La Bourse / Volume IV / Nº IV.05 / Workbook — the executive

A watercolour of rolling hills in late summer, a farmhouse and a line of trees on the far ridge. — Plate IV.05 · Workbook — the executiveThe Ten Thousandth.The ten thousandth panel costs less than the first. Nobody economised. Ten thousand people each learned something small and none of them could tell you what.

WORKBOOK — THE EXECUTIVE

Chapter IV.05 · Energy Systems That Compound

For the person with a P&L, a signature authority and a capital committee that meets monthly. This is not an energy workbook. It is a procurement and capital allocation workbook that uses energy as the case where somebody kept the records for forty-seven years.

THE PREMISE, STATED COMMERCIALLY

Your cost base contains two kinds of item and your schedule shows them in one column.

The first kind gets cheaper as the world makes more of it. The second gets more expensive as the world uses more of it up. They have the same functional form and opposite signs on the exponent, and treating them identically — same discount rate, same forecast model, same contract length, same negotiating posture — is the most expensive unforced error available to a capital committee.

The commercial consequence is specific and it is not about energy. For an item on a falling curve, your own purchase lowers your own future cost, which means commitment is the lever and price is the output. For an item on a rising curve, your own purchase raises it, which means optionality is the lever and commitment is the risk. Most procurement functions run the second playbook on both, because that is the playbook procurement was built for.

You already know one of these curves exists. Solar module cost fell from 106 dollars a watt in 1976 to 0.15 dollars a watt in 2023 — 22.17 doublings at a fitted 22.5 percent each, interval 19.1 to 25.7 percent. What this workbook does is take that instrument off the energy page and point it at your bought-in components, your service lines and your own manufacturing.

PART ONE — DISCOVERY

Days 1–30: find your own exponents

Exercise 1.1 — The three series (one analyst, two weeks)

Pull cumulative volume and real unit cost for three things: one component you buy, one component you make, and one service you deliver repeatedly. Deflate to one currency year — this is the single most common error and it silently flatters every series in a rising-cost period.

Ten points each is enough. You are looking for the shape, not the fourth decimal.

Exercise 1.2 — The exponent register (one afternoon)

Fit each series: log real unit cost on log cumulative volume. Report slope, standard error, R², learning rate, interval for each. Put them in one table and give it a name — the exponent register — and put the date on it.

Expect three outcomes and treat each as information:

A clean fit with a positive rate. You have found a lever. Go to Part Two.
A poor fit. Almost always a regime change inside the series. Cut it at the change; do not average across it.
A negative rate. The most valuable finding on the page. Something is getting more expensive the more of it you do, and nobody has put those two columns side by side before. The chapter's case is a national one — 58 reactors, 5.86 doublings, a cost increase of about 3.50×, a learning rate of −23.8 percent — and the mechanism transfers exactly: large, bespoke, site-assembled, never repeated.

Exercise 1.3 — The appreciative production interview (half a day)

Go to the line, the desk or the depot and ask the people who do the repeated thing: what is different between the tenth one and the thousandth? Write down every mechanism, not every opinion. Kavlak and colleagues decomposed the solar decline into named mechanisms and found the weighting moved over time — research-led before about 2001, scale-led after. Your own decomposition tells you which lever is still available and which is exhausted.

PART TWO — THE ARITHMETIC

Days 31–45: size the prize before you commit anything

Exercise 2.1 — The volume-commitment case, worked

Here is the shape of the paper. A component made internally 2,000 times a year at $46 a unit, with an observed learning rate of 12 percent and five years of history. A proposed commitment triples annual volume for three years.

  cumulative before                      10,000 units
  cumulative after three years           28,000 units
  doublings                              1.49
  unit cost after                        $38.04
  annual saving at the new volume        $47,734

Note what that paper does not claim. It does not claim the saving is free — it is bought with a volume commitment that has to be justified on demand, not on the learning rate. It does not claim the rate will hold; the interval is on the page beside it. And it does not claim the saving is the whole benefit, because the lower cost persists after the commitment ends.

