12. Energy Economics & Policy

Every module in Parts B and D showed you how electricity is made and moved. This one asks a different question: what does it actually cost, and who decides who pays?

It turns out to be one of the more dramatic economic stories of the last decade — and, in India specifically, one of the more consequential policy failures still being repaired in real time.

Putting every technology on one scale

Comparing a coal plant to a solar farm on cost is harder than it looks. Coal has modest construction cost and a large, permanent fuel bill. Solar has meaningful construction cost and essentially zero fuel bill, ever. How do you compare “cheap machine, expensive fuel forever” against “expensive machine, free fuel forever”?

The answer is the levelised cost of electricity:

LCOE = (total lifetime cost, in today’s money) ÷ (total lifetime electricity produced)

“Total lifetime cost” bundles everything: construction cost, financing, operation and maintenance, fuel where relevant, and decommissioning — all converted into today’s rupees, because a rupee spent building the plant this year is worth more than a rupee spent maintaining it in year twenty. Divide that single lifetime cost figure by the total units the plant will ever produce, and every technology, however differently shaped its costs are, reduces to one comparable number: rupees per unit.

Indian LCOE-style figures for FY 2024-25/2023-24 tell a story on their own:

Source	Roughly ₹ per unit
Utility-scale solar (competitive auctions)	₹2.15–2.68
Wind (competitive auctions)	₹3.42–3.81
Nuclear (average tariff)	₹3.83
Existing coal fleet (weighted average, actual utilisation)	₹5.43

Solar’s record low so far is ₹1.99 per unit, discovered in a GUVNL auction in December 2020 — a whisker under the ₹2.00 SECI had set the month before. Notice something important in that table: solar and wind win on LCOE without needing to be perfect — they win because they have no fuel bill, ever, and that single fact dominates the arithmetic once construction costs fell far enough.

The price crash, and why it wasn’t a single breakthrough

Around 2010, Indian solar tariffs were several times today’s. What changed was not a single dramatic invention — it was a pattern that has held with startling consistency since the 1970s, known as a learning curve (or, for solar specifically, Swanson’s law):

Solar module prices have fallen by roughly 20% for every doubling of cumulative global production.

Notice the word cumulative. It isn’t about time passing — it’s about how much has ever been built, worldwide, added together. Double that total, and the price drops by about a fifth. Double it again, and it drops by about a fifth again. This is not a prediction about the future; it’s a description of what has actually happened, repeated for five decades across an industry scaling from research-lab curiosities to a multi-hundred-gigawatt-a-year global business.

Why does it work so reliably? No single cause — a bundle of ordinary forces that all compound with scale: factories get bigger and more automated, manufacturers learn to use less silicon per watt, competition among suppliers intensifies, and the cost of financing a now-proven, low-risk technology falls. None of these is a breakthrough on its own. Together, repeated over enough doublings, they add up to a fivefold real price fall in the decade after 2010 — the crash in this module’s hook question.

The same underlying mechanism is now doing the same thing to battery prices (Module 11) and, more slowly, to electrolysers (Module 9). It is arguably the single most important economic fact in this entire course: cheap, mass-produced hardware gets cheaper in a predictable, self-reinforcing way that fuel never does. Coal and gas have no equivalent curve — a tonne of coal in 2030 will cost roughly what a tonne of coal costs today, adjusted for inflation, because you cannot learn your way to cheaper geology.

Who’s who in Indian power

Three distinct businesses sit between a spinning turbine and your fan, and it matters enormously which one is struggling.

Generation companies build and run power plants. NTPC, India’s largest, alone runs roughly 85.5 GW of capacity (December 2025) — about a sixth of the entire country’s installed base. Generators sell their output at regulated tariffs or through the auctions and exchanges below.

Transmission moves bulk power long distance at the voltages Module 10 described. PowerGrid (Power Grid Corporation of India) is the Central Transmission Utility, operating roughly 84% of India’s inter-regional transmission network — nearly 187,000 circuit-kilometres of lines (August 2026). Transmission charges are a comparatively small, stable slice of your bill.

Distribution — the DISCOMs — deliver power the last mile and, critically, collect the money. India has around 72 DISCOMs (FY 2024-25), mostly state-owned, of wildly varying financial health. This is where the chain breaks.

Why DISCOMs are broke

Electricity flows one way down that diagram: generator → transmission → distribution → you. Money is supposed to flow back up it in exactly the same amounts. In much of India, it doesn’t, and the shortfall has to land somewhere.

