1. The thesis in one sentence
Storage scarcity is shifting from cheap cells toward bankable connections, contracts and delivery, but profit is not moving in only one direction.
This is not a claim that storage necessarily makes money. The question is who controls pricing, who carries performance and market risk, and whether accounting profits become cash. Industry growth, shipment growth, gross-profit growth and shareholder returns are different outcomes. This report is not a securities recommendation.
2. Research boundary and revision log
2.1 Frozen initial boundary
The scope includes lithium and materials, storage cells, packs and DC containers, power conversion systems (PCS), AC integration, engineering and construction, grid connection, dispatch software, financing and project operations. Stationary lithium-ion battery energy storage systems (BESS), especially utility-scale and large commercial installations, are the core. Residential storage remains only where company disclosures cannot separate it; that contamination is explicit.
EV batteries, pumped hydro, hydrogen, compressed air and data-center UPS are not added to the same market total. They matter only as shared upstream demand or alternatives. Regions are the United States, China, Europe and the Gulf. EU statistics are not all-Europe statistics, and an Abu Dhabi example is not a census of Gulf countries.
The baseline is available full-year 2025 reporting, checked against disclosed 2026 quarters or half-years through 26 October 2026. Full-year 2026 actuals are unavailable. Fluence's fiscal year ends on 30 September: its FY2026 Q3 ended on 30 June 2026, not September. Comparisons preserve that difference.
2.2 Boundary revisions made to prevent false comparisons
The following changes narrow what can legitimately be concluded; they do not change the sample to force a preferred story.
| Revision | Reason and treatment | Consequence |
|---|---|---|
| Relabel upstream “Energy Storage” as a lithium-business proxy | Albemarle's segment also serves EV batteries [E11–E12] | Its earnings cannot all be assigned to stationary storage |
| Separate cells, DC systems, AC systems and deployments | Shipment, installation and revenue recognition occur at different stages | GWh rankings cannot substitute for profit rankings |
| Bridge exceptional items separately | Tesla tariff refunds affect both revenue and costs [E07] | Adjust both numerator and denominator; do not invent an official adjusted metric |
| Separate project EBITDA, fund earnings and investment returns | Depreciation, financing and augmentation can be substantial [E15–E16] | EBITDA margins cannot compete directly with manufacturers' gross margins |
| Include tax eligibility and implementation dates | US sourcing rules, Chinese rebates and EU passports affect addressable supply | Physical compliance does not establish financing or tax eligibility |
| Decline to invent global profit shares | Private firms, project SPVs and segment operating expenses are missing | Deliver a sample profit map, not an unsupported global pie chart |
2.3 Reading conventions
GM means gross margin; GP means gross profit; OM means operating margin. Adjusted EBITDA is separately labeled, never renamed operating profit. N/D means no comparable disclosure was obtained, not zero. Chinese accounting “operating profit” may also include items that do not match US operating income.
Confidence distinguishes confidence in a disclosed number from confidence in an industry-wide inference. High: clear definitions and directly verifiable records, or reproducible arithmetic. Medium: a single originating dataset, limited representativeness or conditional inference. Low: secondary reporting with unclear scope. Unresolved: conflicting definitions or evidence prevent a verdict. Repeated coverage of the same filing is not a second independent dataset.
3. Verdicts on four hypotheses
These hypotheses were rebuilt for storage rather than mechanically copied from the AI-compute template.
| Hypothesis | Verdict and evidence | Counterevidence or limitation | Confidence and reversal test |
|---|---|---|---|
| H1: Cheaper batteries necessarily make power plants more profitable | Reject “necessarily.” Pack prices fell [E17], while ERCOT unit revenues weakened [E23]; spreads, utilization, contracts and capital costs determine returns | Lower equipment costs still help new projects; GRID's contracted revenue provides a counterexample to universal deterioration [E16] | Medium. Upgrade if matched projects consistently retain cost savings beyond revenue compression |
| H2: Cells and lithium are the industry's only binding bottleneck | Reject. Connection queues [E24], safety requirements [E25] and financing [E26] can independently block delivery | Qualified, specific-format or locally eligible cells can still be scarce; aggregate capacity is not eligible supply | Medium. Regional delay causes and cost contributions are needed to validate ordering |
| H3: Shipment leadership guarantees profit leadership and a moat | Reject. Sample earnings differ widely, and ranking definitions differ [E20–E22] | Scale can improve procurement, warranties and lender acceptance; scale is not worthless | Medium; the exact global leader is unresolved. Require consistent AC/DC definitions and lifecycle economics |
| H4: Storage is one global standardized market where the lowest factory price wins | Reject. Taxes, eligibility, grid codes and contractual liabilities change deliverable costs [E28–E37] | Cells retain commodity characteristics; regulation does not automatically create durable technology advantages | Medium. Weaken the fragmentation thesis if compliance, financing and warranty costs converge across regions |
3.1 Counterevidence must survive the narrative
Upstream suppliers do not always lose, and integrators do not always win. Yuneng's improving material margins [E13–E14] and Fluence's losses [E08–E10] are important counterexamples. Input prices, product mix, contract pricing, manufacturing ramps and acceptance schedules may all matter. Without a causal decomposition, none is presented as the proven sole explanation.
