Japan's Storage Assets Have No "PPA" —
Which Contract, How Many Years, How Much You Can Lock In

In meetings with investors, domestic and overseas alike, one phrase comes up almost every time: "long-term PPA." The intent — to own a storage asset on a stream of long-term fixed income — is exactly right. Just one word is off. Because a battery does not generate net electricity, the kind of PPA that sells a fixed volume of energy (kWh) at a fixed price, the way solar or wind does, does not map onto a storage asset.

And the return without any fixing has already been put on paper, in a public estimate. In the sensitivity analysis the Mitsubishi Research Institute (MRI) submitted to a Ministry of Economy, Trade and Industry (METI) working group, even with capacity-market income added, the base case at a CAPEX of ¥60,000/kWh returns an IRR of −1.5%. Merchant operation alone does not turn a profit, even on the public assumptions. That is precisely why the question "which contract fixes the revenue" decides whether the project works.

01 — A "PPA" and a storage contract are different things

First, let us align the vocabulary. A generation PPA is a contract to buy and sell energy (kWh) at a fixed price. A virtual PPA goes further: no electrons move at all, only the environmental value (non-fossil certificates) is traded. The virtual PPA signed in June 2026 between ENEOS Renewable Energy and NSK in Kyushu is the textbook case — environmental value alone, supplied for about 15 years from a roughly 54 MW solar plant fitted with about 130 MWh of batteries. Here the battery's job is to firm the solar output; it is not a contract that fixes the revenue of a storage asset itself.

What a long-term storage contract fixes is not energy, but capacity and dispatch rights. Who carries the market-price risk — that is the single axis on which the true nature of these contracts can be read.

In Japan, "offtake contract" and "tolling contract" are used almost interchangeably in practice (they are not perfectly synonymous as a matter of regulatory definition). The scheme Tokyo Gas calls an "offtake contract" is the same shape as a thermal-plant toll: the SPC that owns the battery hands over dispatch rights, receives a fixed fee, and the market-price risk is taken off its hands. We treat the two as synonymous here.

02 — Five options on one map

With the vocabulary aligned, plotting the options on two axes — the firmness of the revenue (predictability) and the room for equity upside — brings out the differences in character. "PPA," a word from the generation world, does not sit on this map at all.

The map shows that predictability and upside are not necessarily inversely related. Because tolling delivers certain revenue, it can carry thick debt, so even a project IRR in the 6% range produces an equity IRR in the 18% range (top right). The LDA, by contrast, has the strongest predictability of all, yet by design it gives up the upside (bottom right, more below). Merchant has the widest room to swing higher, but it swings far enough that even a public estimate can be in the red (top left).

Before going further, it is worth orienting the map against the markets a Western investor already knows. Japan runs the same toolkit — minus the standalone-storage PPA, and minus a federal investment tax credit. The table below maps each Japanese contract onto its closest equivalent in the United States, the United Kingdom and the EU.

Mapping by ScienceX for orientation. The equivalents are approximate; market rules, tenor and counterparties differ. Overseas figures are proxies and cannot be applied directly to Japan.

Read against that mirror, the same five Japanese options line up cleanly on contract, tenor, price, return and domestic track record:

¥ conversions are illustrative at £1 ≈ ¥200, $1 ≈ ¥155, €1 ≈ ¥165, not actual contract values. Domestic toll / offtake / floor unit prices are undisclosed between the parties. The 47 MW equity IRR is a stated-assumption generic model.

03 — What the contracts are: four types and the clauses that matter

A long-term storage contract sorts into four types by who carries the market-price risk. In every case the asset is, as a rule, owned by the owner side (the SPC), and the dispatch rights pass to the offtaker.

What the contract nails down is the same set of clauses an investor would recognise from London, Houston or Frankfurt: availability guarantees (fee reductions on a shortfall), capacity guarantees (including degradation), whether the fee is CPI-linked, responsibility for charging energy, dispatch rights and cycle limits, and credit support (parent guarantee, letter of credit, rating triggers). Because the standard contracts are confidential in Japan, the concrete levels have to be inferred from markets where disclosure runs further — and there, the United States, the United Kingdom and Germany each tell a slightly different version of the same story.

