State of Requirements for Large Dynamic Loads like AI Data Centers

Executive Summary

AI-era data centers behave like “electrified factories”: large, often fast-ramping, and increasingly colocated with generation. The regulatory center of gravity is shifting quickly from ad-hoc treatment of “big customers” to formalized pathways that demand early coordination, full steady-state/short-circuit/dynamics models, power-quality controls, telemetry/visibility, and—in some regions—curtailment or interruptibility provisions. Three developments define the near-term landscape:

codified “Large Load” frameworks (e.g., ERCOT’s Large Load Interconnection process and SPP’s High Impact Large Load policy);
explicit treatment of co-located load at generation (PJM under FERC’s co-located load proceeding and subsequent directive to develop clearer tariff rules); and
interim caps/allocations where demand is outpacing grid capability (AESO in Alberta).

For developers, the success pattern is consistent: open a ticket with the transmission owner (TO) early, bring a complete modeling/data package aligned to NERC FAC-002-4 and MOD-032-1, design in reactive power and harmonics controls from the start, plan for telemetry and potential curtailment, and structure ramp/step profiles that grid studies can actually clear. (NERC)

1. Foundations

1.1 What counts as a “large dynamic load,” and why this matters now

A large dynamic load is a customer facility whose size (tens to hundreds of MW) and time-varying behavior (fast steps/ramps, voltage sensitivity, power-quality impacts) can materially affect bulk power system performance. Modern AI data centers add (a) very high coincident demand, (b) power-electronic front ends (UPS, rectifiers, VFDs) that reshape harmonics and reactive power flows, and (c) backup generation and/or on-site storage that must be represented in studies. Transmission planners increasingly require load modeling with generator-grade fidelity because these facilities can drive voltage dips, thermal overloads, delayed voltage recovery, and mis-operation of protection if not designed and integrated carefully. (NERC)

1.2 Baseline universal requirements (NERC)

Two NERC standards underpin nearly every large-load conversation in North America:

FAC-002-4

FAC-002-4 (Facility Interconnection Requirements) obligates Transmission Planners/Planning Coordinators and TOs to study the reliability impact of new or modified facilities—explicitly including end-user loads—using steady-state, short-circuit, and stability analyses (with required scope determined by the planning authority’s criteria and the characteristics of the interconnecting facility). FAC-002-4 replaced FAC-002-3 in the U.S. footprint; the older “material modification” framing is commonly encountered in legacy utility language, but the current standard’s trigger concept is a “qualified change.” (NERC)

‍Plain-English: Before your project plugs in, the grid owner must run “what-if” simulations for normal and contingency conditions.‍

Example: A 150-MW AI campus at 230 kV triggers thermal and voltage checks, short-circuit duty at nearby breakers, and transient/dynamic studies to verify no unacceptable post-contingency voltage behavior on a local 230/115-kV corridor. (NERC)

MOD-032-1

MOD-032-1 (Data for Power System Modeling) requires consistent modeling data submission for planning cases—covering load composition, power factor targets, reactive devices, dynamic behavior, and related parameters. (NERC)

Plain-English: You must deliver a full “digital twin” of your facility so the grid model is accurate.

Example: You provide a spreadsheet and dynamic model describing 70% non-linear rectifier load behind UPS, 20% induction motor fans/pumps, 10% misc., plus a ±MVAr plan from STATCOM/capacitor banks to hold 0.98–1.00 power factor at the Point of Interconnection (POI). (NERC)

Why this matters: Planning teams will ask for full modeling even at distribution voltage if the bulk system can be affected. Missing or late data is a top schedule killer. (NERC)

1.3 Who pioneered the new playbooks

Historically, large loads were integrated under general interconnection and local TO standards. The recent surge of crypto, hydrogen, and especially AI has pushed regions to harden bespoke pathways: ERCOT’s Large Load Interconnection process, SPP’s High Impact Large Load (HILL) policy, PJM’s evolving co-location treatment under FERC direction, and AESO’s interim cap with staged allocation. Each reflects the same goal: ensure reliability and transparency while moving faster than traditional case-by-case processes. (ERCOT), (SPP), (Reuters), (AESO)

2. Core Concepts & Mechanics

2.1 The minimum modeling package you should expect to provide

Steady-state: P, Q, power factor targets at POI; voltage control scheme; capacitor/reactor banks; STATCOM/SVC settings.

‍Plain-English: How hard you’ll press the gas (MW) and how you’ll keep voltage steady (MVAr).‍

Example: 200 MW at 0.99 lagging PF with 40 MVAr STATCOM to regulate 1.02 pu at the POI.‍

Short-circuit: Equivalent Thevenin contribution (even for “load,” back-up gen and some inverter-based devices can affect fault levels), breaker duty checks at nearest substations.