Storage can also lower the very spreads it monetizes while continuing to create social value. More reliable electricity and flatter prices can transfer benefits to consumers rather than uncontracted asset owners. Social value and private investment returns are not identical.
4. Asset one: The profit map
4.1 Full-year 2025 baseline and latest matched 2026 disclosures
Figures below retain their original currencies. RMB billions, USD billions and GBP millions cannot be added across rows. Derived GP is revenue less reported cost; rounding is explicit.
| Layer / sample | FY2025 revenue and absolute profit | FY2025 GM / OM | Latest corresponding 2026 evidence obtained | What it cannot establish |
|---|---|---|---|---|
| Lithium: Albemarle Energy Storage | Revenue $2.710 bn; adjusted EBITDA $0.697 bn [E11] | GM and OM N/D; adjusted EBITDA margin 25.73 % | Q2 revenue $1.277 bn, adjusted EBITDA $0.723 bn, margin 56.67 % [E12] | Includes EV exposure and other segment-accounting effects; not plant profitability |
| Cathode: Hunan Yuneng phosphate products | Revenue RMB33.910 bn; GP 3.016 bn [E13] | GM 8.89 %; segment OM N/D | H1 revenue RMB34.272 bn, GP 5.077 bn, GM 14.81 % [E14] | Stationary-storage profit cannot be isolated |
| Batteries / systems: CATL storage | Revenue RMB62.440 bn; GP 16.676 bn [E01] | GM 26.71 %; segment OM N/D | H1 revenue RMB53.261 bn, GP 12.760 bn, GM 23.96 % [E02] | The Q3 group update [E03] cannot supply Q3 storage margins |
| Integration: Sungrow storage | Revenue RMB37.287 bn; GP 13.604 bn [E04] | GM 36.49 %; segment OM N/D | H1 revenue approximately RMB15.456 bn; GM 32.43 % [E05] | Group operating profit is not storage operating profit |
| Energy systems: Tesla energy | Revenue $12.771 bn; GP $3.802 bn [E06] | GM 29.8 %; segment OM N/D | Q3 revenue $2.881 bn, GP $0.954 bn, GM 33.1 % [E07] | Includes generation-related activities and exceptional tax effects |
| Integrator: Fluence Energy | Revenue $2.263 bn; GP $0.296 bn; operating loss $0.046 bn [E08] | GM 13.1 %; derived OM −2.05 % | FYQ3 revenue $0.650 bn, GP $0.033 bn, operating loss $0.058 bn; GM 5.1 %, OM −8.85 % [E09] | Not the similarly named water-treatment company Fluence Corp |
| Operations: GRID underlying portfolio | Revenue £60.4 m; EBITDA approximately £38.8 m [E15] | GM and OM N/D; EBITDA margin approximately 64.2 % | H1 revenue £34.7 m, EBITDA £23.5 m; margin 67.7 % [E16] | Not a net margin, dividend yield or project IRR |
| PCS, transformers, cooling, fire protection, software and EPC | No consistent standalone storage sample obtained | N/D, not zero | No substitution of entire-company profits | This gap limits any global profit-pool ranking |
4.2 Absolute profit and margins answer different questions
Within the more comparable same-currency, same-year storage product disclosures, CATL generated more gross profit than Sungrow, while Sungrow retained more GP per unit of revenue. Scale and margin can point to different leaders. Neither comparison establishes net profit or return on invested capital.
Tesla, Fluence and GRID cannot simply extend this ranking: scope, currency, fiscal periods and profit levels differ. In particular, a 67.7 % portfolio EBITDA margin is not evidence that ownership is more lucrative than manufacturing at a 36.49 % gross margin. Owners face depreciation, financing and reinvestment; manufacturers face R&D, sales costs, warranty obligations and working capital.
No global profit-share pie chart is published. Without a complete denominator, statements such as “cells capture 30 % and owners 40 %” would be speculation. The supported conclusion is substantial dispersion and a non-linear redistribution of profits, not a quantified global allocation.
4.3 Exceptional-item audit: Why 33.1 % cannot be extrapolated
Using [E07], the following analyst bridge isolates the specified tariff refund and customer refund only. USD millions:
- Reported revenue 2,881, cost 1,927, GP 954.
- Reverse the customer-refund revenue reduction: 2,881+339=3,220.
- Reverse the tariff-refund cost reduction: 1,927+526=2,453.
- Resulting GP 767 and GM 767/3,220=23.82 %; the net reported GP effect is 526−339=187.
- This is not a company-issued adjusted margin or a sustainable floor. Other mix, warranty and timing effects remain.