Two numbers tell the scale of it, as of the end of FY 2024-25: DISCOMs’ sector-wide accumulated losses stood at roughly ₹6.47 lakh crore, with total outstanding debt around ₹7.26 lakh crore — a lakh crore is 1,00,000 crore, or one trillion rupees. (There was, notably, a first sector-wide positive profit of about ₹2,701 crore that year — a genuinely hopeful signal after years of deepening loss, though tiny against the accumulated debt pile.)

Where does the shortfall come from? Two distinct leaks, both real:

Technical and commercial losses. AT&C losses combine power that is physically lost in transmission and distribution (heat in wires and transformers — Module 10’s territory) with power that is delivered but never billed or never paid for (meter tampering, theft, and simple non-payment). India’s national AT&C loss rate stood around 15% for FY 2024-25 — substantially better than a decade ago, but still meaning roughly one unit in every seven or so a DISCOM sends out earns it nothing.

The tariff gap. Even for units that are billed, DISCOMs often charge some customers less than the power costs to supply — which brings us to the mechanism at the heart of Indian electricity pricing.

The cross-subsidy at the centre of it all

Here is the single fact that explains more about Indian electricity than any piece of engineering in this course:

Industrial and commercial consumers are charged significantly more than their actual cost of supply, so that agricultural and (to a lesser extent) residential consumers can be charged significantly less.

This is cross-subsidy, and it is not a side effect — it is deliberate, long-standing policy, rooted in the political reality that farmers and poor households are numerous, organised, and understandably resistant to full-cost power pricing, particularly for irrigation pumping. National policy has, for years, tried to cap this gap at ±20% of average cost — but in the tariffs regulators actually approved for FY 2018-19, more than half of all electricity units nationally were priced outside that band.

Closing it would take a squeeze from both ends, not just one. The badly underpriced units — those billed below 80% of the average cost of supply, about 28% of everything sold that year — would have needed tariffs roughly 30% higher, about ₹1.17 more per unit. The badly overpriced ones, billed above 120% of cost, would come down by about 14%, which is very nearly the same ₹1.17 going the other way. Those figures come from a 2023 study by the Centre for Social and Economic Progress (CSEP), which added up that year’s tariff orders for 60 DISCOMs — over 90% of all electricity sold through utilities in India. For scale: over the five years to FY 2018-19, the average price actually billed nationally rose about 5% a year.

The consequences ripple outward. Large industrial consumers facing tariffs well above cost have a strong incentive to generate their own power or seek captive supply arrangements — precisely the customers whose steady, well-paying demand a DISCOM can least afford to lose. Meanwhile agricultural power, priced far below cost and often metered poorly if at all, both encourages groundwater overuse (irrigation pumps left running because the marginal cost of another hour is near zero) and gives DISCOMs little financial incentive to improve rural service quality.

Two rounds of national reform have tried to fix the structural debt. UDAY (Ujwal DISCOM Assurance Yojana), launched in November 2015, had state governments take over roughly 75% of DISCOM debt — around ₹4.3 lakh crore at the time — in exchange for commitments to operational reform. AT&C losses did fall meaningfully over the following years, but the underlying tariff-gap problem was not solved, and debt began rebuilding. Its successor, the Revamped Distribution Sector Scheme (RDSS), approved in 2021 with an outlay of roughly ₹3.04 lakh crore, set explicit targets — AT&C losses down to 12–15% nationally, and closing the gap between cost and revenue to zero — alongside a huge smart-metering rollout (Module 10) that has run well behind its original schedule.

The pattern across both schemes is instructive: debt relief without fixing the underlying tariff structure buys time, not a cure. Whether RDSS breaks that pattern is still an open question as its own timelines slip.

Electricity traded like a stock

Since 2008, India has had genuine electricity markets. On the Indian Energy Exchange (IEX) — which handles the overwhelming majority of exchange volume — generators and buyers submit anonymous bids for each 15-minute block of the next day, and the system finds the single market clearing price where supply and demand for that block meet, exactly like a stock exchange matching buy and sell orders.

It genuinely moves fast: in FY 2024-25, day-ahead prices swung from a low of about ₹1.99 a unit on an oversupplied August day to a high of about ₹6.35 at another moment of tight supply, with a full-year average around ₹4.28.

But here is the number worth remembering, because it corrects a common misconception: exchange trading made up under 10% of India’s total electricity supply that year. The overwhelming majority still moves through long-term power purchase agreements — contracts locking in a price for 15 to 25 years, giving both generator and DISCOM the predictable revenue that a fast-moving spot market cannot.