The United States: storage PPAs are usually tolls, plus RA and synthetic floors

In the US, a storage PPA is most often structured as a tolling arrangement: the utility supplies the charging energy, because developers do not want to take input-price risk and the utility is better placed to bear it. In California, load-serving entities carry capacity procurement obligations under the Resource Adequacy (RA) program, and RA contract prices have risen year after year; with an RA contract in place, the average battery could still have earned on the order of $178/kW·yr in total wholesale revenue across 2025. In ERCOT, which has no capacity market, revenue is far more volatile, and tolling is spreading as a de-risking tool — five operating projects under known tolls, with seven more expected online by the end of 2026. Where owners want certainty without a full toll, optimisers such as GridBeyond now offer revenue-floor and synthetic (virtual) toll contracts in ERCOT and CAISO, underwritten by investment-grade partners — the same downside-protection logic as the UK floors and Japan's Bison deal.

Germany: pure tolls, disclosed in euros

German tolls run roughly €110,000–150,000/MW·yr on tenors of 5–7 years (10 at most), with signed examples including Stendal (104.5 MW, 7 years, contract value €85–95m), Vattenfall, and RWE (50 MW, 5 years). These are the clearest disclosed unit prices in the European market.

04 — How many years: domestic offtake is 20 years as standard

On tenor, every domestic offtake we could confirm was 20 years. Hirohara, Ishikari, Ashiya, and the HDRE (HD Renewable Energy) deal added in April 2026 — all about 20 years, in line with the 20-year tenors of the capacity market and the LDA. Shorter domestic offtakes of 10–15 years cannot, at present, be confirmed in public disclosure. Only the floor contract is short, at 10 years, and that is a different form from offtake. Overseas tolling ranges from 5 to 15 years, and the rule of thumb for bankability is roughly 10 years or more with an investment-grade offtaker. Because tenor sets the loan tenor, the floor's 10 years is a constraint in PF design.

05 — How much you make: the floor without fixing, the overseas levels, the LDA ceiling

This is the core. In order: the case without fixing, the overseas fixed levels, the LDA's effective return, and then the same 47 MW model.

The floor without fixing. On the public estimate, the base-case IRR is −1.5%. Turning a profit means either cutting CAPEX below ¥50,000/kWh or wholesale spreads swinging to the upside (on an upside assumption with CAPEX ¥30,000/kWh, IRR is about 14%). Merchant carries the dream of the upside, but on the public assumptions its floor is in the red.

The fixed level is undisclosed in Japan. Tokyo Gas, PowerX, Bison and MIRARTH alike record only "20 years (or 10), fixed amount"; not one publishes a figure that can be normalised to ¥/kW. That is not a gap in the record — it is a finding about the state of the market. The reference level is built from markets where disclosure runs further. In the UK, a roughly 2-year toll is on the order of £57k/MW·yr (≈ ¥11.4m/MW·yr); a 10-year-class floor is £44–52k/MW·yr (≈ ¥8.8–10.4m/MW·yr). Named UK deals add texture: Gresham House × Octopus (≈£57k/MW·yr, including about £10k of capacity market), the Gresham House floor (≈£52k/MW·yr, 7 years and over, performance-conditioned), Eku × SmartestEnergy at Ocker Hill (99 MW / 198 MWh, 10 years — the UK's first debt-backed toll), and Drax's West Burton C (250 MW / 500 MWh, 10 years, CPI-linked).

The LDA is the one contract whose sell-side price is publicly fixed. Weighted-average clearing prices rose across rounds: 5.8, 6.8 and 11.1 (¥10k/kW·yr) for the first, second and third (figures for decarbonised sources as a whole; storage-specific clearings are permanently undisclosed). But the LDA caps the return that can be built into a bid at a pre-tax WACC of 5% (±1%), and then claws back about 90% of any income earned in other markets after the fact. Between these two constraints, the winner's effective return is anchored by the scheme at roughly 5%. Predictability is the strongest of all; the upside is shut off, by design, from the start. From the third round, continuous discharge of six hours or more became a requirement, and the lithium-ion allocation was tightened from 1 GW to 0.4 GW.

Domestic toll fees are undisclosed between the parties. The 47 MW equity IRR is a stated-assumption generic model, not an actual deal value. This 18.6% is on a hybrid assumption where capacity-market income (about 40% of revenue) is held by the owner; on a pure toll that hands the capacity market to the offtaker as well, the same toll fee drops equity IRR to single digits. To reach 18% on a pure toll alone, the toll has to rise to about 2.6× the overseas benchmark (£57k/MW·yr ≈ ¥950/kW·month), or CAPEX has to fall by about half (toward Chinese-cell levels) — see §§06–07.

06 — Running the same 47 MW under four contracts

From here, the contents of that 18.6%. Taking an extra-high-voltage 47 MW / 188 MWh (4-hour) asset as a generic example, we run the same hardware under four contracts — ① balancing-led, ② capacity + wholesale, ③ LDA (6-hour), ④ tolling + capacity market — and line up project IRR, equity IRR, DSCR and payback period, with the assumptions stated.

The remaining assumptions are as follows. Effective tax rate about 31.5% (after size-based enterprise tax and the special defence corporate tax); useful life 17 years; annual degradation about 1% (allowing roughly 4 hours at commissioning to fall to 3 hours at contract end); round-trip efficiency 85%, DoD 90%. On financing, ④ tolling and ③ LDA, with high certainty suited to long-term debt, take about 75% debt at 3.0% over 18 years, while ① and ② merchant, with variable revenue that does not support much debt, take about 40% at 3.5% over 10 years. Market assumptions (base): balancing ΔkW at ¥4/ΔkW·30min (50% clearing), capacity market about ¥11k/kW·yr, JEPX day spread ¥20/kWh, one cycle a day, LDA fixed at ¥30k/kW·yr (about 90% of other-market income clawed back). Only ③ LDA is sized at 6 hours / 282 MWh / CAPEX about ¥15bn. And ④ tolling is a "toll + capacity market" hybrid in which the owner holds the capacity-market income — about 40% of ④'s total revenue. The return on a pure toll that hands the capacity market to the offtaker as well is shown in §07.

The table below is the central result, setting out investment metrics for the same 47 MW hardware run under four contracts.

④ tolling is on a "toll fee + capacity-market income (held by the owner)" hybrid assumption. On a pure toll that hands the capacity market to the offtaker as well, the same toll fee drops equity IRR to single digits (see §07).

Take equity IRR alone and ④ tolling (18.6%) and ① balancing (17.6%) run close. But their insides are opposite. The figure below sets each contract's project IRR (the business's earning power) against its equity IRR (the shareholder's return).

The contract with the lowest project IRR has the highest equity IRR

The flip side: this inversion does not hold unconditionally. If the certainty breaks, the thick debt cannot be raised and the equity IRR sinks. There are three conditions for it to hold.

Reading the four contracts

④ tolling (toll + capacity market) is the lead. A toll fee of ¥1,500/kW·month (¥18k/kW·yr) plus capacity-market income (held by the owner) gives an equity IRR of 18.6% and a minimum DSCR of 1.31x. No market operation is needed, and because the offtaker carries the price risk it is the most stable. The premise is the offtaker's credit. But on a pure toll that hands the capacity market to the offtaker as well, the same ¥1,500/kW raises only a little over 40% in debt, and equity IRR falls to single digits (see §07).

② capacity + wholesale is the upside. You can reach for more through wholesale arbitrage and the capacity market. It is favoured in areas with high JEPX spreads or capacity prices, but because revenue is variable, debt stays around 40%.

① balancing-led is high-return, high-volatility. Project IRR is the highest at 13.3% and payback the fastest at 7 years, but the March 2026 balancing-market reform (price cap ¥15, volume tightened to 1σ) can pull the assumptions apart. The headline IRR and its repeatability are different things.

③ LDA is condition-dependent and the weakest. Six hours / 282 MWh is mandatory, so CAPEX is about 1.5× (¥15bn). It works if you set the fixed income at ¥30k/kW·yr, but at realistic clearing levels (a view in the mid-¥20k thousands) the DSCR falls below 1 and it cannot be structured. The viability depends on a bet on the clearing price. It works only if the equipment cost can be halved, or the clearing comes in high.

07 — The toll fee decides most of the equity (a pure toll is single-digit)

④ tolling's return moves almost in proportion to the toll fee. A difference of a few hundred yen a month swings the equity IRR a long way. First, the case of a hybrid in which the owner holds the capacity-market income.

The table above is on a hybrid assumption with capacity-market income held by the owner. Capacity-market income is about 40% of total revenue. The per-kW unit price is undisclosed between the parties.

So what about a "pure toll" that hands the capacity-market income to the offtaker as well? Because revenue is then the toll fee alone, the same ¥1,500/kW raises only a little over 40% in debt (the DSCR bites, so it cannot be stretched to 75%), and equity IRR sinks to roughly 0%. At the overseas pure-toll benchmark (£57k/MW·yr ≈ ¥950/kW·month) it is negative.