[Conflict E07—headline versus analyst bridge] 33.1 % and 23.82 % describe different adjustment scopes, not contradictory filings. Subtracting 526 from GP while ignoring 339 in revenue is an incorrect bridge. Keep the reported figure and the analyst calculation separately labeled.
4.4 Durability of different profit sources
This table is a mechanism map, not a measured allocation of industry profits.
| Profit source | Structural capability or cyclical benefit? | Evidence to watch |
|---|---|---|
| Low-cost resources and materials | Resource position can persist; selling prices and inventory are cyclical | Realized prices, cash costs, inventory and repricing |
| Cell yield and manufacturing scale | Yield, life and lender acceptance can endure; generic capacity invites competition | Unit GP and cash collection, not only GWh |
| Integration and delivery | Safety history, commissioning and guarantees matter; fixed-price liabilities can erase margins | Acceptance, delay penalties and warranty cash costs |
| Connection rights and land | Location scarcity can generate rents, subject to regulatory change | Actual operation dates, upgrade costs and usable capacity |
| Trading and dispatch | Execution matters, but weather and competitors still shape revenues | Net income after electricity, degradation and trading costs |
| Long-term / capacity contracts | Stability comes from reallocating risk, not creating free income | Counterparty credit, availability penalties and renewal prices |
| Tax incentives and refunds | Policy-dependent; not automatically a technical moat | Eligibility, contractual allocation and uncertainty |
5. Asset two: Regional bottleneck rankings
5.1 Ranking method
These are evidence-constrained research priorities, not precise statistical rankings. The criterion is whether a constraint can prevent commissioning and cash collection, not simply its procurement cost. Resolving the first constraint can reveal the next. Projects within one region can reverse the order. Confidence in exact rankings is no higher than medium.
Numbers below indicate investigation priority and should not be converted into a regional score.
| Region | 1 | 2 | 3 | 4 | Evidence and confidence |
|---|---|---|---|---|---|
| United States | Location, grid queue and upgrade costs | Tax-eligible sourcing and financeable contracts | Accepted system configuration and safety permits | Local manufacturing ramp and site delivery | Queues [E24], eligibility/tariffs [E30–E31], Fluence delays [E10]; medium |
| China | Monetizable tariffs, capacity income and utilization | Dispatch, connection and implementation rules | Cash collection and project financing | Degradation, warranties and continuing performance | Mandatory-storage reform [E28] and capacity mechanism [E29]; medium |
| Europe | Country-specific connection, permits and business model | Merchant saturation and long-term contracts | Safety, product compliance and data chain | Capital costs and supplier service longevity | EU growth [E19], UK portfolio [E16], passport [E35–E37]; medium-low, not all countries |
| Gulf | Bankable offtake / capacity contracts and procurement | Coordinated delivery of large combined projects | High-temperature performance and liability allocation | Execution between financial close and operation | Abu Dhabi case [E26], not a regional sample; medium-low |
5.2 An available cell is not necessarily a usable cell
The qualifications attached to a cell can matter more than the factory price: format, life, a frozen system design, lender and insurer acceptance, origin and tax eligibility, warranty allocation and consistent mass production. A cheap stocked cell may not be substitutable into a tested system without new validation or changes to guarantees.
UL9540 system certification and UL9540A testing are distinct [E25]. Neither can simply replace local permitting. Passing a test is not a universal promise that every installation can never burn. Procurement must identify actual configuration, spacing, fire design, operating conditions and responsible parties.
5.3 Four policy boundaries
United States: Section 301 is only one tariff component. PFE material-assistance restrictions, domestic-content bonuses and ordinary import compliance are separate tests. A legal import may fail to preserve an expected tax benefit. [E30–E31]
China: Removing mandatory storage as a prerequisite does not eliminate the storage market. Capacity payments may help standalone assets, but do not guarantee every project a uniform nationwide return. [E28–E29] A VAT export-rebate reduction is not automatically an identical increase in selling prices. Consumption tax has separate export treatment and should not be mechanically stacked on all exports. [E32–E34]
Europe: The relevant 18 February 2027 passport date differs from 18 August 2027 due-diligence requirements. At the cutoff, access-rights implementation details remain under consultation. [E35–E37]
Gulf: Financial close is a financing milestone, not evidence of an operating power station. Dividing a solar-plus-storage project's combined investment by storage GWh does not reveal battery procurement prices. [E26]
[Conflict E35 versus E36] “EU battery rules were delayed” is too broad. The documents concern different obligations and retain separate dates. The apparent conflict is resolved by scope, not by choosing one date for every requirement.
