AI Power Exposure & Earnings Model
What it calculates
A headline megawatt announcement is not the same as operating load or public-company earnings. This model adjusts capacity for completion probability, utilization, PUE, and commercial timing, then estimates electricity demand, merchant-power revenue, utility rate-base economics, and equipment or EPC opportunity across delay scenarios.
How this ai power exposure & earnings model works
The model probability-adjusts announced capacity, phases it by commercial timing, and converts IT load through utilization and PUE into annual electricity demand. Separate assumptions translate that load into merchant revenue, regulated rate-base opportunity, and equipment or EPC exposure.
Formula
Annual TWh = announced MW × completion probability × utilization × PUE × operating-year share × 8,760 ÷ 1,000,000
Grid queues, permitting, PPAs, fuel costs, regulation, financing, and ownership decide who captures the economics. The output sizes a scenario; it does not identify a winning stock by itself.
Before you use the result
Assumptions
- • Announced megawatts are not treated as operating capacity until adjusted for completion and timing.
- • Merchant power, regulated utility, and supplier economics use separate editable margins and capital assumptions.
- • The result does not assign a project to a company without contract and service-territory evidence.
Quick start
- 1. Enter the announced capacity and its current probability of completion.
- 2. Set utilization, PUE, power price, and the share of a full year expected online.
- 3. Stress delay, cancellation, rate-base, and supplier-content assumptions independently.
Inputs and units
| Input | Unit | Default |
|---|---|---|
| Announced IT capacity | megawatts | 1,000 MW |
| Completion probability | percent | 70% |
| Capacity factor | percent of hours at full load | 90% |
| PUE | ratio of facility power to IT power | 1.2× |
| Operating in first year | months | 6 |
| Electricity price | US$ per megawatt-hour | $55 |
| Utility capex | US$ per kilowatt | $650 |
| Equipment / EPC spend | US$ per kilowatt | $900 |
Worked example (hypothetical)
AI Power Exposure & Earnings Model: worked example
Hypothetical 1 GW announcement using the calculator's default inputs. It does not describe a specific project.
| Announced × probability | 1,000 MW × 70% |
|---|---|
| Capacity factor × PUE | 90% × 1.2 |
| Months online in year one | 6 |
| Probable IT load | 700 MW |
|---|---|
| Annualized electricity | 6.62 TWh |
| First-year electricity | 3.31 TWh |
| First-year electricity spend | $182,120,400 |
1,000 MW announced becomes 700 MW of probable IT load and 3.31 TWh in the first year once completion odds, PUE, load factor, and timing are applied.
How to interpret the result
Announced megawatts overstate near-term demand. The merchant, regulated-utility, and equipment outputs are different economic claims on the same load, so they should not be added together as one company's opportunity.
Frequently asked questions
Why probability-adjust announced megawatts?
Projects can be delayed, resized, or cancelled before they become operating load. Multiplying announced capacity by an explicit completion probability keeps the scenario from treating every press release as certain revenue.
What does PUE change?
Power usage effectiveness compares total facility electricity with electricity used by computing equipment. A higher PUE increases total site demand for the same IT load.
Does electricity spend equal utility profit?
No. Revenue, fuel cost, regulated returns, transmission spending, contract structure, and ownership determine who earns the economics. The model keeps merchant, utility, and equipment cases separate.
Can this identify a winning power stock?
Not by itself. It sizes an exposure scenario. Investors still need project-level contracts, service territory, regulatory treatment, backlog quality, financing, and valuation evidence.
More calculators
The math is easy. The macro is hard.
AI Power Exposure & Earnings Model shows the arithmetic. Members get the monthly call that decides which inputs matter, plus written breakpoints when the facts move. $39/month or $249/year.
Members-only research on this topic
$39/mo or $249/yr · cancel future renewals anytime · sources included