Exercise 2.2 — The three-numbers test (60 minutes, with your analyst)

Take any learning rate already circulating inside your firm and ask what was in the numerator. The chapter's worked case is onshore wind, where one technology produces three defensible numbers over the same 2.42 doublings:

  naive LCOE learning rate           39.4 %
  capital-cost learning rate         23.0 %
  published turbine learning rate    12 %

The decomposition: capital cost fell to a ratio of 0.5324 while capacity factor rose from 27 to 36 percent, a ratio of 0.7500, predicting an LCOE ratio of 0.3993 against an observed 0.2973 and leaving 0.7446 for operating cost, cost of capital, project life and country mix.

Make this a standing question in your capital papers. What is in the numerator, and what else moved?

Exercise 2.3 — Sensitivity, stated the way a board can use it

Never present a learning projection as a number. Present two spreads. For solar at 16,000 GW, the entire measured learning interval moves the answer by 1.30×; the deployment range from 4,000 GW to 16,000 GW moves it by 1.67×. Your paper should say which of the two your firm actually controls. Usually it is the second, and usually it is the one nobody discussed.

Exercise 2.4 — The energy line itself, priced honestly

If your firm is buying renewable energy, your LCOE comparison is incomplete and your operations team already knows it. Add the firming line and compute it: $250/kWh installed, 300 cycles a year, 85 percent round-trip, fifteen-year life, 7 percent cost of capital gives a capital recovery factor of 0.1098, an annual charge of $27.45 per kWh of capacity, 255.0 kWh delivered, and a shifting cost of $0.1076/kWh — making a delivered shifted kilowatt-hour $0.1516 against a prompt $0.0440, or 3.4×.

Worked on a real shape: a 100 MW farm producing 220 GWh a year, with 40 percent of output shifted to the evening, moves 88.0 GWh, adds $9.47 m to a base of $9.68 m, and blends to $0.0871/kWh — 1.98× the headline.

That number belongs in the paper the first time, not in month eight when operations produces it. A procurement that survives its own operations review is worth more than one that wins its approval meeting.

PART THREE — DESIGN

Days 46–60: the instrument

The deployment-indexed offtake agreement.

A conventional fixed-price contract buries a disagreement about the learning rate inside a single number, where it cannot be negotiated. This instrument makes it explicit.

Volume. Firm annual quantity over ten years, with a take-or-pay floor at 80 percent of nominal. That floor is the consideration you are actually offering, and it is what makes the counterparty's line bankable.
The index. A named public series of cumulative world production, by publisher and edition, with a stated successor if publication ceases. Never an index either party maintains.
The price path. P(t) = P0 · (Q(t)/Q0)^(−β). β is the seller's implied learning rate and it is the whole contract.
Collar. A floor protecting the seller against a deployment collapse, a cap protecting you against an index revision, both in currency and not in adjectives.
Re-fit. At year five, re-estimate the learning rate on the intervening data and re-strike β inside the collar.
Firming priced separately, on its own schedule and its own index. A blended price cannot be renegotiated in parts.

The worked case for the board pack. Five hundred gigawatt-hours a year for ten years. A flat offer at $52/MWh is $26.00 m a year and $260.00 m over the term. The same volume indexed gives a ten-year total of $226.15 m at an implied 10 percent learning rate, $198.07 m at 19.1 percent, and $188.22 m at 22.5 percent — savings of $33.85 m to $71.78 m against the flat path, with no change in volume and no change in counterparty.

The balance-sheet treatment. The volume commitment is a purchase obligation and is disclosed as one. The indexed price path is not a derivative if it settles in physical delivery for own use, and the own-use determination is where this transaction stays simple or becomes a quarterly fair-value exercise — take advice in week one, not week nine. Where you fund capacity directly, capitalise against the asset created rather than expensing: a payment that produces a durable cost reduction is capital expenditure, and the chapter's whole argument is that this has been misfiled for twenty years.

Governance. Learning-rate claims in capital papers get the same treatment as forecast revenue: stated with an interval, sourced, and re-estimated annually by somebody who did not write the original paper. One sentence in the capital manual does this.

Exercise 3.1 — The counterparty choice (90 minutes)

Do not start with energy. Start with an existing supplier on an existing product where you already hold five years of cost history, because Part One requires a measured exponent and you cannot measure a counterparty you have never bought from. Energy is the obvious application of this instrument and it is the wrong first one.