6. Asset three: The 2023–2026 profit-migration timeline
The timeline identifies conditional mechanisms, not synchronized movements across every company.
| Period | Observation | Possible profit transfer | Counterevidence and limit |
|---|---|---|---|
| 2023 | Exceptional electricity-market volatility supported some operators; Tesla energy GP $1.141 bn [E06,E23] | Scarce flexibility can earn rents; equipment scale can begin improving economics | One volatile ERCOT year is not a permanent revenue floor |
| 2024 | ERCOT unit revenues fell; Tesla energy GP reached $2.640 bn [E06,E23] | Competition can compress operating income while equipment profits grow | Two samples do not prove a global transfer of all operating profit to manufacturing |
| 2025 | Stationary pack prices declined [E17], EU additions reached 27.1 GWh [E19], integrator earnings diverged [E04,E08] | More projects become affordable while competition and execution redistribute savings | A quote is not completed cost; orders are not cash profit |
| H1 2026 | Material margins improved while some system margins compressed [E02,E05,E14] | Input rebounds and old fixed-price contracts may squeeze integrators | Product mix and other explanations must be separated before blaming lithium |
| H1–Q3 2026 | GRID total revenue rose while revenue/MW fell [E16]; Fluence deliveries slipped [E10]; Tesla tax effects appeared [E07] | Scale, contract coverage, accounting events and production execution explain different samples | 2026 is incomplete; H1 and Q3 must not be blindly annualized |
6.1 A path that should not be drawn as a straight line
Higher resource prices can pull profits upstream. Cheaper cells may temporarily benefit integrators or immediately pass through to owners. Cheaper projects can then enter the same power market, compress scarcity premiums and pass benefits to consumers. Contracts allocate price risk; timing determines who benefits first.
Three clocks must be tracked: input purchasing and repricing, equipment delivery and acceptance, and plant operation and cash settlement. A same-year margin snapshot cannot establish the entire transmission sequence.
6.2 Conflicting rankings: Preserve uncertainty rather than crown a winner
[Conflict E20 versus E22] Wood Mackenzie's 2025 integrator estimate and the alternative ranking reported by EnergyTrend disagree on leading suppliers and shares. The latter's original database and consistent AC/DC scope were not obtained. DC versus AC products, shipments versus deployments, project attribution and coverage are possible explanations, not proven reconciliations. Do not average the shares.
[Conflict E20 versus E21] Even within one research provider, market share and multi-factor capability scores are different rankings. First place in a capability ranking does not prove first place in market share. This apparent conflict is resolved by defining the metric.
6.3 More useful leading indicators
Record delivery terms and configuration alongside prices. For orders, track cancellation rights, deposits, milestones and guarantees. For operations, use both MW and MWh denominators, then deduct electricity purchases, degradation and availability losses. Total revenue may grow through expansion even while existing assets become less productive.
7. Asset four: Three counterfactuals and stress tests
7.1 Calculation discipline
These are analyst-defined mechanical scenarios, not company forecasts, observed quotations or promised returns. Amounts are USD millions unless stated otherwise. Transparent assumptions isolate sensitivities; they are not disguised empirical data. No Monte Carlo simulation or invented probability distribution is used.
7.2 Scenario one: Selling prices decline while procurement costs rebound
Assume an integration contract with revenue 100 and cost 80: cells 50, other costs 30, GP 20. Stress cell procurement costs by+20 % while contract revenue falls 10 %; hold other costs constant. A20 % cell-cost shock is not a 20 % lithium-price shock. Lithium content and pass-through require separate evidence.
The table illustrates the risk to an integrator unable to reprice.
| Item | Baseline | Stress |
|---|---|---|
| Revenue | 100 | 90 |
| Cells | 50 | 60 |
| Other costs | 30 | 30 |
| Gross profit | 20 | 0 |
| Gross margin | 20 % | 0 % |
Transmission: inputs or eligible cells become more expensive; procurement cash leaves first; fixed-price orders cannot pass costs through immediately; margins compress; liquidity and expansion are constrained. Upstream suppliers may benefit while old downstream contracts suffer. Inventory is not an automatic hedge because valuation, compatibility and impairment still matter.
Reversal conditions include locked purchasing costs, accepted indexation, compliant substitution and efficiency gains. [E14] and [E10] motivate investigation of these risks, not a claim that either company experienced this exact scenario.
7.3 Scenario two: CAPEX falls, but new plants consume the spread
Assume a 100 MW/400 MWh plant making 300 full discharges annually: 120,000 MWh delivered. Round-trip efficiency is 85 %, charging price $30/MWh. The base selling price is $80, falling to $60 under stress. CAPEX falls from 100 to 85. Other realizable net service revenue is 3 annually, O&M2 and augmentation reserve 1. Service revenue is assumed net of opportunity costs and operationally compatible with these cycles; otherwise it cannot simply be added. Interest, income taxes and terminal value are excluded.
The table uses a simple cash-contribution ratio, not IRR.
| Metric | Baseline | Only CAPEX −15 % | CAPEX −15 % and selling price $60 |
|---|---|---|---|
| Annual arbitrage: 120,000×(sale price−30/0.85) | 5.365 | 5.365 | 2.965 |
| Cash contribution after services, O&M and augmentation | 5.365 | 5.365 | 2.965 |
| Investment | 100 | 85 | 85 |
| Simple annual cash contribution / investment | 5.36 % | 6.31 % | 3.49 % |
This does not mean project IRR is 3.49 %. IRR requires annual degradation, augmentation timing, debt, tax and residual value. The example proves only that 15 % cheaper capital does not guarantee better economics when revenue compresses more. Fewer cycles or reserved ancillary-service capacity would change the result further.
Transmission: cheaper equipment admits more flexibility to the same node; peak spreads or ancillary premiums narrow; uncontracted owners face pressure. Long-term capacity contracts or better locations can break this chain. GRID's contract coverage is a counterexample, not an inconvenient exception to omit. [E16,E23]
7.4 Scenario three: The cheaper equipment loses an expected tax benefit
Assume US option A has CAPEX 100 and an expected tax benefit 30, leaving net investment 70. A qualified/local supply option B costs 15 % more, CAPEX 115, but retains an assumed 30 % investment credit, leaving 80.5. IfA loses the assumed benefit entirely, its net cost returns to 100, or 19.5 more thanB. The 30 % is a scenario parameter, not a tax determination for an actual project.
The table isolates the effect of eligibility.
| Option | CAPEX | Assumed tax benefit | Scenario net investment |
|---|---|---|---|
| A: lower price, full assumed eligibility | 100 | 30 | 70 |
| A: same equipment, assumed eligibility lost | 100 | 0 | 100 |
| B: 15 % more expensive, assumed 30 % eligibility retained | 115 | 34.5 | 80.5 |
Credit timing, transfer discounts and financing costs are not risk-free; they are omitted here only to isolate eligibility. Real analysis must establish construction timing, material ratios, contractual rights and applicable rules. [E30–E31] Eligibility restrictions are not blanket import bans; local manufacturing alone does not prove every condition is met.
Add a delay sensitivity: capital already deployed 70, annual funding cost 8 %, and a six-month delay imply approximately 2.8 of financing carry alone, before lost operating income or damages. Delivery failures can consume tax advantages. “Eligible” is not synonymous with “more profitable.”
7.5 Common lesson
Ask more than the battery price: How long is the price fixed? What triggers payment? Who carries lifetime and availability risk? Who receives tax benefits? When is grid operation possible? Companies that put these terms into enforceable contracts and deliver on them can convert technical advantages into cash advantages.
8. Asset five: Evidence gaps, falsification tests and next data collection
8.1 Gaps that can materially change the conclusion
The table identifies what is missing and how new evidence could change the thesis. It is not a commitment to automatically initiate additional work.
| Gap | Why it matters | Data needed | What would change the verdict |
|---|---|---|---|
| Global storage operating-profit denominator | Cannot identify the largest absolute profit pool | Private integrators, project SPVs, segment expenses, intercompany eliminations | A complete sample materially differs from these proxies |
| Stationary-storage share of upstream profit | EV demand contaminates material economics | End-use mix and associated costs | Storage customers are far less profitable than segment averages |
| Standalone PCS, safety, cooling and software earnings | Asset-light high-return layers may be missing | Storage-only sales, costs, service and R&D allocation | High sustained returns justify greater weight |
| Consistent AC/DC and shipment/installation rankings | Determines what “leader” means | Full methods, project lists and deduplication | Same-scope disagreement remains unresolved rather than cherry-picked |
| Matched-node lifecycle returns | Tests whether equipment savings reach shareholders | Hourly dispatch, degradation, debt, taxes, augmentation and residuals | Cost reductions persistently exceed revenue compression |
| Causal decomposition of system margins | Prevents attributing everything to lithium | Unit costs, mix, warranty releases and indexation | Mix rather than input pricing explains the change |
| European and Gulf representativeness | UK and UAE cases cannot represent entire regions | Country-level queues, permits, PPAs and availability | Cell shortages prove dominant in specific markets |
| Actual tax-benefit monetization | Nominal credits are not cash received | Legal eligibility, financing, discounts and receipt dates | Different eligibility/value/timing changes scenario three |
| Full-year 2026 and Q4 actuals | Cutoff precedes future disclosure | Update only when published | Seasonality or acceptance reverses the apparent trend |
8.2 Two avoidable research errors
[Conflict E09—excerpt versus complete statement] Fluence's 600.180 million revenue line excludes related parties. Adding 49.668 million related-party revenue gives 649.848 million. Resolve the missing row using the complete statement; do not turn an extraction problem into an allegation against the company.
[Conflict E17 versus E18] “North America costs 44 % more than China” and “China costs approximately 30 % less than North America” can both be true: 1−1/1.44≈30.6 %. Different denominators are not a contradiction. These overall pack comparisons cannot automatically become stationary-only regional price spreads.
8.3 A minimal dashboard for the next update
Track six items: comparable unit GP; backlog conversion into accepted revenue; delay and warranty cash costs; operating net revenue per MW and per MWh; contracted revenue coverage and counterparty credit; and availability, degradation and augmentation budgets. Preserve date, currency, denominator and source, leaving N/D where evidence is absent. Any “explosive growth” narrative should pass these checks.
This deliverable is an evidence-led industry draft, not country-specific legal advice or a project feasibility study. Real procurement, tax structuring or financing requires qualified review of actual project documents.
9. Evidence ledger
Each entry retains its date, source and independence status. A filing and news repeating that filing are one originating source; republication of a regulation is not independent verification. Disclosures may be verifiable while industry extrapolations remain only medium-confidence. Links were used for this research on 26 October 2026.
| ID | Data point / proposition | Value or finding | Time anchor | Source type | Source and link | Independent second source? | Confidence | Conflict flag |
|---|---|---|---|---|---|---|---|---|
| E01 | CATL 2025 annual report | Storage revenue RMB62.440bn; GM 26.71%; derived GP RMB16.676bn. | FY2025 | financial | CATL 2025 annual report | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E02 | CATL 2026 interim report | Storage revenue RMB53.261bn, GM 23.96%; revenue +87.54% YoY, GM −1.56pp. | H1 2026 | financial | CATL 2026 interim report | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E03 | CATL Q3 announcement (filing mirror) | Group quarterly revenue RMB165.221bn; no matched quarterly storage P&L obtained. | Q3 2026; published 2026-10-21 | financial | CATL Q3 announcement (filing mirror) | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E04 | Sungrow 2025 annual report | Storage revenue RMB37.287bn, GM 36.49%, GP 13.604bn; group operating profit RMB16.295bn. | FY2025 | financial | Sungrow 2025 annual report | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E05 | Sungrow 2026 interim report | Storage revenue approximately RMB15.456bn, GM 32.43%, versus 39.92% in H1 2025. | H1 2026 | financial | Sungrow 2026 interim report | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E06 | Tesla 2025 Form 10-K | Energy 2025 revenue $12.771bn, GP $3.802bn, GM 29.8%; 2023 GP $1.141bn. | FY2023–2025 | financial | Tesla 2025 Form 10-K | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E07 | Tesla Q3 2026 Form 10-Q | Energy revenue $2.881bn, costs $1.927bn; tariff refund reduces costs by $526m, customer refunds reduce revenue by $339m. | Q3 2026 | financial | Tesla Q3 2026 Form 10-Q | Not obtained; originating source or mirror | Medium: single origin or sample limits | C1: adjustment scope |
| E08 | Fluence Energy FY2025 results | Revenue $2.262830bn, GP $295.785m; derived operating loss $46.337m. | FY ended 2025-09-30 | financial | Fluence Energy FY2025 results | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E09 | Fluence Energy FY2026 Q3 statement | Revenue $649.848m, GP $33.241m, operating loss $57.520m. | Quarter ended 2026-06-30 | financial | Fluence Energy FY2026 Q3 statement | Cross-checked filing/release; same origin | Medium: single origin or sample limits | C2: excerpt resolved |
| E10 | Fluence FY2026 Q3 results and delivery update | Approximately $400m of deliveries shifted to FY2027; backlog $6.4bn. | 2026-08-05 | financial | Fluence FY2026 Q3 results and delivery update | Cross-checked filing/release; same origin | Medium: single origin or sample limits | — |
| E11 | Albemarle FY2025 results | Energy Storage segment revenue $2.710035bn, adjusted EBITDA $697.215m; also serves EV demand. | FY2025 | financial | Albemarle FY2025 results | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E12 | Albemarle Q2 2026 results | Segment revenue $1.276684bn, adjusted EBITDA $723.457m; not a power-station business. | Q2 2026; 2026-08-05 | financial | Albemarle Q2 2026 results | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E13 | Hunan Yuneng 2025 annual report | Phosphate cathode revenue RMB33.910bn, GM 8.89%; application mix is not profit mix. | FY2025 | financial | Hunan Yuneng 2025 annual report | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E14 | Hunan Yuneng 2026 interim report (filing mirror) | Phosphate cathode revenue RMB34.272bn, GM 14.81%, versus 6.97% a year earlier. | H1 2026 | financial | Hunan Yuneng 2026 interim report (filing mirror) | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E15 | Gresham House Energy Storage Fund FY2025 results | Underlying portfolio revenue £60.4m, EBITDA approximately £38.8m; not fund net income. | FY2025; published 2026-04-21 | financial | Gresham House Energy Storage Fund FY2025 results | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E16 | GRID interim report 2026 (report mirror) | Portfolio revenue £34.7m, EBITDA £23.5m; annualized revenue/MW £63,200 versus £75,100; contracted share 61%. | H1 2026 | financial | GRID interim report 2026 (report mirror) | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E17 | BNEF 2025 battery price survey | Stationary packs $70/kWh, −45% YoY; all-application average $108; neither is full plant CAPEX. | 2025; published 2025-12-09 | industry | BNEF 2025 battery price survey | Not obtained; originating source or mirror | Medium: single origin or sample limits | C3: denominator |
| E18 | IEA Global EV Outlook 2026: batteries | China pack prices approximately 30% below North America and 35% below Europe; not a stationary-only series. | 2025 data; 2026 publication | industry | IEA Global EV Outlook 2026: batteries | Another institution; underlying price data not fully independent | Medium: single origin or sample limits | C3: denominator |
| E19 | SolarPower Europe EU storage report | EU additions 27.1 GWh, utility scale 55%; EU is not all of Europe. | 2025; published 2026-01-28 | industry | SolarPower Europe EU storage report | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E20 | Wood Mackenzie 2025 BESS market share | Chinese integrators 76%; Tesla and Sungrow top two, BYD third; top three 30% combined. | 2025; published 2026-06-29 | industry | Wood Mackenzie 2025 BESS market share | Different estimate exists; definitions not reconciled | Medium: single origin or sample limits | C4: ranking scope |
| E21 | Wood Mackenzie 2026 integrator ranking | Capability ranking: Sungrow, Tesla, CATL; not a market-share ranking. | 2026-07-14 | industry | Wood Mackenzie 2026 integrator ranking | Not obtained; originating source or mirror | Medium: single origin or sample limits | C4: capability vs share |
| E22 | EnergyTrend report citing Benchmark and Electrek | Alternative ranking: BYD 13%, Tesla 10%, Sungrow 9%; original denominator not obtained. | 2025 estimates; published 2026-05-19 | media | EnergyTrend report citing Benchmark and Electrek | Different estimate exists; definitions not reconciled | Low / leader unresolved | C4: unresolved |
| E23 | Modo ERCOT battery revenue analysis | Revenue/kW: 2023 $193, 2024 $56; 2025 $29.4 was a full-year nowcast, not a final actual. | 2023–2025; published 2026-02-03 | industry | Modo ERCOT battery revenue analysis | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E24 | Lawrence Berkeley National Laboratory: Queued Up | Storage queue approximately 749 GW; 2025 completed projects had median 61 months to operation, not storage-only. | End-2025 queue; 2026 release | industry | Lawrence Berkeley National Laboratory: Queued Up | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E25 | UL: ESS codes, standards and testing | UL9540 system certification and UL9540A thermal-runaway propagation testing are different. | Accessed 2026-10-26 | industry | UL: ESS codes, standards and testing | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E26 | Masdar financial close announcement | 5.2 GW solar+19 GWh storage; total project investment $6.1bn; financial close is not operation. | 2026-07-13 | industry | Masdar financial close announcement | Not obtained; originating source or mirror | Medium: single origin or sample limits | — |
| E27 | China NEA storage deployment release | New-type storage 136 GW/351 GWh, average 2.58 hours; excludes pumped hydro. | End 2025; published 2026-01-30 | regulatory | China NEA storage deployment release | Not obtained; originating source or mirror | High for text; medium for application | — |
| E28 | China NDRC policy 136 | Storage installation cannot be a prerequisite for renewable approvals or grid connection; not a storage ban. | 2025-02-09 publication | regulatory | China NDRC policy 136 | Not obtained; originating source or mirror | High for text; medium for application | — |
| E29 | China capacity pricing policy explanation | Grid-side standalone storage capacity pricing mechanism; local implementation and contribution matter. | 2026-01-30 | regulatory | China capacity pricing policy explanation | Not obtained; originating source or mirror | High for text; medium for application | — |
| E30 | IRS Notice 2026-15 publication | PFE material-assistance restrictions concern 45X, 45Y, 48E; distinct from domestic-content bonus requirements. | 2026-02-12 | regulatory | IRS Notice 2026-15 publication | Not obtained; originating source or mirror | High for text; medium for application | — |
| E31 | USTR final Section 301 modifications | Section 301 rate on non-EV lithium-ion batteries rises to 25% in 2026; not total landed tariff. | 2024-09-18; implementation 2026 | regulatory | USTR final Section 301 modifications | Not obtained; originating source or mirror | High for text; medium for application | — |
| E32 | China battery export VAT rebate notice | Listed batteries: VAT export rebate 9%→6% for Apr–Dec 2026; cancellation from Jan 2027. | Issued 2026-01-08 | regulatory | China battery export VAT rebate notice | Not obtained; originating source or mirror | High for text; medium for application | — |
| E33 | China battery consumption tax adjustment | Covered batteries: 2% from Sep 2026, 4% from Sep 2027; product scope matters. | 2026-07-17 | regulatory | China battery consumption tax adjustment | Not obtained; originating source or mirror | High for text; medium for application | — |
| E34 | China tax authority export clarification | Direct exports receive consumption-tax exemption; qualifying tax-paid exports refunds; no blanket extra 2% export charge. | 2026 policy clarification | regulatory | China tax authority export clarification | Not obtained; originating source or mirror | High for text; medium for application | — |
| E35 | European Commission battery-passport guidance | Relevant industrial batteries>2 kWh require passports from 18 Feb 2027; guidance does not add legal obligations. | 2026-08-21; requirement 2027-02-18 | regulatory | European Commission battery-passport guidance | Not obtained; originating source or mirror | High for text; medium for application | C5: obligation/date |
| E36 | EU Council battery due-diligence delay | Relevant due-diligence duties delayed to 18 Aug 2027; this does not postpone passports. | 2025-07-18; application 2027-08-18 | regulatory | EU Council battery due-diligence delay | Not obtained; originating source or mirror | High for text; medium for application | C5: obligation/date |
| E37 | EU battery-passport access-rights consultation | Access-rights implementing rules remain under consultation at the cutoff; not all details are final. | Consultation 2026-10-19–2026-11-16 | regulatory | EU battery-passport access-rights consultation | Not obtained; originating source or mirror | High for text; medium for application | — |
Reproducible analyst calculations
The ledger does not disguise calculations as external data: GP=revenue−cost; OM=operating profit/revenue; Albemarle adjusted EBITDA margins remain separately named. The Tesla bridge is in section 4; every stress-test assumption and formula is in section 7. Arithmetic is high-confidence; real-world representativeness of the assumptions is not established, and no empirical probability is assigned.
10. References by source type
E identifiers match the ledger. Mirrored filings remain classified as company reports, with the mirror identified.
Financial filings and corporate disclosures
- [E01] CATL 2025 annual report — FY2025.
- [E02] CATL 2026 interim report — H1 2026.
- [E03] CATL Q3 announcement (filing mirror) — Q3 2026; published 2026-10-21.
- [E04] Sungrow 2025 annual report — FY2025.
- [E05] Sungrow 2026 interim report — H1 2026.
- [E06] Tesla 2025 Form 10-K — FY2023–2025.
- [E07] Tesla Q3 2026 Form 10-Q — Q3 2026.
- [E08] Fluence Energy FY2025 results — FY ended 2025-09-30.
- [E09] Fluence Energy FY2026 Q3 statement — Quarter ended 2026-06-30.
- [E10] Fluence FY2026 Q3 results and delivery update — 2026-08-05.
- [E11] Albemarle FY2025 results — FY2025.
- [E12] Albemarle Q2 2026 results — Q2 2026; 2026-08-05.
- [E13] Hunan Yuneng 2025 annual report — FY2025.
- [E14] Hunan Yuneng 2026 interim report (filing mirror) — H1 2026.
- [E15] Gresham House Energy Storage Fund FY2025 results — FY2025; published 2026-04-21.
- [E16] GRID interim report 2026 (report mirror) — H1 2026.
Industry research and technical standards
- [E17] BNEF 2025 battery price survey — 2025; published 2025-12-09.
- [E18] IEA Global EV Outlook 2026: batteries — 2025 data; 2026 publication.
- [E19] SolarPower Europe EU storage report — 2025; published 2026-01-28.
- [E20] Wood Mackenzie 2025 BESS market share — 2025; published 2026-06-29.
- [E21] Wood Mackenzie 2026 integrator ranking — 2026-07-14.
- [E23] Modo ERCOT battery revenue analysis — 2023–2025; published 2026-02-03.
- [E24] Lawrence Berkeley National Laboratory: Queued Up — End-2025 queue; 2026 release.
- [E25] UL: ESS codes, standards and testing — Accessed 2026-10-26.
- [E26] Masdar financial close announcement — 2026-07-13.
Media and secondary cross-checks
- [E22] EnergyTrend report citing Benchmark and Electrek — 2025 estimates; published 2026-05-19.
Regulatory and official policy sources
- [E27] China NEA storage deployment release — End 2025; published 2026-01-30.
- [E28] China NDRC policy 136 — 2025-02-09 publication.
- [E29] China capacity pricing policy explanation — 2026-01-30.
- [E30] IRS Notice 2026-15 publication — 2026-02-12.
- [E31] USTR final Section 301 modifications — 2024-09-18; implementation 2026.
- [E32] China battery export VAT rebate notice — Issued 2026-01-08.
- [E33] China battery consumption tax adjustment — 2026-07-17.
- [E34] China tax authority export clarification — 2026 policy clarification.
- [E35] European Commission battery-passport guidance — 2026-08-21; requirement 2027-02-18.
- [E36] EU Council battery due-diligence delay — 2025-07-18; application 2027-08-18.
- [E37] EU battery-passport access-rights consultation — Consultation 2026-10-19–2026-11-16.
This file is the Day02 deep-research draft, bounded by evidence available through 26 October 2026; conflicts and unresolved items are retained, and it is not investment advice.