The AI sector faces shortages in storage, computing power, and optical modules, but at the physical implementation level, the critical bottleneck is electricity.
Last November, Microsoft CEO Satya Nadella noted that the company had a batch of GPUs sitting in inventory due to insufficient power and data center conditions.
Chips may be acquired, but computing capacity cannot be brought online, with the holdup stemming from substations, transmission lines, and connection approvals not keeping pace.
Now, this major bottleneck in the United States has received a policy boost.
The Federal Energy Regulatory Commission (FERC) has directed the six major regional grid operators to re-evaluate their interconnection rules for large power consumers like data centers.
The focus includes streamlining connection processes, clarifying grid upgrade costs, supporting data center siting near power plants or with on-site generation, and designing new interconnection methods for flexible, high-load customers.
FERC has essentially created a "green channel" for data center power connections, providing clearer pathways for project interconnection, cost allocation, and power supply methods.
How will this policy alter the logic of data center power interconnection in the U.S.? Which power assets stand to benefit?
The Core Interconnection Challenge for Data Centers
For a data center to become operational, electricity must travel from a power plant through the high-voltage transmission grid and regional substations before reaching the data center's own substation and distribution system.
The bottleneck often lies not in whether a power plant is nearby, but in whether the grid can reliably deliver the additional power required.
Large AI data centers can have power demands reaching hundreds of megawatts, even approaching gigawatt scale.
Connecting such a massive load requires regional grids to reassess transmission capacity, substation expansions, system reliability, and cost allocation.
These projects typically undergo application, queuing, system studies, and construction phases.
The traditional process was designed for conventional industrial and commercial customers and struggles to accommodate the scale and deployment speed of data centers.
Even with land, chips, and servers ready, projects can stall waiting for sufficient transmission capacity or the completion of interconnection studies.
Thus, the U.S. power bottleneck encompasses both generation resources and the timely delivery of electricity to data centers, with FERC's current push focusing on the latter.
Key Policy Changes Introduced by FERC
FERC has required the six regional grids to justify their existing rules or submit reform plans within 60 days, and to explain within 30 days how they will match sufficient generation resources to new large loads in the future.
The policy shifts focus on five main areas:
First, moving from a "general queue" to a dedicated channel for large loads.
Previously, data centers largely followed the same interconnection process as traditional customers, involving sequential steps of application, queuing, system studies, cost confirmation, and construction.
For loads of hundreds of megawatts or gigawatts, this process often led to repeated requests for additional information and reassessments, dragging on for years.
Now, FERC mandates that regional grids specifically review procedures for large-load interconnections, reduce redundant studies, clarify application standards and timelines, and explore using advanced transmission technologies to unlock existing line capacity.
Second, shifting from "studying costs as you go" to upfront clarity on cost allocation.
Previously, connecting a large data center often required substation expansions and transmission line upgrades, but whether the data center, power generators, or general ratepayers would bear these costs was frequently negotiated during the study process.
Such disputes not only increase costs but also delay projects, as developers might have land and equipment ready but cannot finalize the connection price.
FERC now demands greater transparency in cost allocation, aiming to avoid passing grid upgrade costs triggered by data centers onto residential and small commercial customers.
Data centers may not pay less, but they can assess project costs and feasibility earlier.
Third, moving from ambiguous rules for siting near power plants to clearer "co-location" guidelines.
Previously, even a data center built next to a power plant did not guarantee direct access to power.
Rules regarding how much power could be used, whether transmission charges applied, and how surplus power entered the public grid varied by region, creating long-standing controversies for co-location projects where physical proximity did not guarantee regulatory connection.
FERC now requires clear rules for interconnection, pricing, and reliability when data centers are sited with power plants or storage facilities, providing a clearer path for building data centers adjacent to generation.
Fourth, transitioning from "fixed peak load calculations" to recognizing flexible load capabilities.
Previously, grids typically assessed data center interconnections based on peak load, requiring them to reserve sufficient generation and transmission capacity accordingly.
Even if a data center had storage, backup generation, or was willing to reduce consumption during grid stress, it might not gain significant advantages in the approval process.
FERC now encourages designing new services for flexible loads.
Data centers capable of load shifting, peak reduction, switching to on-site power, or accepting temporary curtailment may receive faster interconnection.
Fifth, shifting from "separate queues for generation and load" to integrated planning.
Previously, new generation projects and data centers typically entered separate application processes and studies.
The result could be a data center ready to break ground while its supporting power source was still awaiting approval, or a completed power plant lacking synchronized transmission and load conditions.
FERC now requires regional grids to consider evaluating new generation, storage, and data center loads together, making it easier to synchronize "power plant + data center" combo projects.
Overall, FERC is establishing a more standardized, transparent process for large-load interconnection that is better suited to the massive scale of data center power demand.
For data centers, connection timelines and costs may become more predictable and assessable.
Potential Beneficiaries Among Power Assets
The benefits from FERC's actions will not be evenly distributed across all generation types.
The degree of benefit depends primarily on three factors: ability to provide stable power, speed to deployment, and proximity to data center demand.
The first tier: Existing nuclear and natural gas power assets.
The most direct beneficiaries are already-operational nuclear plants and natural gas-fired power stations located near data center clusters.
In the PJM region, for example, which is one of the most concentrated areas for U.S. data center demand, there is a significant amount of existing nuclear and gas-fired capacity.
Companies like Constellation Energy, Vistra, and Talen Energy own generation resources in the area.
Talen Energy's data center co-location project around the Susquehanna nuclear plant exemplifies the logic of "directly serving new loads near a power plant."
Nuclear is well-suited for providing long-term, stable baseload power, while natural gas generation offers greater flexibility and scalability.
As co-location and local supply rules become clearer, the locational value of generation assets already positioned near data centers will be enhanced.
The second tier: On-site generation and rapid power supplementation solutions.
Gas turbines, fuel cells, and microgrids can be deployed directly near data centers, reducing wait times for public grid upgrades.
GE Vernova stands to benefit from increased demand for gas turbines, while Bloom Energy offers modular fuel cell solutions.
The core value of these solutions is reducing time-to-power, though they remain subject to equipment lead times, gas pipeline availability, and cost constraints.
The third tier: Renewable energy paired with storage.
Solar and wind can meet the long-term power purchase agreements of cloud providers and align with clean energy goals, but their intermittent output typically requires pairing with storage, natural gas backup, or the public grid.
Therefore, this segment is more of a long-term beneficiary; its short-term ability to provide stable, rapidly deployable power remains weaker than that of existing nuclear, natural gas plants, and on-site generation.
Comments