‍Plain-English: When there’s a “short,” how big is the surge and can breakers survive?

‍Example: Four 10-MVA diesel gens modeled for near-source fault duty contribution (noting typical controls and any fault-current limiting behavior).‍

Dynamics: Composite load model (motors, electronic loads), UPS/inverter behavior (often fault-current limited; ride-through logic matters), ramp/step profiles, under-frequency/under-voltage load shedding (UFLS/UVLS) interactions.

‍Plain-English: How your site behaves over seconds and minutes during disturbances.

‍Example: A 60-MW step change over 2 s; UPS ride-through for 10 s without tripping; motor stalling not allowed for a single-contingency event. (NERC) ‍

Power quality and harmonics: Compliance with IEEE-519-type harmonic limits and local flicker requirements is typically embedded in TO/PTO interconnection standards; filters (passive/active) and multi-pulse rectifiers or front-end converters are common mitigations.

‍Plain-English: Keep your “electrical noise” quiet.

‍Example: 12-pulse rectifiers and tuned filters to keep 5th/7th harmonics within PCC limits; active filters for higher-order components. (ATC) ‍

Telemetry/visibility: Real-time MW/MVAr/voltage, breaker status, alarms, and EMS/SCADA integration requirements where applicable.

Plain-English: The grid operator needs to “see” your site like a power plant.

Example: A defined telemetry list (MW, MVAr, POI voltage, breaker status, reactive device status/setpoints) included early in the interconnection package. (ERCOT)

2.2 Ride-through, curtailment, and interruptibility

ERCOT has formalized a Large Load Interconnection process for new or modified load facilities ≥75 MW (per ERCOT Planning Guide Section 9.2.1 as referenced on ERCOT’s Large Load Integration page). (ERCOT)

Separately, ERCOT also advanced modeling expectations for loads greater than 25 MW via a planning-focused revision (often discussed alongside large-load integration), which developers should treat as an “early data and modeling” signal even when not in the ≥75 MW process. (ERCOT) (Gibson Dunn)

SPP’s HILL policy targets an accelerated pathway with an advertised ~90-day study-and-approval timeline and integrated design/study/registration concepts, while maintaining system-security safeguards. (SPP), (Utility Dive) (SPP)

Plain-English: Big loads must stay connected through routine bumps, and be prepared to ramp down when the grid is stressed.

Example: A 120-MW AI cluster defines interruptible blocks and provides 5-minute and 30-minute ramp-down profiles tied to operator/TO instructions or agreed curtailment provisions (where applicable). (ERCOT), (SPP) (ERCOT)

2.3 Co-location with generation: pricing and “rules of the road”

FERC initiated a Section 206 proceeding focused on co-located load in PJM (Docket EL25-49), directing PJM and PJM Transmission Owners to show cause or propose tariff changes addressing rates/terms/conditions and reliability concerns for co-location arrangements. (PJM)

On December 18, 2025, Reuters and AP reported that FERC directed PJM to implement clearer rules governing connection of AI-driven data centers and other large loads located near power plants, and found PJM’s existing tariff unclear/inconsistent (unjust and unreasonable), requiring revisions. (Reuters), (AP) (Reuters)

Plain-English: If you want to plug your data center into a plant’s backyard, PJM is being pushed toward clearer terms, costs, and visibility/studies to match.

Example: A data center at a nuclear site may require bespoke agreements and tariff-compliant treatment to ensure the public network is protected and costs aren’t shifted improperly. (Reuters), (AP) (Reuters)

2.4 How planning criteria show up in your design

Voltage and reactive power: Many TOs/PTOs enforce POI voltage ranges and minimum PF/VAR capability; modern STATCOMs give dynamic MVAr support without large steps. (ATC)
Ramp/step shapes: Grid studies clear smoother ramps; step limits are negotiated and may be codified in interconnection agreements.
UFLS/UVLS: You must identify which parts of your load can be automatically shed to support system integrity‍
Backup generation and emissions: Air permits and operational use profiles must be aligned with study assumptions; UPS vs. genset bridging shapes disturbance behavior.

