Department stated that this privately funded investment, exceeding $100 billion, is one of the largest in Kentucky's history, expected to create 8,000 construction jobs and 600 permanent positions. Brookfield Asset Management will develop and operate the 1.8-gigawatt data center campus, which will occupy part of the Department of Energy's 3,556-acre site. This site was previously used for producing weapons-grade uranium and nuclear reactor fuel. Production using the now-obsolete gaseous diffusion uranium enrichment technology ceased in 2013, and the site is currently in the cleanup phase. The Energy Department indicated that construction is expected to be completed by 2031. The project also includes collaboration with local electric cooperatives.
A global wave of data center expansion has been sparked by the surging demand for AI tools. This construction boom has raised concerns about water and electricity costs across the United States. The Trump administration has been trying to alleviate these concerns before the midterm elections, including by requiring tech companies to commit to bearing related costs.
Brookfield is clearly positioning itself as a full-stack infrastructure holder/organizer for the AI computing era. In other words, Brookfield is not targeting single-point AI GPU computing resource leasing. Instead, it is packaging "AI chip computing supply, server rooms, AI data center power chains, and underlying energy assets" into an integrated core infrastructure sale/lease capability for the AI era. It is betting that government organizations and global tech companies will need massive amounts of core AI infrastructure, including AI chips and power resources, to win the increasingly intense AI competition.
In its statement, Brookfield estimated that developing AI technology or updating cutting-edge AI large models by global governments or tech giants will require massive capital investment. Its AI infrastructure fund is actively seeking commitments from investors to achieve higher returns than its flagship infrastructure fund. Brookfield estimates that achieving global AI prosperity will require $7 trillion in capital investment, with AI computing infrastructure alone needing at least $3 trillion. SoftBank Group, led by legendary investor Masayoshi Son, earlier this year indicated that the data center campus it is developing at another former Department of Energy uranium enrichment site in Ohio could be as large as $500 billion.
The $100 Billion AI Power Fortress Pierces the "AI Computing Power Demand Collapse" Theory
The plan by NextEra and Brookfield to transform the Cold War-era uranium enrichment base in Kentucky into a 1.8-gigawatt AI data center campus, equipped with approximately 2 GW of natural gas generation capacity and up to 2.6 GW of battery storage, is not just about adding another server room. Its core significance lies in the fact that AI infrastructure has upgraded from "large-scale procurement of AI GPUs/TPUs" to a heavy-asset industrial system encompassing land, independent power sources, storage systems, transmission and distribution, liquid cooling, and data center buildings.
The 2 GW of generation and 2.6 GW of storage capacity cannot be simply added to equal 4.6 GW of continuous power supply, as the duration of the storage has not been disclosed. However, this model of "power and computing co-planning" represents the most reproducible engineering path for hyperscale campuses now that public grid access queues and supply reliability have become the primary bottlenecks.
The U.S. Department of Energy has previously designated Paducah as a key site for developing AI data centers and supporting power sources on federal land. This project is not isolated but resonates strongly with the latest construction plans of tech giants. As of the close of U.S. stock trading on July 29, Meta's latest official guidance for 2026 capital expenditure is $125 billion to $145 billion, up from the previous $115 billion to $135 billion. In July, the company also launched its first 1 GW AI data center in Canada, with an investment exceeding C$13 billion, and expanded its Louisiana campus to 5 GW of computing capacity with an investment exceeding $50 billion. Its power plan includes seven new natural gas power plants, three grid-scale battery systems, and nuclear capacity expansion.
Meanwhile, Alphabet raised its 2026 capital expenditure target to $195 billion to $205 billion. Its Google Cloud revenue in the second quarter grew 82% year-over-year to $24.768 billion, directly indicating that demand for AI infrastructure is still accelerating, not peaking. NextEra and Google have also been jointly promoting multiple gigawatt-scale campuses in the U.S., with the first three projects under development and approximately 3.5 GW of resources already in operation or under contract.
This project's blueprint, combined with the latest data center plans from tech giants like Meta and Alphabet, provides a significant positive fundamental catalyst for the globally battered AI computing power theme and offers hard industrial chain evidence to empirically refute the "AI computing power glut" theory. However, it is not a sufficient condition for an immediate V-shaped reversal for all AI stocks. The core issue being traded in the current sell-off of AI computing power stocks is not "data center project plans," but when massive capital expenditures will translate into sustainable free cash flow, and whether debt, finance leases, and project financing will erode shareholder returns.
Alphabet's second-quarter capital expenditure reached $44.924 billion, exceeding its operating cash flow of $39.069 billion, resulting in negative free cash flow of $5.855 billion for the quarter. Credit default swaps (CDS) for AI-related companies have also widened noticeably, with Meta at about 93 basis points, Nvidia at about 78 bps, and the investment-grade CDS index hovering around 53 bps. In other words, while physical AI computing orders are still in a super cycle, the capital market has shifted from "rewarding investment" to "judging AI return on investment." The Kentucky project is scheduled for completion by 2031, so it serves more as an industrial order and power demand signal for the coming years, rather than an immediate catalyst for profits across the entire chain. The AI computing power industry cycle has not stalled, but the valuation system is switching from "computing scarcity" to "cash flow and capital efficiency scarcity."
Comments