Why grid-scale storage is bought with cash —
the “can’t take it as collateral” myth, and why ¥6.2 billion over 20 years got signed

Power infrastructure assets worth hundreds of millions to billions of yen change hands for cash rather than debt — that is an ordinary day in the market for grid-scale battery storage. A high-voltage 2MW/8MWh unit carries an indicative initial investment of ¥500–600 million (Nikkei Energy Next). Purchases of a size that would automatically be financed in real estate are, for most buyers in this market, settled out of their own pocket. "Batteries can't be taken as collateral, so you won't get a loan" — an explanation heard almost weekly on the deal floor.

The disclosed record says otherwise. A 40MW/200MWh project in Nishigo Village, Fukushima secured a syndicated loan with a contract amount of ¥6.2 billion and a 20-year term against a total project cost of roughly ¥6.5 billion. A 49MW/231MWh project in Niigata secured Japan's first project bond raised for development capital — ¥10 billion, rated A- by R&I. A portfolio of 14 high-voltage storage sites across the country secured a non-recourse facility with a ¥4.9 billion commitment limit; and even a 99MW full-merchant project, with no contracted revenue at all, secured project finance (PF) repaid solely out of market revenue.

What separates the projects that get debt from the ones that don't is not collateral. To state this article's conclusion up front: collateral is the result; the contract is the cause. And once the contract has determined the shape of the debt, a revenue model still has to go on the credit committee's table. "What is the basis for this bid price?" — this article covers how to turn the argument that stalls right there into one that can actually be tested. Part 1 deals with the vehicle and the contract; Part 2 with the scrutiny of the numbers. In 2026, the market is moving from a stage where cash was the only option to one where buyers can design how cash at the entrance and debt at the exit fit together.

01 — Why cash: the asymmetry of speed (field observation)

Let me start with what I see day to day in originating and broking these deals. This section is field observation rather than published statistics, and I want that clear from the outset.

In sales of storage projects — whether development rights or completed assets — sellers frequently want settlement within about a month. There are three reasons. First, the seller's own cash cycle: grid connection cost contributions and connection deposits have payment deadlines, and since the April 2026 measures against speculative capacity reservations (COLUMN 09), deposit burdens have moved up in both size and timing. Second, there is always more than one candidate buyer. Third, freshness: the closer a project's scheduled connection date, the more it is worth, and elapsed time is straightforward product decay.

Now look at the buyer's clock. A corporate loan needs 1.5 to 2 months for internal approval plus bank credit review; project finance needs six months to a year. The project will not wait. The result is that cash that can decide today beats debt that is cheaper but slower, and owner-run companies and foreign funds that can commit on the spot do not lose out to listed corporates and financial institutions bound by committee. There is corroboration. When Tokyo Century, one of Japan's major leasing companies, announced its own wholly self-developed pipeline of four extra-high-voltage sites totalling 101MW, the strength it cited was the decision-making speed that comes from going it alone (December 2025). Even professional institutional investors treat speed as a competitive resource in this market. (No reliable published statistics on standard credit review timelines could be confirmed; the figures above are a practitioner's sense of it.)

The point to hold on to is this: cash settlement becoming the norm is not proof that debt is unavailable. It isn't that debt won't come — it's that it won't come in time. Whether it comes at all, and on what terms, is what the published record can answer. That is where we go next.

02 — How much of "you can't take it as collateral" is true?

Start with the law. Battery equipment — containerised BESS, PCS, substation gear — is movable property, not real property, so a mortgage cannot be granted over the equipment itself. That does not mean it cannot be taken as collateral. There are three principal routes.

There are worked examples. A crowdfunded lending fund for grid-scale storage took a revolving mortgage over the land and security over the battery equipment, and disclosed a planned loan of ¥495 million against an external appraiser's collateral value of ¥602 million — an advance rate of roughly 80% of appraised value. Collateral appraisal of storage assets is already happening in the market.

