Energy Cases 2026: Spark Spread and Capacity Factor

The conversion that unlocks an energy case interview

The number an energy prompt gives you is a heat rate, and the number you need is a cost per kilowatt hour. Converting between them is the first thing to get right, and it is two multiplications.

A heat rate is British thermal units of fuel per kilowatt hour of output. A gas price is dollars per million British thermal units. So fuel cost per kilowatt hour equals the heat rate divided by one million, times the gas price.

At a 7,000 Btu heat rate and $4.00 gas, that is 7,000 divided by 1,000,000, which is 0.007, times $4.00, so $0.028 a kilowatt hour. Written the way the industry writes it, 2.8 cents.

Add variable operations and maintenance of 0.4 cents and the variable cost of one kilowatt hour is 3.2 cents. Subtract that from the power price and you have the spark spread, which is the contribution margin of a gas plant under another name. At a 4.5 cent power price the spread is 1.3 cents.

Two habits protect this step. Write the unit beside every figure before multiplying, because Btu, million Btu, kilowatt hour and megawatt hour are four bases and three of them are wrong. And convert megawatt hours to kilowatt hours early, since prices are quoted per kilowatt hour and output is quoted in megawatt hours.

Costing one plant, and reading the impossible answer

Here is the prompt. Every cost figure in it is ours, posed as an assumption, and a candidate should label their own the same way.

Cardinal Power owns one combined cycle gas plant and wants to know whether to keep it open.

1. Nameplate. 500 megawatts.
2. Heat rate. 7,000 Btu per kilowatt hour.
3. Fuel. $4.00 per million Btu.
4. Variable operations and maintenance. 0.4 cents a kilowatt hour.
5. Power price received. 4.5 cents a kilowatt hour.
6. Fixed cost. $60 million a year, covering the capital charge, fixed maintenance, staff and interconnection.
7. Current dispatch. A 50 percent capacity factor, set by where the plant sits in the merit order.

At nameplate, 500 megawatts for 8,760 hours is 4,380,000 megawatt hours. At a 50 percent capacity factor that is 2,190,000 megawatt hours, which is 2.19 billion kilowatt hours.

Contribution is 2.19 billion times 1.3 cents, so $28,470,000. Against $60 million of fixed cost the plant loses $31,530,000 a year.

Now solve for the capacity factor that would break even. Full output contribution is 4.38 billion kilowatt hours times 1.3 cents, so $56,940,000. Divide $60,000,000 by that and you get 105.37 percent.

Do not treat that as an error. It is the answer, and it is the most decisive sentence available: at a 1.3 cent spread this plant cannot cover its fixed cost even running every hour of the year. The recommendation is therefore not about dispatch. It is about the spread, the fixed cost, or closure.

Solving the energy case backwards for a number the client can act on

A capacity factor above 100 percent closes one door and opens the useful one. Hold the 50 percent capacity factor, which reflects where the plant sits in the merit order, and solve for the spread it needs.

$60,000,000 divided by 2.19 billion kilowatt hours is $0.0274, so 2.74 cents. Against the 1.3 cents the plant earns today, it needs more than double.

Translate that into the price the market would have to pay. Variable cost is 3.2 cents, so the required power price is 3.2 plus 2.74, which is 5.94 cents. Against 4.5 cents received, the plant needs 32 percent more than the market gives it.

That single figure reframes the case. Nothing inside the plant closes a 32 percent price gap, so the branches worth spending time on are the ones that change the denominator or the fixed cost. Contract the output at a fixed price above market. Retire the plant and stop the $60 million. Repower it, which resets both the heat rate and the fixed cost. Or bid it into a capacity market, which pays for availability rather than energy and is a different revenue line altogether.

Which sensitivity actually moves the answer

Two levers get proposed in every energy case, and they are not the same size. Test both.

The middle row is a 10 percent heat rate improvement, which is a major capital project. It adds $6,132,000 of annual contribution. The bottom row is a 25 percent fall in the gas price, which the plant does not control at all. It adds $15,330,000.