To reach an equity IRR of 18.6% on a pure toll (no capacity market), the toll has to rise to about ¥2,500/kW·month (≈ ¥30k/kW·yr ≈ 2.6× the overseas benchmark), or — holding the toll at ¥1,500/kW — CAPEX has to fall to about ¥5.2bn (≈ ¥28k/kWh, Chinese-cell level, about half of current). At Japanese CAPEX (¥10bn, ≈ ¥53k/kWh), a realistic pure toll does not produce a double-digit IRR. Overseas investors can aim for a high shareholder return in Japan when their procurement can cut CAPEX. The disclosed pure-toll benchmark is roughly ¥10–15k/kW·yr (overseas); the per-kW unit price is undisclosed between the parties.

08 — Who is contracting, and over how many years

To fix, you need a counterparty that can fix. For grid-scale storage, the one party publishing 20-year, dispatch-rights-transfer, fixed-fee offtake at scale and across multiple deals, as the offtaker (the party taking dispatch rights), is at present clearly Tokyo Gas. On top of Hirohara 30 MW, Ishikari 30 MW and Ashiya 50 MW, April 2026 added about 190 MW with HDRE (HD Renewable Energy).

Sources: Tokyo Gas (2024-04-24 / 2025-06-30), HDRE (2026-04-16), MIRARTH (2025-09-08), Kyodo PR Wire (Bison, 2025-07-07). Undisclosed items are marked "undisclosed."

Look to other forms and the monopoly is starting to crack. MIRARTH and PowerX are tolling; Bison and Englehart CTP are Japan's first 10-year floor. The picture of "the only seller of fixed revenue = Tokyo Gas" is no longer accurate. The base of providers is also thickening, in the form of equity and operations.

Sources: each company's timely disclosure / IR and reporting (2024–2026). Saibu Gas × Bank of Fukuoka is at the agreed-in-principle stage; the SPC and contract are unannounced.

The financing precedents are now all in place. All arranged by MUFG Bank: (1) Hirohara, a fixed-revenue type built on offtake as the credit base (Japan's first storage PF); (2) Ishikari, an offtake type (about ¥5bn); and (3) Tanagawa, a full-merchant non-recourse deal repaid from market revenue alone (a Japanese first). The convention is to credit-enhance with fixed income; full merchant presumes operating expertise on the order of Kansai Electric Power.

09 — CAPEX, and the rules that tailwind fixing

What squeezes the return most is CAPEX. A domestic grid-scale turnkey is estimated at about ¥68,000/kWh (about $440). BloombergNEF's global turnkey average (including system installation) is $117/kWh, and Chinese-made is $73/kWh (about ¥11,000/kWh) — a gap of more than 3× between Japan and abroad. The LDA's third round made six hours or more of continuous discharge a requirement, tightened the lithium-ion allocation from 1 GW to 0.4 GW, and imposed a "30% per-country cell-manufacture cap." The scheme brought the trade-off between reliance on cheap Chinese product and security into the design.

The rules tailwind fixed contracts in one more way. The balancing market's price cap was cut from ¥19.51 to ¥15/ΔkW·30min (in force), and if competition does not improve, a staged reduction to ¥10 and ¥7.21 is under study. The procurement volume was tightened too. The more the upside of merchant operation is trimmed, the more the relative value of fixed income (tolling, floor, LDA) rises.

10 — Three gates specific to the foreign investor

The gates that stand only in front of overseas capital, in brief. Each can be built into the plan if grasped in advance.

JC-STAR (security certification) — the battery itself is out of scope. What is certified is the communications and control layer (IP communications equipment such as PCS, EMS and gateways); the battery cells and packs themselves are out of scope. Adopting foreign-made cells does not, in itself, run into a constraint. EHV and HV are slated for April 2027, and a working group has agreed on making the trigger the "grid-connection-contract application," though the text of the grid interconnection rules and the like is not yet promulgated. The fact that the trigger is the "application," not the "interconnection date," binds procurement and EPC schedules earlier, so deals aiming for an in-year interconnection need to confirm this sooner.

11 — Hold the risk in numbers

Closing — Choose by the numbers, not the contract's name

The first question an investor should put in Japan is not "can we do a long-term PPA." Storage has no PPA. The questions to ask are: which of merchant, tolling, floor or LDA; over how many years; at what price; and who is the counterparty. Without fixing, even a public estimate is in the red; with fixing, leverage lifts the equity. Investment judgement begins by confirming that structure not by the contract's name, but by the numbers. And for an investor coming from the US, the UK or the EU, the toolkit is the one you already know — the work is reading it through Japan's prices, Japan's counterparties and Japan's gates.