So why the refusal at the bank counter? The answer lies in liquidation value. The secondary market for used BESS is immature, and uncertainty remains around how to assess degradation (state of health), whether units can be relocated or repurposed, and whether manufacturer warranties and long-term service agreements can be transferred. Seizing the equipment and putting it to auction offers no clear prospect of recovering the loan — in that sense, "the collateral value is thin" is broadly correct. Seen from the buyer's side, the consequence is an absence of leverage. In real estate you borrow against one asset and compound into the next; storage is, in principle, all-equity, and buying one unit ties up the capital. The investor's complaint that "the yield is good but the capital efficiency is worse than real estate" comes straight out of this collateral structure.

Reduced to a single table, the verdict looks like this.

03 — What lenders actually take: collateral is for taking over, not for selling off

Project finance for generation and storage runs on an all-asset security principle. Security is layered over the borrower SPC and the whole of its assets — the equipment (movables), the land or the right to use it, receivables and contractual positions under project agreements such as EPC, O&M, aggregator and tolling contracts, insurance claims, bank accounts, and the SPC equity itself.

What matters is the purpose of that package. Legal practitioners divide the function of security in two. One is the positive function — the ability to force a step-in: when the business falters, the lender takes over the SPC equity and contractual positions, replaces the sponsor and keeps the business running. The other is the defensive function — preventing assets from leaking out through third-party attachment. Security assignment over the equipment blocks a forced sale by way of a third-party objection, protecting the value of the business as a going concern.

In other words, lenders do not take security in order to break the batteries up and sell them. They take it in order to take the whole business over. That is why an immature secondary market does not directly determine whether PF can be arranged. What determines it is whether the cash flow remains legible after the takeover — which is to say, the contracts. Transferring contractual positions requires the consent of counterparties to the connection agreement, tolling agreement and so on (the transmission and distribution utility, the offtaker), and PF locks this down in advance through pre-agreed assignments and direct agreements. Grid connection rights enter the security package through the same mechanism.

And on 25 May 2026, one premise in this field changed. The enterprise value security interest under the Act on the Promotion of Business-Based Lending came into force (enacted June 2024). It is a new form of security that captures the debtor's total property as a single pool — tangible and intangible assets through to future cash flow — is created through a trust structure, and restricts the use of personal guarantees from owner-managers when granted. Structurally it suits a storage business, which is movables-heavy and derives its value from future revenue, and it could simplify today's practice of stacking all-asset security item by item. No storage deal using it has been publicly confirmed yet, but it is worth watching as a change that could bring down the wall in §02 head-on.

04 — The other wall: revenue a bank cannot read

Alongside collateral, the second structural factor is revenue predictability. FIT solar offered 20 years of legible cash flow at a fixed price for a fixed period, which is precisely why PF was arranged at scale. A grid-scale storage asset earns by combining three markets day by day — JEPX, the balancing market and the capacity market (COLUMN 02) — and most of that income is merchant revenue at the mercy of market conditions. On top of which, the domestic operating record is still shallow. IEEFA's analysis puts Japanese BESS grid connection applications at 170.8GW while only 0.62GW is actually connected (as of 2025). The sample of track record a lender can lean on is simply small.

SMFL Mirai Partners, which arranged Japan's first BESS project financing, has itself reflected that it could not assume FIT-style stable income and had to build both the project appraisal and the contractual terms from scratch. Standard practice in renewable PF is for lenders to size debt off a conservative P90 scenario (the revenue level exceeded with 90% probability) and to require a DSCR of roughly 1.2 to 1.35 times. So who underwrites the P90 of three-market revenue, and how? That is where the granularity of aggregator disclosure becomes the crux — and public data at a level investors and lenders can verify is still limited. The practical answer to this — bracketing your assumptions with market outturn, the policy path and the contracted revenue ratio — is set out in Part 2, from §08.

