In recent months, the discussion around U.S. AI infrastructure has shifted from chip supply to power access. BNEF has raised its data center electricity usage forecast, National Grid Ventures has invested in large-scale load projects in the U.S., and New York State has paused state environmental permits for some new super-sized data centers.
These three clues point to the same question: After AI companies purchase GPUs, will they still be able to secure land, cooling, water, grid access, and long-term power supply as planned?
For investors, data centers are no longer just capital expenditure projects for tech companies. The power consumption of large data center campuses may be comparable to that of a small city. As long as AI training and inference continue to expand, utilities, natural gas, nuclear energy, grid equipment, copper, and data center REITs will all be integrated into the AI pricing chain.
Revised Electricity Forecast, Pressure on Grid Nodes
In its report at the end of 2025, BNEF predicted that the U.S. data center power demand could reach 106GW by 2035, an increase of 36% from its forecast 7 months ago. This is not the realized electricity consumption but a market reference for future demand.
106GW can be understood as the capacity supply level of a group of large power plants operating at full load over the long term. What's more troublesome is that these loads will not be evenly distributed across the U.S. but concentrated in a few data center states and grid regions.
BNEF also forecasts that data center capacity in the PJM grid region could reach 31GW by 2030. PJM covers multiple states in the U.S. East Coast and is one of the power markets where data centers and industrial loads are concentrated. When load concentration grows, the issue shifts from national power generation capacity to whether local nodes can handle it.
This explains why electricity forecasts impact asset pricing. Buying GPUs is just the first step for AI companies, as data center commissioning requires power sources, transformers, transmission lines, backup power, and long-term power purchase agreements. Compared to servers, grid access and permits are more challenging to replicate quickly.
Energy Capital Begins AI Load Tie-Up
National Grid Ventures announced on July 1 that it had acquired a 35% stake in Joulent for $1.75 billion, participating in a large-scale electricity infrastructure project in the United States. The signal of this transaction indicates that traditional energy capital is starting to view AI data centers as long-term load assets.
Joulent's inaugural Project Kilby is located in West Texas and is a 2.67GW complementary power facility, with Chevron and ExxonMobil each holding a 50% stake. The project is planned to provide power through a 20-year PPA to a Microsoft-operated data center, with the goal of commencing operations in 2028.
The focus is not on a single power source winning out, but on data centers shifting from "waiting in line for power" to "locking in power ahead of time." Long-term PPAs and self-generation are becoming prerequisites for compute expansion.
The logic of self-generation is straightforward. If data centers continue to draw power from the grid, the expansion costs may be passed on to all users. By developing their own power sources or signing long-term contracts, project owners can increase power supply certainty and more easily demonstrate to regulators that they won't encroach on residential electricity.
There are still risks. Gas projects face fuel price and emissions pressure, nuclear projects encounter regulatory and construction timelines, and grid expansion is constrained by transformers, transmission lines, and local permits. Capital binding power sources in advance indicates that demand is being taken seriously and that the bottlenecks are specific enough.
New York Puts Cost Allocation in the Spotlight
New York Governor Kathy Hochul signed an executive order on July 14 to halt state environmental permits for new super-sized data centers, with a maximum duration of one year. More specifically, the suspension applies to incomplete related permit applications, not all data center constructions.
The state government's reasons include protecting consumers, the environment, the grid, and communities, while focusing on water resources and electricity cost impacts. This places clear constraints on the AI power narrative: data centers can expand, but cannot pass on grid upgrades, water resource pressures, and residential electricity price hikes to local communities.
New York State also stated that the Energize NY program will require data centers to pay higher energy costs, or provide their own power, and consider grid acceleration funds and dedicated clean power requirements. This is not merely opposition to AI, but a demand for high-power-consuming projects to internalize the true costs.
This will change the competitive landscape for data center operators and REITs. Previously, the market valued land, leases, customer quality, and financing capabilities. Now, it is also necessary to assess whether projects can secure electricity, obtain permits, and prove that they will not drive up residential electricity prices.
New York may not represent the entire United States. However, if high-load areas such as Virginia, Georgia, and Texas also experience similar pressure, the timeline for AI infrastructure expansion will shift from "who gets the chips first" to "who obtains local permits and verifiable power sources first."
Nuclear Power Provides Long-term Optionality
The Trump administration's 2025 initiative to advance small modular reactors, AI data centers, and federal site deployments has brought nuclear power back into the AI electricity narrative. For data centers, nuclear power's appeal lies in its stability, low carbon emissions, and long-term power supply.
Companies like Oklo, X-Energy, Aalo Atomics, Valar Atomics, and Helion are drawing market attention, stemming from this vision. Investors are seeking power solutions that are more reliable than the traditional grid and more stable than intermittent renewable energy sources.
However, policy support does not guarantee commercial delivery. Traditional nuclear construction has long lead times, and advanced reactors face challenges related to regulation, supply chains, fuel, financing, and public acceptance. Even with accelerated approvals, large-scale power delivery to data centers cannot be achieved in the short term.
Therefore, nuclear power is more akin to a long-term option. It can explain why nuclear companies, uranium, engineering services, and grid equipment are being included in the AI trading chain but cannot prove that the AI electricity bottleneck has a definitive solution. Equating policy catalysis directly to order fulfillment is a key risk in this narrative.
Cost Grounding Determines Market Sentiment Trajectory
Whether AI electricity trading can transition from a thematic market sentiment to a performance-based sentiment depends on two variables: whether data centers can access electricity as planned and whether additional costs can be clearly allocated.
If on-site power generation and long-term power purchase agreements can be implemented at scale, utilities, grid equipment providers, gas, and some nuclear assets will have a more defined anchor for orders. AI companies can also trade off higher electricity costs for computational certainty.
If local moratoriums spread or residential electricity cost pressure becomes a political issue, the pace of data center construction may be prolonged. The pressure may not necessarily be on AI demand itself, but on the siting, grid interconnection, and profit margin assumptions of high-power-consuming projects.
The essence of this storyline is not "AI is definitely facing a power shortage" or "nuclear power will immediately solve the problem." The more tradable judgment is that AI expansion is turning the electricity system into a new pricing anchor. Those who can prove they are bringing electricity are more likely to translate their computational demand into revenue and valuation.
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia
