A single hyperscale facility can draw +75 MW of load. Build five, and that’s nearly half a gigawatt. Add the current pipeline, and the demand growth rivals what PJM used to see over an entire decade.
PJM’s 2025 summer peak forecast just shot up to 210 GW in 2035, or 228 GW by 2045 - a sharp break from years of flat demand.
By 2030, PJM expects peak load to climb by 32 GW. Data centers drive 94% of that growth.
To keep reliability intact, PJM is proposing a new framework for how these projects interconnect - one that rewrites the rules for loads of 50 MW or more.
The new category: Non-Capacity-Backed-Load
PJM’s August proposal introduces a new class of demand: Non-Capacity-Backed-Load (NCBL).
- No capacity charges. NCBL does not participate in PJM’s capacity auctions. PJM removes this load from each utility’s capacity obligation, so it no longer counts toward the demand curve that sets prices.
- First to be curtailed. PJM can curtail NCBL ahead of capacity-backed demand response or maximum generation steps during emergencies.
- Voluntary - until it isn’t. Loads can choose voluntary NCBL status, but if PJM forecasts a supply shortfall, it will mandate NCBL. Any Bring Your Own Generation (BYOG) contracted by the load is credited in this calculation.
- Still on the hook for wires. NCBL avoids capacity charges by being curtail-able during peak periods, lowering its contribution to peak demand. But it remains liable for transmission charges, because PJM adds curtailed load back when calculating Network Service Peak Load (NSPL).
For data center operators, the trade-off is lower upfront cost, but with curtailment exposure if reliability is strained.
Who qualifies? Most new loads over 50 MW
Threshold: Load additions ≥ 50 MW. The load-serving entity (or utility) can approve smaller projects case-by-case.
Exclusions: Critical infrastructure like hospitals, 911 centers, wastewater plants, gas pumping stations, telecommunication facilities cannot be designated NCBL.
Credits: Loads participating in BYOG or demand response are exempt up to their accredited contribution.
NCBL lowers reliability requirements and softens capacity prices
Assigning NCBL reduces the Reliability Requirement (RR) for PJM and for affected zones.
This shifts the Variable Resource Requirement (VRR) curve - PJM’s demand curve for capacity - downward, softening capacity clearing prices.
This is how NCBL impacts the capacity auctions:
1. Pre-auction: BYOG resources, demand response, and voluntary NCBL are nominated.
2. During the auction: PJM compares supply to the RR. If short, voluntary NCBL is allocated first. If a gap remains, PJM assigns mandatory NCBL by zone, pro rata, until balance is met.
3. After the auction: PJM adjusts the RR and VRR curve downward to reflect allocated NCBL. Then, it recalculates each LSE’s capacity obligations.
In practice, large loads willing to curtail or back their demand with new generation lower the system’s reliability requirement. That in turn lowers capacity prices, or at least stops them from climbing further in an already tight-supply market.
For context, at a $325/MW-day clearing price, a 500 MW NCBL site could run backup diesel for up to 417 hours in a year and still profit.
Data centers can speed interconnection by pairing with new generation
PJM’s proposal is exploring ways to tie data center load to new generation to speed interconnection.
If a data center campus signs an off-take contract with new gas, solar, or storage, that project could move faster through the queue.
Developers also have existing tools to accelerate projects:
- Upgrade requests - pay for transmission upgrades upfront and build in parallel. A gas turbine colocated with a hyperscale campus could advance this way if the developer funds the upgrades.
- Surplus Interconnection Service - tap spare capacity at an existing point of interconnection. A solar-plus-storage project at a retiring coal plant site could share the connection, improve utilization, and deliver power to nearby data centers sooner.
Load Curtailment or BYOG: the developer’s choice
NCBL is effectively a formalized, large-scale interruptible class. PJM can curtail these loads before calling on capacity-backed demand response or maximum generation.
That leaves developers three options, each with trade-offs:
- NCBL with curtailment only: Accept curtailment risk but avoid capacity charges. Reliability could fall to ~98% uptime. By contrast, data centers often require “five nines” - 99.999% uptime, or just 5 minutes of downtime in a year.
- NCBL with diesel backup: Avoid capacity charges, cover curtailments with onsite generators. Diesel provides short-term coverage but with limited duration and lower reliability than grid-backed supply.
- BYOG (Bring Your Own Generation): Pair new load with new supply, such as gas, solar-plus-storage, or hybrid projects. This can provide firm capacity but raises questions on cost, financing, supply chain, and queue certainty.
But not all supply matches hyperscale demand.
The chart below compares data center demand with resource output under different interconnection pathways, underscoring how each option provides a different degree of reliability.
Data center load is treated as flat and continuous. In reality it can swing by hundreds of megawatts in milliseconds.
Gas turbines can track demand but face cost and emissions hurdles.
Diesel generators, while dependable during emergencies and designed to operate in isolation from the grid, can face stricter emission controls and capped runtime under EPA rules.
Solar-plus-storage reduces peaks but leaves gaps.
For developers, the trade-off is clear
NCBL lowers costs but exposes projects to curtailment. Diesel backup only goes so far. Firming with new generation means higher capital and interconnection hurdles.
The siting decision becomes critical. Strong-grid zones will carry less risk, while speculative projects in weak, capacity-tight areas may struggle to clear financing without BYOG.
On the flip side, if BYOG is used as a stopgap while awaiting transmission upgrades, the longevity of this initiative could be a risk factor. If the NCBL rules change or grid conditions improve sooner than expected, developers may be left with underutilized, stranded assets.
PJM’s proposal has drawn fire from multiple directions
Hyperscale players like Amazon, Google, Microsoft argue the NCBL concept undermines tariff integrity and market design.
LS Power and East Kentucky Power warn it could cut capacity prices, erode investor confidence, and drive data centers to other RTOs. This would stall economic development in PJM.
Some state governors urge PJM to focus instead on better load forecasting, transmission planning faster interconnection.
The takeaway
PJM’s proposal remains conceptual.
Member debate is underway through the Critical Issue Fast Path (CIFP) initiative, with a proposal filing target to FERC by year-end.
The goal is for any solution to be in place by the 2028/2029 capacity auction, scheduled for June 2026.
But the trajectory is clear: hyperscale loads in PJM will face a new rulebook.
This is not a cap on PJM’s load growth. Lower realized demand is more likely to come from project attrition in the queue than from PJM blocking new interconnections.
Watch this space next week for when we translate PJM’s proposed framework into our load forecast.
