Editorial Article
The Cheaper Cell and the Dearer Megawatt-Hour
The most repeated sentence in the energy transition is that storage gets cheaper every year. It does. It is also, for a developer signing a power purchase agreement in 2026, the least useful sentence in the industry, because the thing a developer buys is not a cell. It is a megawatt-hour that has passed an interconnection study, cleared an eligibility test and been paid for by a contract. And that product got substantially more expensive last year.
The two numbers are both true and they do not sit easily together. IRENA records battery storage costs falling 93 per cent since 2010, to USD 192/kWh in 2024, with cell prices down 20 per cent in 2024 alone to USD 115/kWh. Lazard’s 2026 analysis puts utility-scale levelised cost of storage without the investment tax credit at USD 210 to 292/MWh — against USD 115 to 254/MWh in 2025. On midpoints that is USD 184.5 to USD 251, a rise of 36.0 per cent in twelve months.
The hardware fell 20 per cent in a year in which the delivered megawatt-hour rose 36 per cent. Nothing in the battery changed. Something in the eligibility rules did.
What the FEOC rules actually removed
The United States’ foreign-entity-of-concern restrictions, introduced in 2025, limit investment tax credit eligibility for projects using Chinese-manufactured cells. That is a single sentence with a very large consequence, because the cheapest cells in the world are Chinese cells, and an ITC that cannot be claimed is not a subsidy — it is a line item the developer has to fund from equity instead.
Lazard’s own numbers show how completely the credit is doing the work. With the ITC, 2026 storage lands at USD 148 to 209/MWh — a midpoint of USD 178.5, which is 3.25 per cent below the 2025 midpoint without any credit at all. Strip the credit and the same technology, in the same year, costs 36.0 per cent more. The credit is not a sweetener on top of a falling cost curve. On these figures it is the cost curve.
That reverses the usual reading of a storage forecast. The market this describes is not priced; it is underwritten. Which means the forecast has to be read as a schedule of policy events rather than as a projection of a cost trend — and, unusually, that makes it more reliable than most of the chemistry forecasts in this chain, because a statute has a date on it.
Three demand pulls, all with dates
Germany began market-based procurement of inertia — instantaneous reserve — by transmission system operators in January 2026, offering long-term contracts specifically for grid-forming battery storage. That is a new revenue stack for a capability that batteries have and gas plants also have, and it is being bought on contract rather than on price.
India‘s POWERGRID invited bids in July 2026 for a 500 MW / 1,000 MWh battery system across multiple locations — a two-hour asset bought by tender, which tells you the buyer is procuring reliability rather than arbitrage.
New Hampshire‘s HB 1718, passed in 2026, sets a framework for net metering credits for solar-charged stationary storage, which is the residential version of the same mechanism: the value is created by the tariff, not by the cell price.
Add the announced factories and the picture is consistent: Peak Energy is building a 4 GWh sodium-ion plant in California with shipments from Q1 2027 and more than 6 GWh of customer commitments already in hand; Zenergy signed a 50 GWh long-duration LFP project in Suzhou, phase one in Q3 2027 and phase two in 2028, aimed at 4 to 12 hour discharge; and Ultium Cells began making LFP cells for storage in July 2026 by converting an EV line in Tennessee, which is precisely what domestic-content compliance looks like when it is cheaper to convert a plant than to build one.
The constraint is not the cell
Here is the part that most storage forecasts miss by sizing the battery and forgetting the box around it. The IGBT and silicon-carbide module market faces a demand-versus-foundry-capacity gap exceeding 18 per cent in 2026, with lead times of 30 to 45 weeks. At least one manufacturer had ON Semiconductor IGBT allocations pushed out by 15 weeks, which is not a procurement inconvenience — it is a commissioning date moving.
An inverter shortage is a storage shortage. The industry’s response has been to make the cell bigger and the count smaller: CATL at 587 Ah, EVE Energy at 628 Ah, HiTHIUM at 1,175 Ah, which improves module-level energy density by 60 to 80 per cent and cuts balance-of-system cost per installed kilowatt-hour. That is a real answer, and it is an answer to a different question than price per cell — it buys down the cost of everything that is not the cell.