05 — It got done anyway: five archetypes emerged between 2023 and 2026

Structurally difficult — and yet the disclosed record has built up steadily over three years. Sorted, it comes to five archetypes.

Beyond these, ORIX Bank executed its first non-recourse PF for a solar plant with co-located storage (Usa, Oita) in 2026; and on the equity side, Itochu arranged Japan's first fund dedicated to grid-scale storage (over ¥8bn, 2024, with Fuyo General Lease, Tokyo Century, Bank of Yokohama and others investing), while Fuyo General Lease and seven co-investors took part in an SPC covering six sites totalling about 174MW nationwide (2026).

Read what the archetypes mean. LDA structures attract debt most readily, because the regime secures fixed income for 20 years in principle, and the tenor lines up with the term of that income. Tolling structures convert volatile market revenue into fixed cash flow using the offtaker's credit; the key is the counterparty's standing, and the number of domestic players able to take that on is still small (COLUMN 30). Merchant structures are the hardest, but at Tanagawa the strong sponsor line-up of the Kansai Electric group, and at Nippon Chikudenchi the diversification across 14 sites, appear to have served as the backing for volatile revenue. Put the other way round: whether long-term money reaches a single-site merchant project depends on sponsor credit, operating capability and operating track record — and that boundary started moving in 2026 (§16).

06 — Why ¥6.2 billion over 20 years, and a ¥10 billion bond, got signed

Take apart the two most eloquent deals in that lineage.

Nishigo (LDA). SMBC's release states the reasoning plainly: the project uses the Long-term Decarbonization Auction (LDA), fixed income is guaranteed for 20 years in principle, and stable project revenue is therefore secured over the long run — hence a 20-year loan. The 20-year term maps exactly onto the 20 years of LDA capacity income, so the "matching of tenors" (contract term ≥ loan term) set out in COLUMN 30 has now been demonstrated in a disclosed deal. A contract amount of ¥6.2 billion against a total project cost of about ¥6.5 billion is, on a simple comparison, debt exceeding 90%. (Because recourse, parent guarantees and the equity breakdown are undisclosed, this cannot be read straight across as an LTV. What is certain is that this number and the received wisdom that "storage doesn't get debt" cannot both be true.)

Niigata Yamaya (tolling). Underneath sits a tolling agreement in which Tokyo Gas pays a fixed usage fee for 20 years in return for dispatch rights — and that contract structure is what made both the A- rating from R&I and a ¥10 billion raise from the capital markets, rather than a bank, possible. Note the order of events. Because there was fixed income, a rating agency could verify the revenue; because it could be verified, capital markets money arrived. The contracted revenue ratio can be restated as the share of revenue a third party is willing to underwrite the verification of. Fixed income unlocks not just loans but bonds — and the arranger has signalled that ultra-long-term funding is indispensable for infrastructure in an environment of heightened rate volatility, with the intention of recovering development capital and rolling it into the next project. The "cash at the entrance, debt at the exit" design described in §14 is already being implemented on large deals.

By contrast, Tanagawa and the 14 high-voltage sites — full-merchant deals with no revenue contract — are either modest or undisclosed on amount and tenor. Even among deals that "got done," the depth of fixed income sets the depth of the debt. That is the ranking in the next section, exactly.

07 — The ranking is universal: "contracted revenue ratio" as a shared language

Law firm practitioners describe financeability in storage in terms of revenue contract type. OCCTO-type structures (capacity income, including LDA) are relatively easy to finance because the payer's credit is strong. Tolling suits long-term finance because the owner carries no market price risk, but offtakers with sufficient credit are few. Merchant structures are hard to read on future power prices, and financial institutions are still feeling their way on whether long-term finance is possible at all. That is the ranking.

Distil that ranking into a single number for the conversation with lenders and you get the contracted revenue ratio — the share of total revenue locked in by contract (LDA, capacity market, tolling, floors). In Japan three sources of contracted income make up that ratio. The capacity market: the sixth main auction cleared at a national average of ¥13,303/kW-year after transitional adjustments, a record high, with Tohoku, Tokyo and Kyushu hitting the cap (trends and regional differences in COLUMN 28). The Long-term Decarbonization Auction (LDA): fixed capacity income for 20 years in principle, with the weighted average for decarbonised sources rising from roughly ¥68,000/kW-year in the second round (held in FY2024) to roughly ¥111,000/kW-year in the third (FY2025) — which includes the effect of the raised cap threshold. In that third round the award rate for batteries was 46% (against 1.251GW bid), and with a six-hour-plus duration requirement and cell manufacturing-country restrictions on top, it remains a narrow gate (COLUMN 01 and 03). And tolling agreements (COLUMN 30). Overseas disclosure shows, bluntly, that this ratio is what sets the debt.

In short, "contracted revenue unlocks debt" is a universal principle. And here Japan's peculiarity comes into view. Japan has no public backstop equivalent to Australia's CIS or the US ITC (no battery guarantee has been confirmed from the GX Promotion Organization either — §16), and LDA fills that gap with contracted revenue supplied by regulation. Conversely, the merchant PFs at Tanagawa and the 14 high-voltage sites, arranged without any such backstop, are ambitious even by global standards. As for the floor when nothing is fixed, COLUMN 30 has it: public modelling puts the base-case IRR at −1.5% even including capacity market income (at CAPEX of ¥60,000/kWh). A lender's conservatism is not sentiment; it is DSCR arithmetic.

Interest rates are not irrelevant either. A rising discount rate and cost of debt thin out the present value of merchant revenue and raise the relative value of fixed income (the rate figures are in §16). In a world with interest rates, the contracted revenue ratio matters more, not less.

08 — The numbers that still can't be verified: why it becomes a standoff

That covers the vehicle and the contract. But settling on a contract type does not remove the revenue model from the credit committee's table — and the discussion always stalls in the same place. "What is the basis for this bid price?" "Aren't the aggregator's track record figures rather bullish?" The challenge is fair. Yet ask back what the appropriate level would be, and the other side has no alternative number either. With no fixed price of the FIT kind, there is no "right answer" for future market prices. What follows is how to turn that standoff into a testable discussion. As Part 1 showed, a lender's scepticism is not aimed at your model being sloppy; it is aimed at the structure of an asset class with a small sample of track record (§04). Which is exactly why the correct response is not to argue from instinct, but to present the parts that can be verified.

Look at the structure. What financial institutions verified in solar project finance was, in the end, irradiation. The tariff was fixed for 20 years. Only output varied, and output can be processed statistically from meteorological records as P50 (the median, exceeded with 50% probability) and P90 (the conservative level exceeded with 90% probability). Hence sizing debt at P90 and requiring a DSCR of roughly 1.2 to 1.35 times — the convention took hold (§04).

Storage is the reverse. Volume is controllable to a degree, but price is not fixed. Revenue is stacked across three markets — JEPX, the balancing market and the capacity market (COLUMN 02) — and most of it is merchant revenue at the mercy of price. As far back as 2018, Moody's characterised value stacking, which combines contracted and market revenue, as carrying greater cash flow volatility risk than a purely long-term contracted model. When Australian lenders to battery projects say plainly that they do not size debt on one-off high-revenue events, the logic is the same.

There is no objective foundation for the revenue distribution equivalent to meteorological statistics — so when a developer's assumptions meet a lender's intuition, you get a standoff. The way out is to bracket the assumptions with numbers that can be verified. A storage revenue model has three of them. (1) Past outturn — EPRX cleared prices. (2) A settled future — the price cap reductions the regime has already announced. (3) A contracted present — fixed income from the capacity market, LDA and tolling. Since (3) is precisely the contracted revenue ratio from §07, what follows takes (1) and (2) in turn.

09 — Verifiable number 1, market outturn: give your assumptions a coordinate

Whether an assumption of "bid at X yen, win Y%" is bullish or conservative can be shown not by instinct but by where it sits on the range of actual outturn.

According to trading results published by EPRX (the balancing capacity exchange), average cleared prices by resource type in the first half of FY2025 ran from 9.6 to 13.5 JPY/ΔkW per 30 minutes for batteries in primary reserve, on a monthly basis. That is several times the level at which thermal (2.1–3.5) and pumped storage (1.1–3.4) cleared. Behind it lies a supply shortfall: the fill rate for primary reserve (cleared volume against volume sought) stayed at around 40% online in FY2025, and in the Tokyo and Chubu areas there were summer months in which over 90% of the volume sought went unfilled (regional and product-level detail in COLUMN 16).

Set against that outturn, a common modelling assumption — bidding primary reserve at 9 JPY/ΔkW per 30 minutes and clearing 70 to 90% of it — sits near the bottom of the observed range on price, which makes it conservative to reasonable; and on clearing rate it is reasonable while the shortfall persists.

What matters is writing that positioning into the credit paper. Not "we used actual figures," but "here is where on the observed range we placed the assumption, and why." Make the coordinate of the assumption the subject of debate rather than its rightness or wrongness — get that far and it is no longer a standoff.

10 — Verifiable number 2, the future the regime has announced: the 15 JPY cap and what comes next

Extrapolating past outturn over 20 years draws criticism not because extrapolation is bad in itself, but because it looks as though it ignores declines that have already been decided.

In the balancing market, alongside the move to day-ahead trading in March 2026 (COLUMN 29), the price cap for primary reserve, secondary reserve ① and the combined product was cut from 19.51 to 15 JPY/ΔkW per 30 minutes. During deliberations a flat 7.21 JPY across all products was also floated (October 2025, ANRE's institutional design working group), but concerns about the effect on investment predictability landed the outcome at 15. That said, the regulatory papers expressly state a phased path down to 10 and then 7.21 JPY if competitive conditions do not improve. At the same time, the volume sought for primary and secondary ① was cut from a 3σ equivalent covering about 99% of demand variation to a 1σ equivalent covering about 84%. Both price and volume, in other words, are being tightened by regulation.

The outturn is already moving. The interim review in May 2026 (Working Group on Stable Electricity Supply) confirmed increased bid volumes and an improvement in the primary reserve shortfall, while noting that it "has not been fully resolved" and that "bids near the price cap continue to clear"; it also showed the average cleared price for the combined product falling to 3.15 JPY/ΔkW per 30 minutes in January 2026. On the next cut (15 → 10 JPY), that May 2026 review deferred its judgement on the grounds that one month of post-day-ahead data is not enough to assess operator behaviour, which makes the second-half deliberations this fiscal year the main event. Note that some trade media have begun to report the cut to 10 JPY as a settled matter; what the primary sources confirm as of July 2026 is only the conditional policy that the cap will be "reduced in stages, to 10 and then 7.21 JPY/ΔkW per 30 minutes and so on, if no improvement is seen in competitive conditions in the market" — not a decision with an effective date. Because EPRX publishes any change on its price-cap page at least two weeks before the corresponding delivery date, that page is the practical place for developers and lenders to keep watch.

Feeding this into a revenue model is simple. Assume the 15 JPY cap and the post-1σ environment from year one, then overlay a sensitivity that steps 10 and 7.21 JPY into later years — and attach it yourself, before you are asked. A plan that has already priced in the decline the regime announced changes the level of trust on its own. Even METI's own public modelling uses the fiscal year with the narrowest wholesale price spread and the year with the widest to bracket a downside and an upside. Showing a range is not weakness; it is form.

11 — The five questions lenders actually ask, and how to answer

Here is everything above, reordered into the sequence in which the questions actually arrive.

Q1. "Are that cleared price and award rate consistent with the market?"

Attach EPRX trading results and regional fill-rate data, and show where your assumption sits on the observed range. If you place it above the median, state why (area, product mix, dispatch policy). Aligning definitions — shortfall rate or your own clearing rate — is a precondition (§09).

Q2. "Have you priced in the cap reductions?"

Reflect 15 JPY from year one and overlay the phased reductions to 10 and 7.21 JPY as a scenario. Reflect the narrowing of clearing opportunities from the 1σ volume change in your volume assumptions too (§10).

Q3. "What share of revenue is fixed by contract?"

Give the contracted revenue ratio from the capacity market, LDA and tolling as a single figure. If it falls short of the overseas benchmark of 50%, explain why and add the alternative mitigants (portfolio diversification, third-party verification as in Q5) (§07).

Q4. "Isn't a 20-year extrapolation optimistic?"

Show that DSCR holds its threshold in a downside case that incorporates a declining price scenario and the battery degradation curve (COLUMN 35). The standard in renewable PF is P90 sizing with DSCR of roughly 1.2 to 1.35 times. For full-merchant projects with no contract to lean on, overseas discussion runs a thicker range of 1.2 to 2 times (COLUMN 07, §04).

Q5. "Who verified that extrapolation of operating performance?"

Aggregator disclosure is still not granular enough for lenders to verify — a point widely made in the industry, and precisely why third-party revenue forecasts carry weight. In the UK, revenue curves from independent research houses have been accepted as the basis for debt sizing at several banks, and in Japan in 2026, financing was drawn on several projects (148MW in total) that had been through an independent house's economic analysis. Replacing numbers that cannot be verified with numbers a third party has verified — that is what completes Q1 through Q4.

12 — When the language of electricity doesn't land, translate

A word too on the cases that stumble before credit review even begins: real estate investors.

Open with an explanation of ΔkW and tertiary reserve ① to someone who has invested in hotels and apartment blocks, and the conversation will almost certainly stop. It isn't that they can't understand; they are working from a different dictionary. What is needed is not more precision but translation.

A storage asset is machinery with a statutory useful life of 17 years, and in combining land, equipment, operations and finance it is inherently a familiar shape for a real estate investor. Niigata Yamaya can be described entirely in real estate terms: a 20-year fixed-term lease (the tolling contract) was signed, so a rating could be obtained, so a bond could be issued. Translate the contracted revenue ratio from §07 and it becomes "what share of NOI is locked in by lease agreements." One number, and it speaks to bankers and property investors alike.

13 — Leasing as a separate circuit: lender and buyer at the same time

"If the bank says no, try a lessor" is the usual line, but leasing companies are doing more here than substituting for a loan. On the published record, they are entering the storage market wearing two hats.

First, as lender and investor. Japan's first BESS project financing was arranged not by a megabank but by SMFL Mirai Partners, part of the SMFL group (Himeji — sole PF arrangement plus a 5% equity stake). Ricoh Leasing on the 14 high-voltage sites (a ¥4.9 billion non-recourse commitment), Fuyo General Lease and Tokyo Century investing in Itochu's dedicated fund, and Fuyo General Lease with seven others taking part in an SPC of roughly 174MW all belong to the same lineage. An industry accustomed to movable property, and quicker to decide than a bank, is structurally well suited to battery finance.

Second, as principal and buyer. According to Nikkei, Tokyo Century plans to invest around ¥100 billion by the fiscal year ending March 2030 and Sumitomo Mitsui Finance and Leasing around ¥200 billion by March 2032, and both have maintained that stance since the balancing market price cap was cut. Tokyo Century duly announced in December 2025 an investment in four wholly self-developed extra-high-voltage sites totalling 101MW (citing the decision-making speed of going it alone as its strength), and SMFL was reported in September 2025 to have acquired a large storage asset from West Holdings. Leasing companies are therefore candidate lenders for your project and candidate buyers of completed assets and development rights at the same time — which means sellers have two exits.

Two things to watch when using a lease. First, accounting: the new lease accounting standard becomes mandatory from April 2027, and lessees will in principle bring leases on balance sheet (using a lease for off-balance-sheet purposes therefore needs a different design — see COLUMN 30 §03). Second, tax.

14 — The realistic answer for high-voltage 2MW/8MWh, and three designs for buyers

The large PFs and bonds above are mostly extra-high-voltage stories. What about the high-voltage 2MW/8MWh class most readers are considering, with an indicative initial investment of ¥500–600 million?

Honestly: PF on a single site is structurally disadvantaged. PF carries heavy fixed costs — due diligence, documentation, agency work — which a project of a few hundred million yen cannot absorb into the interest rate. The bias towards cash settlement in standalone high-voltage deals reflects that cost logic as much as the speed described in §01. But there are three circuits. First, the portfolio route: Ricoh Leasing with Nippon Chikudenchi bundled 14 high-voltage sites nationwide under a ¥4.9 billion non-recourse commitment — a replicable template that divides PF fixed costs a single site cannot carry across fourteen. Second, small-ticket finance based on collateral appraisal: the crowdfunding structure in §02 (¥495 million lent against a ¥602 million appraisal). The rate is higher than bank PF, but it works at small scale. Third, lease and instalment finance, and sale to a leasing company: as §13 shows, both the lender side and the buyer side are available.

Recast as buyer strategy, that gives three options.

Option 1: cash settlement, then refinancing. Win the acquisition on the speed of cash, build an operating record once the asset is running, then recover capital through debt, lease or the capital markets and roll it into the next project — capital recycling. The Niigata Yamaya bond is exactly this in practice. The sequence has a second advantage: once an asset is operating, its economics can be shown as measured rather than assumed — the assumption-backing described in §09 can be done with your own operating data. An operating record is the best credit evidence a storage asset has. Separating the entrance (acquisition) from the exit (financing) is, in my view, the most reproducible model available today.

Option 2: lock in a long-term contract first, then raise debt. If you can acquire an LDA-awarded project, or secure tolling or a long-term offtake with a creditworthy counterparty, a Nishigo-style senior facility of 20-year vintage comes into view. This is the classic route for new-build and large investments not under settlement pressure.

Option 3: use lease or instalment finance. Arranging with a leasing company — faster to decide than bank PF, and comfortable with movable property — can be the middle path between acquisition speed and leverage.

One trap common to all three deserves a mention: collateral restrictions on subsidised projects. Under the storage system deployment support programme administered by SII, pledging subsidised property as collateral also counts as "disposal" under the grant regulations, and prior approval from the secretariat is required within the disposal restriction period. The more a project has used subsidies to bring CAPEX down, the less freedom it has to add debt afterwards. Combining subsidies with debt has to be designed from the application stage onwards (the framework for deciding whether to use subsidies at all is in COLUMN 20; see also checklist item 11 in §15).

Cash is a tactic. Capital design is a strategy. The strongest buyer in this market is the one holding both.

15 — Getting a project into financeable shape: a lender's-eye checklist

Debt is not something you apply for; it is something you design. Before a transfer or an arrangement, run through this in the order a lender will.

Items 1 to 4 double, in a sale or transfer of rights, as the conditions for a project that sells well. Getting an asset into financeable shape widens the buyer's funding options — and thereby widens the pool of buyers.

16 — What turned in 2026, and what is still missing

The environment is moving too. Three changes to note, and four things that cannot be confirmed.

What changed. First, the enterprise value security interest came into force (25 May 2026; §03). Second, the GX Promotion Organization — operational since July 2024, with a ¥1 trillion guarantee capacity intended to cover risk that private lenders cannot fully absorb. No guarantee for an individual grid-scale storage project has been publicly confirmed as at the time of writing, but the vehicle exists. A first such deal would move the bankability of merchant structures up a notch. Third, interest rates. The Bank of Japan raised its policy rate to 1.0% in June 2026 (a level not seen for about 31 years), and the 10-year JGB yield touched 2.88% on 9 July 2026 (its highest since September 1996), trading in the 2.7% range later that month. A higher cost of debt squeezes IRR on merchant projects and raises the relative advantage of fixed-income structures such as LDA and tolling. At the same time, the opportunity cost of parking equity has risen for cash buyers too — in a world with interest rates, the logic of cash-only is no longer self-evident.

What accumulated. The case record keeps growing. In March 2026, the acquisition of an operating merchant storage asset (Helios, Sapporo, 50MW/104MWh, operating since November 2025) was funded by what was announced as Japan's first green project bond — roughly ¥5.4 billion with a tenor of up to 19 years (arranged by Nomura Securities, rated BBB by R&I, using a trust beneficiary interest structure). Nineteen years of capital markets money reached a single-site merchant project on the strength of its operating record — the merchant implementation of "cash at the entrance, refinancing at the exit" from §14. In May the same year came a full-merchant non-recourse PF for Hiji, Oita, 51MW/204MWh (arranged by Aozora Bank, with silent partnership equity from four domestic companies including Itochu Enex, operations expected around September 2028). Disclosed full-merchant PFs are stacking up behind Tanagawa and the 14 high-voltage sites, and the roster of lenders has widened from megabanks and lessors to securities houses, trust banks and newer entrants.

What is still missing. (1) Published benchmarks for DSCR, LTV and pricing on domestic BESS project finance (Nishigo's ¥6.2bn against ¥6.5bn is an exceptional case where a ratio can be read, but recourse is undisclosed, so it cannot be stated as an LTV). (2) An executed GX Promotion Organization guarantee for a BESS project. (3) Disclosed instances of sale-and-leaseback or Japanese operating leases (JOL) on storage assets. (4) A secondary market for used BESS — the trading market that would underpin liquidation value remains immature. Put positively: as secondary circulation and SOH assessment standardise, the treatment of batteries as collateral can improve. What is unknown is left as unknown, recorded as a factor for judgement.

Conclusion — Collateral is the result; the contract is the cause. And the discipline of the numbers gets debt through

Storage assets are bought with cash not because buyers are ignorant, nor because sellers rush them unfairly. It is the market's rational adaptation to three conditions: the asymmetry of speed, a collateral structure built on movable property, and volatile revenue with a shallow track record.

But the sentence "you can't take it as collateral, so you can't get a loan" is half wrong as a matter of law and entirely wrong as a matter of causation. Batteries can be taken as collateral. What cannot be taken is a reliable prospect of recovering the loan by selling that collateral. So lenders assemble a security package that lets them take over the whole business, and decide whether to lend on whether the cash flow of that business is legible — under which contract, for how many years, at what fixed amount. ¥6.2 billion over 20 years went to Nishigo not because the collateral was impressive but because 20 years of LDA fixed income made the repayment schedule work. A ¥10 billion bond went to Niigata Yamaya because Tokyo Gas's 20-year tolling agreement made an A- rating possible.

At the same time, you cannot eliminate the unverifiable numbers from a revenue model. This is not FIT, so that is normal. What you can do is bracket the assumptions with the three verifiable numbers — market outturn, the policy path, contracted income — apply sensitivities and DSCR stress to the residue that cannot be verified, and add third-party verification where possible. An honest revenue model is not one that predicts the future correctly; it is one that also shows what happens when it is wrong. What financial institutions are ultimately looking at, in my view, is less the numbers themselves than how you deal with them.

In under three years since the first domestic PF in August 2023, five archetypes have appeared, leasing companies have entered in earnest, the enterprise value security interest has come into force, and a world with interest rates has returned. Contracts call debt in, and the discipline of the numbers gets it through. Buyers secure the competitive edge of instant cash at the entrance and design refinancing at the exit. Sellers get an asset into financeable shape before selling it. As of 2026, that is the order in which equity works hardest in this market.