Researched by BEAST Library Curator | Verified by Harper | Quality: 8.85/10
Grid Demand Crunch Hits CRE: PJM Cleared at the Cap, Taipower Froze Northern Capacity, and What FMs Should Do in the Next 90 Days
BLUF: Two grid events this quarter changed the math for every commercial building owner with a peak demand charge or a tenant data closet. PJM's 2026/2027 capacity auction cleared at the FERC-approved $329.17/MW-day cap across the entire footprint — meaning every utility passing through capacity costs is now collecting record peak charges from tenants. Simultaneously, Taipower has stopped accepting new ≥5MW supply requests in northern Taoyuan since August 2023 and will require energy-use plans for any 5MW+ project from 2026 onward. Together these signal one thing: the grid is now a paying customer for building flexibility, not a passive supplier of cheap power. This report is a 90-day playbook for facility GMs across the AISB footprint — US PJM zone, Singapore, and Taiwan.
What just happened on the supply side
The 2026/2027 PJM Base Residual Auction procured 134,311 MW of unforced capacity at the price cap of $329.17/MW-day across all 13 states and DC. Demand response cleared roughly 8 GW (5% of resource mix), holding flat with the prior auction. PJM has separately proposed an emergency procurement of 15 GW of new capacity by June 2031, explicitly aimed at the data center load wedge. ISO-NE will begin running its FERC Order 2222 markets on November 1, 2026, meaning aggregated DERs (including office-building battery storage and HVAC flexibility) become bid-eligible against utility-scale generation in the energy and ancillary services stacks.
Taipower is moving in the same direction with different mechanics. Taiwan's AI-related electrical load is forecast to grow from 0.24 GW in 2023 to 2.24 GW in 2028 — more than 8x. Taipower has not approved new ≥5MW supply requests in northern Taoyuan since August 2023; the chairman has explicitly directed AI data centers toward sites near power sources in central and southern Taiwan. From 2026, all new 5MW+ projects must submit energy-use plans before construction; hyperscale data centers must hit PUE ≤1.3 and colocation facilities PUE ≤1.4.
Why this matters for non-data-center CRE
This is the trap most CRE owners are walking into: "my building isn't a data center, this doesn't apply to me." Wrong. The capacity-cost pass-through and the GEB economics affect every commercial customer with a peak demand charge — which in PJM territory means almost everyone above 100 kW peak. The grid does not care whether your load is a chiller plant, a tenant server room, or an AI training cluster. It cares about three things: your peak (kW), your shape (kWh by hour), and your dispatchability (response time and curtailment depth).
The 2026 grid economics table
| Market / Region | Headline Number | What it Means for a 250,000 sqft Office | Source |
|---|---|---|---|
| PJM 2026/2027 BRA | $329.17/MW-day cap (UCAP) | ~$120K/yr capacity-cost pass-through on a 1MW peak building (assuming full pass-through). Was ~$10/MW-day in 2024/25. | PJM Inside Lines, July 2025 |
| PJM Demand Response | 8 GW cleared, ~$330/MW-day | An office shedding 200 kW for ≤10 events/yr can earn ~$24K/yr capacity payment via aggregator | PJM 2026/27 BRA Report |
| ISO-NE FERC 2222 | Nov 1, 2026 go-live | Behind-the-meter battery + HVAC flex becomes bid-eligible in energy + ancillary stacks via DERA | ISO-NE Order 2222 Project Page |
| Taipower 2026 mandate | Energy-use plan required for ≥5MW projects | Any new fit-out with material server load now needs documented load profile + reduction commitment | Digitimes, Jan 2026 |
| Taipower northern freeze | No new ≥5MW Taoyuan since Aug 2023 | Office buildings hosting tenant tech tenants in Taipei/Taoyuan have hard headroom limits | Digitimes, Dec 2025 |
| Singapore tariff (Apr-Jun 2026) | 27.27¢/kWh (excl. GST) | ~$0.30/kWh effective; demand charge structure rewards peak shaving on 30-min intervals | SP Group / EMA |
| DOE GEB potential | Up to 30% of commercial peak curtailable | A 1MW office can typically expose 200-300 kW of dispatchable load via existing BMS controls | DOE BTO / RMI |
The 90-day FM playbook
Here's what I'd do if this were my building, sequenced by week and dependency:
Weeks 1-2: Establish the baseline (you cannot manage what you do not measure)
- Pull 12 months of interval data from your meter (15-min in PJM, 30-min in Singapore, 15-min in Taiwan). If you only have monthly bills, that is the first capital ask — interval submetering pays back in <90 days at current capacity prices.
