Power We Sell Cheap and Buy Dear

The full numbers

Bhutan’s hydropower fleet is mostly run-of-river, so dry-season (December–April) output collapses to 25–35% of installed capacity. Since 2022, peak demand has outgrown dry-season generation (peak demand rose from 375 MW in 2020 to 1,477 MW in November 2025 — see paradox #5). The country now imports from the Indian Energy Exchange (IEX) every winter to cover the gap: 1,406 GWh in 2024; 1,102 GWh in 2025 (per [BPC Power Data Book 2024][^bpc-power-data-book-2024]).

Lean-season IEX clearing prices typically run INR 4–6/kWh. BPC’s posted retail tariff to HV1 industrial customers — the same wires, the same hours, often the same physical electrons — is Nu 1.60/kWh (ERA tariff order, 1 July 2025, continuing the September 2022 structure). The arithmetic:

This is the cash-out version of the implicit subsidy quantified in #51 and #62. It is not opportunity cost. It is real money out the door: BPC pays Indian discoms in INR for power purchased at scarcity prices, and recovers 25–32% of that cash from the industrial customers who consume it. The shortfall lands on BPC’s P&L (compressing the dividend to DHI) or is socialised across the LV ratepayer base via the tariff structure.

The 2025–28 BPC proposed tariff revision (LV +115%, MV +228%, HV +75%) is partly an attempt to plug this hole. Even at the proposed HV rate of Nu 2.80/kWh, the lean-season import gap (Nu 5.00 vs Nu 2.80) is still ~Nu 2.20/kWh = $0.026/kWh = **$28M/year loss** on the same 1,102 GWh — compressed but not eliminated.

Imagine this

It is 6 AM on 14 January 2026 in Thimphu. The peak-load registers 1,420 MW. Domestic hydropower output, after losses, is delivering roughly 1,000 MW. The gap — 420 MW for the morning peak — flows in via the Birpara–Malbase tie line from the Eastern Regional grid in India. The Day-Ahead Market clearing price for that block was INR 5.20/kWh. BPC pays the bill in rupees.

Across town, a ferro-silicon furnace in Pasakha is consuming 38 MW continuously. Of those 38 MW in this hour, roughly 11 MW are physically the imported Indian electrons — the rest is domestic hydro running into the same wires. The furnace’s electricity meter ticks over at the HV1 industrial tariff: Nu 1.60/kWh. BPC bills the factory Nu 60.80 for that hour. The same 11 MW that came from across the border cost BPC roughly Nu 197.60 to bring in.

Nobody in this transaction is doing anything wrong. The PPA volumes are locked. The IEX clearing price is set by the All-India market. The HV1 tariff is set by ERA under the Electricity Act. The furnace pays its bill on time. But the system, taken as a whole, loses Nu 137 on this single hour of supply to this single furnace — and this hour repeats across thousands of customers, across 90 winter days, every year. The bill comes in. The bill is paid. The loss is structural.

Where this came from

Three things compounded over a decade:

1. The HV1 tariff was set when the country was a net exporter year-round. The Nu 1.60/kWh rate (originally established in 2022, reinstated 1 July 2025) was calibrated against a generation cost structure that assumed surplus generation in every month. Once domestic demand started forcing lean-season imports (around 2022), the tariff stopped reflecting the actual delivered cost of supply.
2. The import volume grew faster than the political appetite to revise. Lean-season imports ran ~200 GWh/year in 2018; by 2024 they were 1,406 GWh. The structural deficit became permanent before the tariff structure caught up.
3. The HV1 customer base is small, organised, and strategically protected. Twenty-three customers (the sovereign BTC miner, the BitDeer–DHI joint venture, large cement and ferro-alloy exporters) consume ~88% of domestic electricity. They are too economically and politically important to absorb a tariff increase that would price them out of competitiveness — and they know it. The 2026 proposal raises HV1 by 75% (still cheap globally) precisely because anything sharper would invite negotiated side-letters or operational scale-back.

Why this matters now

The lean-season deficit is projected to widen to 600–800 MW by 2026 and 900+ MW by 2027 (see paradox #5). At unchanged tariffs, the cash loss on imported-resold MWh could reach USD 60–80 million per year by 2028 — comparable in size to the entire annual GoI grant transfer. This is not a forecasting exercise; the physical reality of the storage gap, plus the existing PPA volume commitments, plus the existing HV1 tariff, mechanically produces this loss every winter for as long as none of those three move.

It is the closest thing in the Bhutanese energy system to a paradox-grade negative-margin sale: a state utility paying market price to import and recovering one-third of that cost on resale, with the gap absorbed silently across the BPC balance sheet and the LV ratepayer base.

What it should be

• HV1 retail tariff dynamically indexed to lean-season IEX clearing prices — at minimum during Dec–Mar (a “winter tariff” surcharge structure exists in Norway, Quebec, and most US utilities)
• A pumped-storage commissioning timeline that closes the lean-season gap before 2032 (Gongri-Jericho 1,800 MW currently in DPR — see paradox #5)
• Demand-side management contracts with HV1 customers (curtailment payments for accepting reduced supply Dec–Mar in exchange for preserved Nu 1.60/kWh rate in May–Oct)
• Public reporting of BPC’s cost-of-supply by season alongside the retail tariff, so the cross-subsidy is legible to policymakers and ratepayers

How others do it

• Norway — Nord Pool day-ahead pricing flows through to industrial customers; LV customers face progressive tariff with seasonal variation; no static below-cost retail tariff
• Quebec (Hydro-Québec) — large industrial customers face market-indexed pricing for incremental consumption; periodic moratoriums on new BTC connections precisely to prevent this dynamic
• Texas (ERCOT) — wholesale-passthrough is the norm; large industrial customers absorb spot-price volatility directly
• California (PG&E, SCE) — winter/summer tariff differentials embedded in retail; industrial customers can opt for time-of-use or market-indexed
• Bhutan: flat Nu 1.60/kWh HV1 tariff year-round, with BPC absorbing the import-price differential as a structural loss

The question we should be sitting with

If a private utility paid Nu 5/kWh to procure power and resold it at Nu 1.60/kWh to its largest customers, the auditors would call it a structural impairment and the board would replace management. When a state-owned utility does the same thing, we call it the cost of policy. What is the cost of pretending the loss isn’t there?