3. Applications & Implications

3.1 Regional snapshots (what you will actually be asked to do)

ERCOT (Texas)

Status: ERCOT has a published “Large Load Integration” resource hub and a Planning-Guide-anchored process for ≥75 MW load facilities. (ERCOT)
ERCOT’s stakeholder documents show Board consideration/approval actions on April 8, 2025 tied to large-load interconnection process revisions, with implementation staged (e.g., portions effective June 1, 2025 and later updates effective December 15, 2025). (ERCOT)

What to expect: Early submission of steady-state and dynamic models, PF/MVAr plans, step/ramp profiles, and required interconnection forms/communications per ERCOT’s posted templates and Planning Guide references. (ERCOT)

Implication: Treat your campus like a grid-significant facility—arrive with a polished modeling package and a control strategy that can be studied and, where applicable, operationally constrained. (ERCOT)

PJM (Mid-Atlantic/Midwest)

Status: Load interconnections proceed via PJM/TO New Service Requests; PJM Manual 14H is the primary process reference for New Service Requests, and Manual 14C governs construction/as-built alignment concepts used to ensure built facilities match studied assumptions. (PJM)
FERC’s co-located load proceeding (EL25-49) and the December 18, 2025 directive reported by Reuters/AP signal formal tariff clarification for AI/co-located large loads. (PJM), (Reuters), (AP)

Developer actions: File the appropriate New Service Request, align with the host TO, and prepare dynamics/short-circuit data plus robust as-built and model-validation packages where required. Monitor PJM’s tariff filing(s) and compliance timeline under the FERC directive for project structuring in co-location scenarios. (PJM), (Reuters)

SPP (Central/South-Central)

Status: SPP announced Board approval of its High Impact Large Load (HILL) policy (including a 90-day target), with supporting artifacts such as a HILL process “one-pager.” (SPP)

SPP also filed tariff revisions for HILL/HILLGA processes (transmittal letter and docketed submission). (SPP filing materials)

Separately, Utility Dive reported FERC approval of SPP’s Provisional Load Process, while noting the HILL policy still required FERC approval at the time of reporting. (Utility Dive)

Developer actions: Submit HILL screening early with size, ramping, PF, harmonics/filters, and backup generation plans; be ready for targeted mitigations in an accelerated window and track the tariff’s FERC posture. (SPP)

MISO (Midwest/South)

Status: Load interconnections are typically TO-led; MISO provides planning coordination and has published guidance describing load interconnection principles and data expectations. (MISO)

Developer actions: Start with the host TO’s interconnection guide (e.g., ATC). Align on PF, flicker, IEEE-519-type harmonic limits, starting/inrush profiles, and operational flexibility assumptions. (MISO), (ATC)

NYISO (New York)

Status: NYISO’s TEI framework explicitly covers studies for generation, load, and transmission facilities; the TEI Manual and TEI User’s Guide describe processes and note ongoing revisions to incorporate newer interconnection procedures. (NYISO)

Developer actions: Engage the host TO and NYISO; provide steady-state and dynamics models, PF/VAR plan, demand profiles (including steps/ramps), telemetry expectations where applicable, and any islanding/black-start interactions. (NYISO)

CAISO (California)

Status: In CAISO territory, technical interconnection requirements for load-only or load + generation interconnections are often implemented via Participating Transmission Owner (PTO) tariffs/handbooks (e.g., PG&E’s Transmission Interconnection Handbook explicitly documents technical requirements for interconnection of loads and generators to the PTO system). (PG&E)

Developer actions: Start with the PTO handbook/standards; deliver full models and power-quality controls; coordinate early on metering/protection/control requirements, and expect ISO involvement where transmission impacts or tariff requirements apply. (PG&E)

AESO (Alberta, Canada)

Status: AESO adopted an interim staged approach for large load projects; Phase 1 allocated 1,200 MW of capacity (with executed load contracts) and Phase 2 consultations are underway to establish a longer-term framework. (AESO)

Developer actions: Prepare a comprehensive technical package (ramping, PF, UPS/backup behavior, harmonics) and track Phase 2 rulemaking to position for future allocations. (AESO)

3.2 Why this matters for schedule and cost

Study acceptance hinges on model quality: poorly specified dynamic behavior and unclear ramp/step logic are a primary cause of study rework.
Power-quality non-compliance (harmonics/flicker) drives late redesigns (filters, STATCOM sizing). (ATC)
Telemetry and curtailment/interruptibility terms increasingly affect commercial viability in ERCOT/SPP and are a focal point for PJM co-location clarity under FERC direction. (ERCOT), (SPP), (Reuters)

3.3 The GridStrong integration playbook (field-tested)

Initiate with the TO even if you plan to connect at distribution. Anchor on NERC FAC-002-4 and MOD-032-1 data scopes; bring a one-page “study-ready” parameter list to the kickoff. (NERC)

Engineer PF/VAR and harmonics from day zero: pre-size STATCOMs/filters and model them explicitly; treat PF at the POI as a design constraint, not a site-ops afterthought. (ATC)

Declare ramp/step envelopes the planners can clear: set default and emergency ramps, plus contingency block-shedding.