- Identify your top 100 peak hours. Roughly 80% of capacity-charge exposure typically lives in <5% of hours. The shape is non-uniform — Taipei peaks are afternoon-summer; PJM peaks are evening-summer + winter-morning; Singapore is mostly afternoon-flat.
- Score your dispatchable load. Chiller staging, AHU fan VFDs, lighting dimming, EV chargers, tenant server-room cooling setpoints. Be honest about what you can move 2°F without a tenant complaint.
Weeks 3-6: Pick your aggregator and contract path
- PJM zone: Talk to 2-3 DERAs (Enel X, Voltus, CPower, NRG Curtailment Solutions). Get firm capacity-payment quotes for 100kW, 200kW, 500kW shed levels at 30-minute and 10-minute response times. The 10-minute tier pays meaningfully more.
- Singapore: EMA's Demand Response programme through SP PowerGrid. Less mature than PJM but operational. Low-stakes way to dry-run the operational discipline.
- Taiwan: Taipower's iDR (interruptible demand response) and time-of-use tariff optimization. With Taipower's tiered rates for data centers now active, peak-shaving directly reduces tariff tier — an even larger lever than DR payments alone.
- Cross-cutting: Negotiate "no penalty for non-performance in summer cooling extremes." This is the trap clause. You want capacity payments, not exposure to involuntary curtailment penalties when a heat wave hits.
Weeks 7-12: Wire it into the BMS — and prove it works
- Pre-cool sequences: Drop building setpoint by 1.5°F starting 90 minutes before predicted peak. Recover during peak by allowing 1.5°F drift. Most ASHRAE 55 thermal comfort envelopes tolerate this without complaint logs spiking — verify with your tenant satisfaction baseline.
- Chiller staging: Stage down one chiller during peak windows, with a hard 30-min recovery. This alone moves 100-300 kW on a typical 1MW office.
- EV charger throttling: If you have Level 2 charging, dynamic curtailment to 50% rated power during peak hours is invisible to drivers (most cars charge faster than the building peak window).
- Run 3 dry-run events per month for the first quarter — measure actual shed depth vs. promised, document tenant impact, refine setpoints. The dispatch settlement methodology in PJM is unforgiving; you want to be over-delivering by 10-15% on the test events.
- IPMVP-grade M&V: Adopt Option C (whole-facility) baseline with a 5-day pre-event regression model. If you cannot defend the baseline, you cannot defend the savings.
What about battery storage?
The economics flipped in 2026 for two reasons. First, capacity-payment thresholds in PJM now justify dedicated battery sizing — a 200 kWh battery cycling once daily for peak shaving plus capacity participation can pay back in 4-6 years on a 5,000 sqft tenant. Second, FERC Order 2222 in ISO-NE makes the same battery bid-eligible into ancillary services markets after November 2026, stacking a third revenue stream. The 2026 commercial energy storage ROI curve has shifted from "marginal" to "first-mover advantage" in PJM.zones with high capacity prices.
For Taiwan, the calculation is different but more urgent. With northern Taoyuan capacity frozen, BTM battery storage is the only path to expanding electrical capacity for tenants who need more load. Pair with rooftop PV and you have a defensible answer to Taipower's "no new supply" letter.
The trap to avoid: tenant communication
The single biggest failure mode in commercial DR programs is not technical — it's tenant relations. If a tenant gets one uncomfortable afternoon and complains to their broker, you have lost the asset's reputation for years. Three rules I would not break:
- Communicate before you dispatch. Email ops contacts at 9am for an afternoon event. Weekly digest of upcoming flex windows.
- Cap the comfort drift. ±1.5°F maximum, not ±3°F. The capacity payment is not worth an angry CFO at the anchor tenant.
- Share the upside. Some buildings now offer tenants a small rebate (10-20% of capacity revenue attributable to their floor) to align incentives. This converts tenants from "victims of cost-cutting" into "partners in flexibility."
Where this connects to the broader AISB stack
The grid-energy story is a specific instance of a general pattern we've documented across the Library: buildings are no longer passive infrastructure. They are continuously-bid assets in markets that did not exist five years ago. The same control-system maturity that enables AI-HVAC optimization also enables grid-services participation — same sensors, same setpoint logic, same M&V protocol. If your BMS is not exposing dispatchable load via API, that is the foundational gap to close before any of this revenue is reachable. See our companion piece on M&V standards (Option D's comeback) for the audit-trail discipline this requires, and our BaaS platform contract substrate piece for how multi-vendor stacks now coordinate.
Have a question about your specific building?
If you want to know whether your peak shape, location, and BMS maturity make grid-services participation a real opportunity or a distraction, talk to our agent. We can walk you through the specific PJM zone math, Singapore tariff structure, or Taipower iDR enrollment process for your asset.
Have a question about this topic? Ask our CRE AI Agent →