And the materials did not cooperate either
The upstream record in 2026 is a list of single points of failure. The Democratic Republic of Congo produces about 70 per cent of the world’s cobalt; its regulator ARECOMS set an annual export cap of 96,600 tonnes for 2026 and 2027, and in July 2026 a customs-platform administrative failure threatened as much as 20,000 tonnes worth USD 1.1 billion, with 60 to 75 per cent of companies judged unlikely to meet the quota deadline. Cobalt is up 160 per cent since February 2025 to about USD 26/lb, or USD 57,320 a tonne.
Nickel went the other way and still hurt: BHP suspended Nickel West from October 2024 on oversupply, taking an underlying EBITDA loss of about USD 300 million, with a review due by February 2027 and roughly AUD 450 million a year still being spent on the assets. Graphite from Mozambique’s Balama mine was halted for nearly a year on political unrest. Lithium extraction in Chile faces water-rights disputes with Indigenous communities. Indonesia’s share of nickel supply went from 31.5 per cent in 2020 to about 60 per cent in 2024 after a raw-ore export ban, and the state is now tightening control of the sector it created.
None of that is a cost curve. It is a map of where the next interruption comes from, and it argues the same point as the FEOC rules from the other direction: the delivered price of storage is set by jurisdiction, not by chemistry.
What the number is, and what it is not
The market is USD 64.54 billion in 2025, projected to USD 247.68 billion by 2033 at an 18.31 per cent baseline CAGR, with an optimistic case at 20.81 per cent and a conservative case at 15.81 per cent. That is 3.84 times in eight years. Growth is front-loaded, with the strongest years between 2028 and 2030 — which is to say the forecast’s steepest section sits immediately after the capacity coming online in 2027 and 2028, and immediately after the EU battery passport obligation lands in February 2027.
One caveat belongs in the article rather than in a footnote. This base year is derived from a single web estimate, and the record says so. The level should therefore be treated as indicative; the growth shape is a different proposition, because it is driven by dated procurement, dated statutes and dated factory start-ups, and those can be checked one by one. Use the rate. Do not defend the base.
And hold the scale in mind against the chemistry that is supposed to serve it. Sodium-ion cell output in 2033 is forecast at 500 GWh, with more than 60 per cent of sodium cell demand going to stationary storage — inside a market whose 2033 value is USD 247.68 billion. Peak Energy’s 4 GWh is, against that, a rounding error today and a credible wedge tomorrow. The gap between those two statements is the entire investment case, and it closes or it does not on whether the megawatt-hour gets cheaper — which is now a policy question with a date on it, not a manufacturing one.
This analysis draws on the stationary energy storage systems market report, and on the sodium-ion battery materials and electrolyte work, which place more than 60 per cent of sodium cell demand in stationary applications.
Frequently Asked Questions
How large is the stationary energy storage systems market?
USD 64.54 billion in 2025, rising to USD 247.68 billion by 2033 at an 18.31% baseline CAGR — 3.84 times in eight years. The optimistic case is 20.81% and the conservative case 15.81%. The base year is derived from a single web estimate, so the level should be treated as indicative while the growth shape is supported by dated procurement and legislation.
Are storage costs rising or falling?
No — they are different products. IRENA records battery storage costs falling 93% since 2010 to USD 192/kWh in 2024, and cell prices fell 20% in 2024 to USD 115/kWh. But Lazard's 2026 analysis puts utility-scale LCOS without the ITC at USD 210–292/MWh against USD 115–254/MWh in 2025 — a 36.0% rise on midpoints. The delivered megawatt-hour includes cells, inverters, interconnection, permitting and tax-credit eligibility.
Why did storage get more expensive in 2026?
FEOC (foreign entity of concern) restrictions introduced in 2025 limit investment tax credit eligibility for projects using Chinese-manufactured cells. With the ITC, 2026 LCOS lands at USD 148–209/MWh, whose midpoint is 3.25% below the 2025 midpoint with no credit at all. Without it, the same technology costs 36.0% more year on year — so on these figures the credit is effectively the cost curve.
What is the binding constraint on storage deployment?
Power electronics. IGBT and silicon-carbide module demand exceeds foundry capacity by more than 18% in 2026, with lead times of 30–45 weeks, and at least one manufacturer had ON Semiconductor IGBT allocations pushed out by 15 weeks. Large-format cells — CATL 587 Ah, EVE 628 Ah, HiTHIUM 1,175 Ah — improve module-level energy density 60–80% and cut balance-of-system cost, but they do not fix an inverter shortage.