Build a telemetry/visibility plan that mirrors “operator-usable” points found in regional large-load processes. (ERCOT), (SPP)

If co-locating with generation (PJM region), pre-align on cost allocation, islanding/perimeter protection, and tariff treatment in light of FERC’s ongoing actions. (PJM), (Reuters), (AP)In ERCOT/SPP, identify interruptible blocks and curtailment logic up front to shorten study cycles. (ERCOT), (SPP)

4. Integration & Broader Context

4.1 Adjacent domains that will touch your project

Power electronics & controls: UPS/inverter fault behavior, current-limiting, and PLL settings define ride-through success.
Plain-English: Your electronics decide whether you “blink” or “sail through” a grid wobble.
Example: Setting inverter current limit at 1.2 pu with priority to voltage support can help stabilize POI voltage during an N-1 line trip.
Protection coordination: High electronic load fractions change short-circuit behavior and time-current curves; UFLS/UVLS participation requires careful block definitions so you don’t under- or over-shed.
Thermal/EMT planning: Heavily electronic loads can justify EMT studies in weak-grid corridors, especially with nearby STATCOMs/SVCs‍
Policy & siting: Rapid load growth is altering state resource plans and triggering exceptional proceedings. For example, AP reported Georgia’s Public Service Commission approved a major Georgia Power expansion plan primarily tied to data-center/AI load growth. (AP News)

4.2 From theory to practice: how study artifacts translate to steel-in-the-ground

Voltage/reactive devices: capacitor banks vs. STATCOMs—trade static cost for dynamic control; include space and ducts for future MVAr expansion.
Harmonics: reserve bays for filters in the POI substation; specify filter tuning that considers network resonance variation across operating states.
Telemetry/SCADA: specify RTU points during ~30% design; test with the TO/ISO well before energization.
‍Agreements: embed ramp/step envelopes, PF requirements, curtailment/interruptible blocks, telemetry points, and power-quality limits into interconnection and service agreements so operations match the studied model. (PJM), (ERCOT), (SPP)

4.3 What’s coming next (12–24 months)

PJM: Tariff language clarifying co-located large loads, with pricing and reliability/visibility protections, following FERC’s co-located load proceeding and the December 18, 2025 directive reported by Reuters/AP. (PJM), (Reuters), (AP)
SPP: HILL artifacts maturing (including published process documentation) and ongoing FERC action on filed tariff revisions; watch whether the 90-day target holds under complex constraints. (SPP), (Utility Dive)
ERCOT: Continued refinement of LLI templates and forms (notably updates posted in late 2025). (ERCOT)
MISO/NYISO: More explicit “substantial load” materials and utility/TO guidebooks; NYISO is actively revising TEI materials to incorporate newer interconnection requirements. (NYISO), (MISO)
AESO: Phase 2 consultations to move from interim caps to a durable, rules-based framework; expect clearer sequencing and capability screens. (AESO)

5. Practical Developer Toolkit (works in most places)

One-pager for kickoff (copy/paste into your spec):
- Site size by phase (MW), min/max PF, scheduled ramps (normal/emergency), maximum 1-s and 10-s steps.
- Composition (% UPS/electronic load, % motors, % misc.), UPS topology, inverter settings, expected fault-current behavior.
- VAR plan at POI (device ratings, setpoints, deadbands), voltage control strategy.
- Harmonics and flicker limits/design (IEEE-519 references, filter types, expected THD/TDD).
- Telemetry points (MW, MVAr, V, breaker status, device setpoints), curtailment blocks, UFLS/UVLS blocks.
- Backup gen and emissions use profile; black-start/islanding interactions if any. (NERC), (ATC)
Modeling cheat-codes:
- Provide EMT and dynamic model files as requested; include a “study harness” with ramp events, block-shed scripts, and POI voltage holds.
- Share a sensitivity matrix (capability map) showing MVAr vs. POI voltage for seasonal cases; planners clear faster when they can see margins.
- Deliver harmonics studies with multiple system short-circuit levels; filter performance can change with topology. (ATC)
Common failure modes to avoid:
- Treating PF as a site-ops variable rather than a POI contract requirement.
- Submitting static VAR devices for a problem that is explicitly dynamic (fast steps).
- Omitting operator-usable ramp/step and (where applicable) curtailment logic in ERCOT/SPP and co-location scenarios under PJM’s evolving rules.
- Under-modeling inverter/UPS fault behavior; this leads to protection surprises during TO reviews.

State of Requirements for Large Dynamic Loads like AI Data Centers — What’s Here and What’s Coming