The White Rabbit in the Elevator, Part 2

The row that cannot be handed to anyone

The first half of this piece began in an elevator, with a rabbit advertising electricity to a household that had already left for a mortgage discount, and ended in TEPCO Holdings’ own segment table — three of the group’s four businesses sorted into instruments a financier would recognize: generation as an equity line, the wires as bonds, retail as an option. The 2020 legal separation reached the legal entity and stopped. What the sorting could not place was the holding company’s own row: revenues of dividends and contribution income, a nuclear capacity utilization rate printed as a dash. This half reads that row.

The holding company’s own row

That row explains why the rest does not come apart. Its revenues are dividends, decommissioning contribution income, and support fees; its costs are repairs and depreciation on nuclear plant, and contributions to the state support corporation. Plant that generates nothing still incurs repairs, depreciation, and levies.

The volatility is on a different scale from anything else in the table. In the first quarter, a subcommittee’s guidance on how preparatory work for fuel debris retrieval should proceed led to ¥903 billion of additional cost; over the nine months the equity ratio fell from 25.1 to 20.6 percent. A committee described a method, and several years of ordinary income disappeared.

The natural objection is restart, and restart has arrived: Unit 6 at Kashiwazaki-Kariwa has been in commercial operation since April, the first reactor the company has run since the accident — the dash belongs to the nine months to December, not to the present. But a gate stands ahead of the running plant. The specified severe-accident facilities — backup structures required against deliberate aircraft impact — carry statutory installation deadlines, and a plant without them may not run. TEPCO’s own chart shows Unit 6 with a deadline of September 2029 and completion projected two years later, the gap labelled, in the company’s words, the period in which operation is not permitted. The plant that finally runs is already scheduled to stop again. The availability of this asset into the early 2030s is being decided by a construction programme.

Then the number that makes the structure legible. Under the fifth comprehensive special business plan, approved in January 2026, the funds required for Fukushima total ¥23.4 trillion, most of it fronted by the state and recovered afterwards. The group’s own share is stated as approximately ¥500 billion a year. The deck also discloses what has actually been produced, year by year, for eight years: the annual total has cleared ¥500 billion twice.

The discretionary piece underneath is the special contribution, which TEPCO alone pays and which the support corporation notifies each year. Zero one year, ¥230 billion the next. That is not a repayment schedule; it is a figure set annually against what the operator can carry — and the deck confirms the character of it in the smallest print available. In the full-year earnings forecast, the special contribution appears as a footnote: ¥50 billion, provisionally placed.

Read the direction carefully. The damage is fixed, published, and measured in tens of trillions. What flexes is the rate at which it is collected, and it flexes to whatever leaves the operator standing.

One more fact belongs in this row, because the word “support” understates it. The support corporation has held a majority of the holding company’s voting rights since 2012. The state is not standing behind this balance sheet; it has been the controlling shareholder on top of it for over a decade. The entity that sets the annual figure and the entity that owns the debtor are the same entity.

TEPCO publishes a diagram of its return-on-invested-capital management. The chain from investment to corporate value passes through a box labelled Fukushima responsibility, with compensation and decommissioning costs secured at ¥500 billion a year as a named driver.

No other company’s ROIC tree has that box.

Read the retail row again with it in view. Energy Partner is not retained because retail is a good business; it is retained because an obligation that generates no cash of its own needs every leg that does. And the same obligation reaches the regulated leg by a separate route: the provisioning that should have been accumulated before 2011 and was not is being recovered over roughly forty years inside the wheeling charge, nationwide, from households regardless of which retailer they chose — put there by ministerial ordinance.

Vertical integration here is not held together by synergy. It is held together by a debt.

The question that was asked, twice

The vocabulary that sorted the first three businesses has a fourth entry, and it is the instrument no private book issues or holds: an obligation of sovereign scale, no maturity, no schedule — serviced the way restructured sovereign debt is serviced: at capacity, year by year.

Three of the four have natural investors. The fourth has exactly one natural holder, and it is not a company.

That is the portfolio question. What Japan debated at length was neutrality — the question every unbundling asks. But the portfolio question was asked too. Twice. In the years just after the accident, as the argument over legal insolvency: put the company through bankruptcy and cut the obligation loose from the businesses. The answer was no. And again in 2016, when a government committee on TEPCO reform put restructuring on the table — reorganizing the network and nuclear businesses jointly with other operators. Read what the reorganization was for. Not so that each business could stand on its own terms, but so that the earnings of the reorganized businesses could fund Fukushima. The question was asked at cabinet level, and the answer reversed its premise: the businesses belong wherever they produce the most cash for the obligation. Separation was considered — as a funding device.

Europe answered with the opposite grammar. E.ON’s spin-off of Uniper did not put the nuclear plant into Uniper — that leg had no buyer. In 2017 the German operators paid €24.1 billion into a state fund and were released from responsibility for storing the waste.

They bought an inheritor, once, in cash.

The two settlements are inversions of each other. Germany priced the obligation, moved it to the state, and left the equity private. Japan moved the equity to the state and left the obligation with the company, unpriced, collected at whatever rate the year permits.

Uniper was nationalized five years later — for reasons that had nothing to do with nuclear; Russian gas did it — so separation plainly does not make sovereign exposure disappear. What it decides is whether the other three businesses are still standing underneath it.

One observation would tell me I have read this wrong. If the special contribution stops being a discretionary annual figure and becomes a fixed one — set by a published rule rather than by what the year allows — then the obligation has a price, and a priced obligation can be transferred. A fixed figure is a coupon, and an instrument with a coupon can be held by someone other than a sovereign. What would no longer hold is the explanation this piece runs on: that the arrangement is stable because the debt bends to fit the business.

Where it stops

The separation reached the legal entity, and stopped. Not for lack of analysis — everyone involved can read a segment table, and the table is published by the group itself. It stopped at the question of who inherits an obligation measured in trillions and in decades, because every available answer is one no participant will say out loud: the network company, and its tariff; an independent generator no one would fund; the state, explicitly rather than through a support corporation.

What has changed since 2020 is not the argument. It is that the arithmetic has begun moving without one — the wires into their first demand growth in fifteen years, the equity down, the debt up — while the obligation’s largest discretionary component sits in the forecast as a placeholder.

And one relationship in the story is not a choice, which is where it reaches back to the corridor. I changed retailer in an afternoon of paperwork. The wires I cannot change — nobody can; the leg with no churn rate is the leg every household in the region hangs from, whatever name is on the bill. Nothing in the disclosure says that leg is short of money. But its decade of investment is funded through the balance sheet that lost four and a half points of equity for reasons located in none of the operating businesses, and the room for that spending is set, in part, by the same annual figure that flexes to what the group can carry. That is not a prediction of outages. It is an observation about address: if reliability were ever to be rationed, the rationing would be decided on that balance sheet — and the households deciding nothing would be the ones with no exit.

A rabbit advertises electricity to a household that is grateful for it all summer and that left for a mortgage discount. A chain of reasoning that was true until the thirty-first of March runs its full length through people nobody has asked to shorten it. And underneath both, a debt that no one can hand to anyone keeps every leg attached to it, one levy and one cash flow at a time — at a rate set, each year, by how much the legs can still carry.