🔥Why electricity markets sometimes produce negative pricesAdvanced · optional

In parts of Europe and the United States, wholesale electricity prices genuinely go negative — a generator effectively pays the grid to take its power. Germany saw negative prices in over 5% of all hours in 2024; California’s market saw it in roughly 13% of hours that year, mostly around midday when solar floods the system exactly as Module 7’s duck curve predicted.

The mechanism has two ingredients:

1. Some generation is cheaper to keep running at a loss than to switch off. Nuclear plants and some renewable installations face costly or slow restart processes, or — for subsidised wind and solar in some markets — actually earn a fixed subsidy per unit generated, regardless of the market price. A generator earning a subsidy of, say, $22 per megawatt-hour is still profitable bidding as low as −$21.
2. Demand cannot always absorb the surplus, and storage is limited. When supply outstrips what the grid and its batteries can use, the market-clearing price falls to whatever level makes someone willing to either reduce demand or accept payment to keep absorbing supply — and if enough generators would rather pay than switch off, that price falls below zero.

Does this happen in India? Not yet — and the reason is instructive. Indian regulations set a price floor of ₹0 per unit on the power exchanges; bids cannot go negative by rule. The closest documented approach came on 26 May 2025, when unseasonal rain across northern India cut demand sharply while generation ran near-full: the real-time market price fell all the way to ₹0 per unit for several hours — visibly brushing the floor — but by regulatory design, no lower.

This is a genuinely useful contrast for understanding market design. Negative prices are not an inevitable law of physics; they are what happens in a specific regulatory choice — no price floor, plus subsidies paid per unit generated — combined with high renewable penetration and limited flexibility elsewhere in the system. India has the renewable growth but not (yet) the regulatory conditions. As Indian renewable capacity keeps climbing toward the targets in Module 13, whether that floor holds, and whether regulators choose to remove it, is a live and legitimately open policy question — not a foregone conclusion either way.

🔥🔥The learning-curve mathematics, and why LCOE hides a real assumptionDeep · optional

Formalising the learning curve. If cost per unit C depends on cumulative production Q as a power law:

C(Q) = C₀ · (Q / Q₀)^(−b)

then the learning rate LR — the fractional cost reduction per doubling — relates to the exponent b by:

LR = 1 − 2^(−b)

Solar’s long-run learning rate of ≈20% corresponds to b ≈ 0.32. This is exactly the kind of power-law relationship that produces a straight line on a log-log plot of price against cumulative volume — the classic “experience curve” chart used across manufacturing industries, not invented for solar but arguably nowhere demonstrated more cleanly than in solar.

The formal LCOE calculation, spelled out:

LCOE = [ Σₜ (Capexₜ + Opexₜ + Fuelₜ) / (1+r)ᵗ ] ÷ [ Σₜ Energyₜ / (1+r)ᵗ ]

summed over the plant’s operating life, where r is the discount rate — the rate at which future rupees are worth less than present ones, reflecting both the time value of money and project risk.

Here is the assumption LCOE tables rarely make explicit, and it matters enormously for comparing technologies fairly: the discount rate you assume changes the answer more than almost any engineering parameter. Solar and wind have nearly all their cost upfront (construction) and almost none later (no fuel); coal and gas have costs spread evenly across decades (continuous fuel purchases). Raising the discount rate — treating future costs as less painful — therefore helps fuel-burning technologies relative to capital-heavy ones, because it discounts away more of their future fuel bills. A country or investor facing high borrowing costs (a high discount rate) will calculate a relatively worse LCOE for solar than a country with cheap, patient capital — which is one real, underappreciated reason renewable auction prices vary so much between countries even when hardware costs are identical.

What LCOE does not capture at all. It says nothing about when a plant’s output arrives relative to when it’s needed (Module 7’s duck curve, Module 10’s evening peak), nor about the cost of integrating a variable resource into a stable grid — inertia, balancing, transmission to reach the sunniest or windiest sites. Comparing raw LCOE figures across dispatchable and weather-driven sources is therefore useful but incomplete; serious system planning increasingly uses more complete measures that price in exactly the timing and reliability costs LCOE was never designed to capture.

What comes next

You now understand the economics underneath everything else in this course: why solar kept getting cheaper, why India’s power sector keeps generating debt even as it builds record renewable capacity, and how price signals — from fifteen-minute exchange auctions down to the cross-subsidy on your own bill — shape what actually gets built. Module 13 turns from economics to atmosphere: the greenhouse effect properly explained, honest emissions numbers by source, and what India’s headline climate pledges actually require, year by year.