Researched by BEAST Library Curator | Verified by Harper | Quality: 8.3/10
BLUF: PJM's 2026/27 capacity auction cleared at the cap of $329.17/MW-day — a 22% spike on top of last year's record. Bills land June 2026. Data centers drove ~40% of that $16.4B cost. CAISO's Extended Day-Ahead Market launches today (May 1, 2026), reshaping Western dispatch economics. Taipower's VPP just crossed 30 MW from ~50 C&I sites. The same owner playbook — measure, flex, monetize — works in all three regions. Here's the 90-day version.
1. What just changed (and why your June bill is the canary)
Three grid signals converged in the last 60 days, and they tell the same story: structural scarcity is here, and CRE owners who treat their building as a passive load are about to subsidize hyperscale AI infrastructure on their utility bill.
| Signal | Date | Hard Number | What It Means for CRE |
|---|---|---|---|
| PJM 2026/27 BRA cleared | July 2025 → bills June 2026 | $329.17/MW-day (cap), +22% YoY, all zones cleared at cap | Capacity-line items rise ~22%; total bill +1.5% to +5% depending on state/utility mix |
| Data-center share of capacity cost | PJM IMM Q4 2025 | $6.5B of $16.4B (~40%) | Non-data-center customers cross-subsidizing AI load growth |
| CAISO EDAM go-live | May 1, 2026 (today) | $543M/yr operational savings, 2.92M t CO₂/yr (West-wide) | Day-ahead price discovery improves; flexibility gets paid more accurately |
| Taipower VPP / Enel X | Live 2024–2026 | ~30 MW from ~50 C&I sites in Supplemental Reserves | APAC parallel: TOU + CPP + ancillary services pay flexible buildings now |
| C&I BESS LCOE | 2026 deployments | $250–550/kWh installed; 3–6 yr payback; ITC 30–40% | Demand-charge arbitrage is bankable inside lease term |
Sources: PJM 2026/27 Base Residual Auction Report (July 2025); PJM Independent Market Monitor; CAISO EDAM Fact Sheet (April 2026); Enel X Taiwan press materials; vendor LCOE benchmarks aggregated April 2026.
The PJM number deserves a closer read
$329.17/MW-day was the auction cap, not the equilibrium price — supply was insufficient to clear below it. PJM procured 134,311 MW of unforced capacity for the delivery year. The IMM's data-center attribution (~40%) is the load-side story behind that cap-clear: forecasted peak-load growth for 2027/28 is ~5,250 MW, of which ~5,100 MW is data-center expansion. Translation: the grid is building for AI, and every other ratepayer is paying the construction tab through capacity charges.
If your portfolio sits in PJM's footprint (13 states + DC), this is not abstract. Your June 2026 invoice will show it, and it compounds: the 2025/26 delivery year already had a record-clearing of $269.92/MW-day; 2026/27 is on top of that base.
2. The owner-operator escape valve: from price-taker to grid-revenue earner
Here's the asymmetry: capacity prices are clearing higher because flexible capacity is scarce. A commercial building that can shift 200 kW for 4 hours in summer afternoons is precisely the resource the grid is now willing to pay $329/MW-day for. The DOE's Grid-Interactive Efficient Buildings (GEB) framework has been pointing here since 2021; the economics finally caught up.
The four revenue/cost-avoidance vectors a typical 250–500k sf Class A office can stack:
| Vector | Typical $ Impact (per 100 kW shed) | Activation Time | Required Capability |
|---|---|---|---|
| Capacity-payment (utility/aggregator) | $8K–$15K/yr | 2–4 mo onboarding | Verifiable curtailment, M&V Option C baseline |
| Demand-charge management (peak shaving) | $25K–$60K/yr | Day-1 once BESS or load-control online | 15-min interval data + automated dispatch |
| TOU/CPP arbitrage | $5K–$20K/yr | Immediate (rate election) | Schedulable HVAC pre-cool / BESS |
| Ancillary services (frequency reg, reserves) | $15K–$40K/yr | 3–6 mo (telemetry build) | Sub-second telemetry, OpenADR 2.0b |
Ranges are based on PJM/CAISO/ISO-NE C&I aggregator program disclosures and vendor case studies. Actual values are deeply portfolio-specific — utility, climate zone, occupancy, lease structure — and require a site-level M&V baseline to underwrite.
What I'd do if this were my building: Stop optimizing the energy line item in isolation. Treat the building as a small grid-services asset and target $80–$120/sf-yr in stacked revenue/avoidance over a 3-year ramp. The economics break even faster than most office-leasing assumptions; the hard part is operational discipline, not capex.
3. The Taiwan / APAC parallel — same physics, different paperwork
For owners with Asia exposure (or anyone reading this from Taipei), Taipower's path validates the Western direction:
- AMI coverage — Taipower has deployed ~3.4M smart meters covering ~81.5% of system load. Sub-15-minute data is the precondition for everything that follows; without it, you can't bid into ancillary services or even prove demand-charge savings.
- Rate-design instruments live — IBR, Seasonal Rate, TOU, and Critical Peak Pricing (CPP) are all in production. CPP is the highest-leverage instrument for offices: a single peak event can move a quarterly bill 3–6%.
- VPP aggregation operational — Enel X's Taiwan VPP aggregated ~30 MW from ~50 C&I users within a 30-minute notification window for Taipower's Supplemental Reserves program. Taipei 101 has been a public anchor since 2022.