Score your candidates on four things, in this order:

Do we have the series? No series, no contract. This eliminates most of the list and it eliminates it correctly.
Is the annual spend large enough to matter and small enough to sign? The test is the same one the ninety-day method uses throughout this edition: the effect must be visible above ordinary period variation, and the commitment must sit inside one persuaded person's authority.
Is the product architecture one that learns? Small, repeated, standardised, factory-made. A bespoke, site-assembled, rarely-repeated item will not deliver an exponent however hard you negotiate.
Is the counterparty capacity-constrained? A supplier who cannot expand has nothing to sell you in exchange for your volume floor, and the whole consideration of this instrument evaporates.

Exercise 3.2 — Rehearse the negotiation on paper (60 minutes)

Write the four sentences you will actually say when the seller's price path implies a learning rate below the measured lower bound. They are shorter than you expect and they are more effective written down than improvised:

Your path implies a learning rate of X. The published measurement on this class of product is between 19.1 and 25.7 percent, here is the fit and here is the standard error. At your implied rate we would be paying you for a cost reduction that the world is going to deliver anyway. We are offering a ten-year volume floor; price the floor, not the forecast.

Then write the seller's three best replies and your answer to each. The strongest is usually that their own volume is not the world's volume — which is true, is the honest core of the negotiation, and is exactly what the collar is for.

PART FOUR — DESTINY AND DELIGHT

Days 61–90: make it hold, and tell one true story

Three things make it survive your tenure.

It is in the capital template. A question in the standing paper — what is the exponent on cumulative volume for this item, and what is its interval — persists long after the person who introduced it. A question that lives in a template outlives every champion.

Somebody owns the exponent register and re-fits it annually. Name them. Fifteen minutes a year per series, and it is the only defence against quoting a rate that stopped being true in 2021.

One contract exists. A single executed deployment-indexed agreement is worth more than any amount of advocacy, because the second one is a precedent rather than a proposal.

The failure modes, named so you can see them coming.

The curve flattens at a material floor. Modules are 19.8 percent of a utility system's cost; if only modules learned the system rate would be 4.5 percent and free modules would leave $0.0353/kWh standing. Watch for three consecutive periods of shallowing slope, then re-fit rather than extrapolate.
The world market fragments. The rate is a global rate on global volume. Tariffs and local-content rules split one curve into several smaller ones that double less often. Choose that trade if you wish, but put the number beside the choice so it is a decision and not a drift.
Learning is assumed rather than measured. The moment a rate is assumed because a technology is on the right side of an argument, the instrument has been inverted and will produce confident nonsense.
The soft cost is never attacked. The 3.0× gap between $3.00/W and $1.00/W residential solar in two countries with identical hardware is made of forms and queues. Your own equivalent is real and it is entirely inside your control, which is why nobody has an excuse and nobody has done it.

The board paper, one page.

Section	Content
The finding	Three series fitted; rates and intervals; the negative one named first
The prize	Annual saving at committed volume, with the interval, not a point
The instrument	Volume floor, public index, `β`, collar, five-year re-fit
The risk	Demand risk on the volume floor — sized against the balance sheet
The omission	Firming and integration, priced separately and shown
The decision	`β` implied by the seller against the measured interval

The number that decides it. If the seller's implied learning rate sits below the measured lower bound — below 19.1 percent on the chapter's solar estimate — you are paying for a cost reduction the world is about to hand you for nothing. If it sits above the upper bound of 25.7 percent, the seller has promised more than the evidence supports and the collar is the only thing protecting delivery. Between the bounds is a real negotiation, and it is the only region in which this contract should be signed.

Delight, and it is a commercial asset. The pleasure in this work is the moment a finance team realises it has been holding two different asset classes in one column, and that one of them has been quietly getting cheaper every time anybody bought any. It changes the posture of a whole function — from defending a price to buying a curve — and people who have made that move do not go back.

APPRECIATIVE QUESTIONS FOR YOUR LEADERSHIP TEAM

Which of our costs has fallen without any programme making it fall, and who noticed first?
Where have we already committed volume rather than negotiated price, and what did the supplier do with the certainty?
If every capital paper carried an exponent and an interval, which decision from the last two years would have gone differently?
What is the largest line in our cost base that we have never plotted against cumulative volume?
Who in this business could re-fit the register annually in fifteen minutes a series, and what would we have to stop asking them to do?
Where is our own soft cost — the part that is made of forms rather than things — and what would a third of it be worth?