The ratio is $15,330,000 over $6,132,000, which is exactly 2.5. So at this heat rate the fuel price moves the result two and a half times as hard as a substantial efficiency gain. Neither row reaches the $60 million fixed cost on its own, and saying so is part of the answer.

The engineering reading of this is worth stating plainly. A heat rate gain scales with the fuel price, because the saving is a percentage of a fuel bill. When fuel is cheap, efficiency projects are worth less, which is the opposite of the intuition most candidates bring.

Never blend a regulated and a merchant business in an energy case

Utilities routinely contain both a regulated network and a merchant generation arm, and their economics are unrelated. A regulated network earns an allowed return on a rate base set by a commission, so its revenue is a function of invested capital and its volume risk is limited. A merchant generator earns a spread, so its revenue is a function of a market price it does not set.

A blended margin across the two produces a number that describes neither, and it is the most common structural error on this case type. Split them in your first branch and keep them split.

The Energy Information Administration shows how far a blended price can mislead even inside one business. Its table 5.3 gives 2025 US averages of 17.30 cents per kilowatt hour for residential customers, 13.41 cents for commercial, 8.62 cents for industrial and 13.63 cents for all sectors, checked 30 September 2026 at eia.gov. The residential average is 2.01 times the industrial average.

So when a prompt says customers pay a given number of cents, ask which class. The same utility serving the same grid earns roughly double per kilowatt hour from a household compared with a factory, and the difference is cost of service and rate design rather than margin.

For the asset side, the Federal Reserve's release of 18 September 2026 reports the following. "The operating rate for mining rose 0.1 percentage point to 86.3 percent, and the operating rate for utilities increased 1.1 percentage points to 71.3 percent", published at federalreserve.gov. The same release puts the long-run utilities average at 84.0 percent, so the sector was running 12.7 points below its own normal rate.

Ask which horizon an energy case recommendation is measured on

Before recommending anything, ask what bar the client underwrites. Published cases disagree sharply, and an energy asset has a life long enough that the disagreement changes the answer.

Oliver Wyman's own practice case is the clearest example in print. An Indian gas operator considers distributing drinking water, and the published lines are "Annual revenue: Rs. 750,000,000, assuming 300 operational days per year" and "Total annual costs: Rs. 575,000,000 or 57.5 crore". Recomputing from the stated inputs: 250,000 kiloliters a day across 300 days is 75 million kiloliters, at 10 rupees that is 750 million rupees or 75 crore, variable cost at 5 rupees is 375 million or 37.5 crore, plus 20 crore of fixed cost gives 57.5 crore, so annual profit before tax is 17.5 crore. The firm's three lines all reconcile.

Then it refuses the project, because entry "does not align with the client's investment criteria for a five-year payback period", checked 30 September 2026 at oliverwyman.com. A profitable business, declined on a time criterion.

Commodity risk is why that convention exists here. Gulf Coast jet fuel spot went from $2.024 a gallon in August 2025 to $3.724 in August 2026, a rise of roughly 84 percent inside twelve months, published at eia.gov. A five year payback on an asset exposed to moves of that size is a different risk from a five year payback on a regulated network.

Where the points go on a generation case

The scheme below is ours. We read the weights out of the grading code in src/lib/grading, and no firm publishes a marking scheme for a case interview, so these are our dimensions rather than anyone's official ones.

Where the public record runs out

We looked for a first-hand candidate account of an energy case and found none we can cite. Reddit refuses connections from this network, and no video transcript tooling was available in the session. Oliver Wyman publishes two consultant profiles and neither discusses its six practice cases. McKinsey publishes two recruiting blog posts with the same silence. Competitor preparation blogs cover energy cases at length and are not citable sources.

Five figures a case needs are not published anywhere we could reach. Nobody publishes the heat rate of a named plant, so the 7,000 Btu figure is ours. Nobody publishes fixed cost per megawatt of capacity, so the $60 million is ours. The Energy Information Administration publishes average prices paid by customer class and not the wholesale price any particular generator receives, so the 4.5 cents is ours. Capacity market clearing prices exist in public auction results, and we did not read them, so this page does not quantify that branch. And no firm publishes a hurdle rate for approving a generation asset, which is exactly why the question above is asked rather than assumed.