- Energy Trading Platform (since 2021) — Taipower's market is the APAC analog to PJM/CAISO ancillary services. Aggregators bid in; buildings get paid through the aggregator.
The differences from PJM/CAISO are jurisdictional, not technical: contract templates, notification windows, settlement cycles, and FERC-vs-MOEA approval paths. The control-system architecture, the M&V protocol, and the financial logic translate cleanly. See our other library entries for IPMVP-anchored approaches.
4. The 90-day GEB playbook
Days 0–30: Instrument and baseline
- Pull 24 months of 15-minute interval data from every utility account. If you only have monthly bills, request hourly via the utility's Green Button or AMI portal — Taipower, ConEd, PG&E, BGE, Pepco all support this. Without 15-min data, you cannot bid into capacity programs and cannot underwrite a BESS.
- Compute load shape, peak-hour cluster, and demand-charge contribution. A useful diagnostic: what % of your annual electricity bill comes from the worst 15-minute interval each month? In Class A office, the answer is usually 20–35%. That's your headroom.
- Establish IPMVP Option C whole-facility baseline. Pre-period of 12+ months. Document independent variables (HDD/CDD, occupancy, plug load). This becomes the verification artifact for every program below.
Days 30–60: Pick two revenue vectors and onboard
- Vector 1 — Demand-charge management. Lowest-friction. Two paths: (a) automated load-control on chillers/AHUs via your BAS with a peak-prediction overlay; (b) BESS sized for the 4-hour ratchet (typically 200–800 kWh per 100 kW peak). Vendor LCOE in 2026 sits at $250–550/kWh installed for commercial systems; 3–6 yr payback in high-tariff zones, faster with the IRA ITC 30% (40% with domestic content) still in force.
- Vector 2 — Capacity-payment program. Onboard with an aggregator (Enel X, CPower, Voltus, Enchanted Rock) rather than direct ISO participation — the aggregator handles telemetry, M&V, settlement. Standard contract: 2–4 hour curtailment, ~10 events/yr in PJM/NYISO summers; payment based on registered capacity × clearing price.
Days 60–90: Document, monetize, prove
- Run two test events in Q3. Document the curtailment delta against baseline using your IPMVP Option C method. This is the artifact your CFO, your insurer, and your tenant will all want to see.
- Update the lease language for cost recovery. Critical and frequently missed: in a triple-net lease, demand-charge savings flow to the tenant unless the lease addresses grid-services revenue. Most pre-2024 leases are silent. New lease drafts and renewals should explicitly carve out grid-services revenue (capacity payments, ancillary services) as the landlord's, while pass-through energy savings remain with the tenant. Get this in front of legal before signing the aggregator contract.
- Stack vector 3 (TOU/CPP) once the schedulable HVAC is proven. Pre-cooling 1–2°F starting 90 minutes before peak is invisible to occupants and routinely worth $5K–$20K/yr per 100k sf. Ask our CRE AI Agent to run the math on a specific building.
5. What to watch through Q3 2026
| Watch Item | Date Window | Decision Trigger |
|---|---|---|
| First PJM utility bill cycles with 2026/27 capacity charges | June–August 2026 | If capacity line item >5% of total → escalate GEB enrollment from "explore" to "execute" |
| CAISO EDAM real-time price discovery | May–July 2026 (post-launch settling) | PacifiCorp/PGE territories: re-evaluate TOU election once 60-day price data is visible |
| FERC Order on PJM co-located data-center tariff | Filings due Jan/Feb 2026, decision Q2–Q3 2026 | Outcome materially shifts who pays for grid expansion; could ease or worsen the cross-subsidy |
| PJM 2027/28 BRA results | Auction expected mid-2026 | If clears at cap again, plan capacity-line-item budget at +20–25% YoY through 2028 |
| Taipower CPP event frequency 2026 summer | June–September 2026 | Taiwan portfolios: validate VPP enrollment ROI on hottest 10 days |
6. The honest caveats
Three things that don't fit on the spreadsheet but matter operationally:
- Tenant comfort is the silent kill criterion. A 2°F pre-cool is invisible. A 4°F set-point relaxation during a CPP event is not. Run the first three events with conservative dispatch and an FM in the building. The dollar value of one angry CFO tenant exceeds two years of demand-response payments.
- BESS warranty cycles vs. dispatch frequency. Most LFP commercial systems are warranted to 6,000–8,000 cycles at 80% DoD. If you dispatch daily for arbitrage and for capacity events, you may exit warranty inside 8 years. Model the cycle budget before sizing.
- Aggregator contracts have penalty clauses. Under-delivery on a registered capacity event is typically penalized at 1.5–2× the payment rate. Conservative registration beats optimistic registration; let observed performance pull the registration up over 2–3 cycles.
7. Bottom line
The grid told CRE owners three things in the last 60 days: (1) capacity is structurally scarce and getting more expensive; (2) hyperscale AI is the load-growth story, and you're paying for it on your bill; (3) the same flexibility the grid is paying premium prices for is sitting unused in your HVAC plant and BAS. The 90-day playbook above is not novel — DOE's GEB framework, IPMVP Option C, and aggregator programs have been mature for years. What's new is that the math now works inside a 3-year hold, in PJM, in CAISO, and in Taipei. The owners who move in Q2 will have receipts by Q4. The owners who wait will have invoices.
Have a question about this topic? Ask our CRE AI Agent →