AI investment thesis
The margin does not shrink. It moves to whoever instruments first.
Every deregulation cycle in US power has triggered the same pattern: incumbents lose margin, new entrants capture it through operational speed. AI is the current cycle. The companies that embed intelligence into grid operations, billing, and asset management will set margin floors their competitors cannot reach.
The thesis in three sentences
US power is a $500B industry where profit concentrates in nine activities spanning generation, transmission, distribution, and retail. AI does not eliminate these activities — it redistributes margin from slow operators to fast ones. The companies that instrument grid operations, automate billing, and build asset intelligence from operational telemetry will capture 8-15% net margin expansion while competitors absorb the same commodity price exposure with no hedge.
The margin migrates. The question is whether it migrates toward you or away.
What displacement looks like
| Before AI | After AI | |
|---|---|---|
| Load forecasting | 8-12% MAPE, day-ahead only, weather-adjusted regression | 2-4% MAPE, sub-hourly, operational signal fusion (load + grid + weather) |
| Billing operations | Manual dispute resolution, 5-7 day cycle, 8-12% annual churn | Real-time resolution, voice AI end-to-end, churn drops 12-18% |
| Asset valuation | Book depreciation, annual inspection, estimated useful life | Continuous telemetry, predictive maintenance, residual curves from operational data |
| Grid optimization | Static switching schedules, conservative margins, manual SCADA | Dynamic optimization, real-time DER coordination, autonomous fault isolation |
| Energy trading | Historical pattern matching, human traders, hourly granularity | ML price forecasting, 5-minute granularity, automated dispatch optimization |
5 of 9 activities
What acceleration looks like
AI displacement replaces manual processes. Acceleration compounds each stage through the energy value chain — better forecasting feeds better dispatch, which feeds better trading.
2-4% MAPE
Renewable forecasting
Deep learning models fuse satellite imagery, weather data, and historical generation to predict solar and wind output sub-hourly. Accurate forecasts reduce curtailment and enable firm power contracts.
15-30% revenue uplift
Grid dispatch optimization
Real-time dispatch algorithms place generation and storage where congestion premiums are highest. Each cycle of optimization data improves the next forecast.
$25-40/MWh spread
Battery arbitrage
Charge/discharge timing moves from static schedules to dynamic arbitrage. AI captures intra-day spreads that manual operators cannot see at 5-minute granularity.
60% faster resolution
Billing automation
Voice AI handles rate plan inquiries, dispute resolution, and usage explanations end-to-end. Faster resolution drives lower churn and higher customer lifetime value.
Real-time valuation
Residual curve pricing
Operational telemetry from battery, solar, and turbine assets generates continuous residual curves. Asset-backed financial instruments reflect actual condition, not book depreciation.
"The highest-margin activities in power are the most exposed to AI displacement. They are high-margin precisely because they require operational intelligence that is expensive to build. AI replicates that intelligence. The margin migrates to whoever instruments first."
Moative power and utilities thesis, May 2026
The two activities that break the pattern
Physical grid maintenance is augmented, not displaced. Lineworkers, substation technicians, and field crews remain human. AI assists at the edges: predictive maintenance flags transformers likely to fail, drone inspection routes optimize themselves, digital twins simulate grid configurations before physical switching. But the physical work — the climbing, the switching, the emergency restoration — stays manual. The margin impact is indirect. Better prediction reduces unplanned outages, which reduces overtime costs and regulatory penalties.
Regulatory proceedings stay human-driven. Rate cases, interconnection queues, and compliance filings require legal judgment and political navigation that AI cannot replicate. AI accelerates the analysis — load studies, cost-of-service models, rate design simulations — but the testimony, the negotiation, the stakeholder management remain human capital.
The rebuild order is not arbitrary. Which activities you instrument first determines whether the next phase gets clean inputs or garbage.
Map the profit pool
Identify where your organization sits on the displacement curve for each of the nine activities. Not all are equal. Start where the margin moves fastest — typically forecasting and dispatch.
Sequence the instrumentation
Forecasting drives dispatch accuracy. Dispatch drives trading margin. Trading margin funds the next phase. Instrument in that exact sequence.
Deploy activity by activity
Each activity is a standalone rebuild. Ship one, measure margin impact, use the operational data to improve the next. The compounding effect accelerates — by activity three, the data layer is rich enough to train cross-functional models.
Moative arrives with the thesis written and operates through rebuild on equity: cash at risk, not hours billed.
Where the thesis applies
Read the thesis for each activity
Renewable forecasting→
Solar and wind output prediction drives curtailment costs and grid balancing.
Grid optimization→
Real-time switching, fault detection, and DER coordination.
Battery storage→
Charge/discharge optimization for grid-scale storage.
Customer analytics→
Churn prediction, rate optimization, consumption patterns.
Energy trading→
Price forecasting for ERCOT, PJM, CAISO wholesale markets.
Demand response→
Industrial and commercial load curtailment scheduling.
EV charging→
Load management across charging networks.
Data center power→
Cooling optimization, infrastructure sizing, procurement.
Billing platforms→
Rate plan complexity, dispute resolution, invoice automation.
The profit pool shows where the thesis lands
The interactive visualization maps all nine activities by revenue share, operating margin, and AI impact. Click any activity to see the thesis in numbers.
Open the profit poolExplore the cluster
More on power and utilities
Cluster overview→
Nine profit pools, three structural transitions, and the AI activities reshaping US energy.
Profit pool→
Which activities capture margin today and how AI restructures the value chain over five years.
AI shift timeline→
How each profit pool activity transforms over the next five years as AI adoption accelerates.
Questions about the power and utilities AI thesis
Why does AI favor first movers in energy?
AI models improve with operational data volume. The company that instruments first builds a compounding advantage — better data feeds better models, which drive better decisions, which generate more data. By month 18, the leader's models have been refined through millions of dispatch cycles the second mover hasn't seen.
Which activities are most AI-addressable?
Load forecasting, battery dispatch optimization, and billing automation show the fastest ROI. These activities have high data density, clear performance metrics, and existing operational workflows that AI can improve without replacing infrastructure.
How much margin expansion is realistic?
First movers in grid optimization and trading have demonstrated 8-15% net margin expansion in the first 24 months. The range depends on starting operational data maturity and the specific market structure.
Does this thesis apply to regulated utilities?
The thesis applies differently. Regulated utilities capture value through rate case efficiency gains and reliability improvements rather than trading margin. The instrumentation phase is the same — the financialization phase routes through regulatory proceedings rather than wholesale markets.
What is the capital requirement to start?
Instrumentation typically costs $500K-2M for a mid-sized utility. Most of the data already exists in SCADA, EMS, and billing systems — the gap is making it queryable and real-time. The ROI timeline is 6-12 months for forecasting improvements, 12-18 months for billing automation.