Storage Prices Are Splitting in Two: What CEA's Q3-2026 Forecast Means for BESS Project Capex

Intertek CEA's Q3-2026 price forecast puts a number on the split that storage developers have been arguing about all year: U.S. DC block prices fall roughly 15% to 30% by 2030, while import markets capture reductions exceeding 40%. Take a Q3-2026 U.S. DC block near US$195/kWh, apply the mid-case 26% decline, and you land near US$145/kWh. Apply the import-market trajectory of more than 40% to a China-sourced block and you reach roughly US$110/kWh. That is a US$35/kWh delta — US$7 million on a 200 MWh project, before EPC, interconnection or tax-credit effects.

This is not a story about batteries getting cheaper. Batteries have been getting cheaper for a decade. It is a story about one global market fragmenting into at least three pricing tiers — U.S. Section 45X-eligible, allied but non-U.S. (South Korean cathode), and China-sourced — each with its own cost floor, its own compliance risk and its own bankability profile. Anyone building an LCOE model with a single "BESS capex" scalar input is now modelling a market that no longer exists.

What CEA actually forecast

The Q3-2026 ESS Price Forecasting Report is a four-year forecast of DC container pricing — cells, modules, racking and balance of container. The headline published by pv magazine USA on 26 October was the 15–30% U.S. decline to 2030 against import markets exceeding 40%. The companion ESS Supply, Technology and Policy Report explains the mechanics.

Three drivers do most of the work:

• Cell form factor migration. Chinese manufacturers are moving from standard prismatic cells to larger formats — the 500–600 Ah class — which cut cell count, busbar, harness and rack hardware per MWh. CEA's own headline for the PFR is the 320 Ah versus 587 Ah LFP cost comparison, which is where most of the import-market reduction originates.
• Lithium and commodity relief. Lithium prices pulled back roughly 15% quarter-over-quarter, stabilising near US$20/kg, which offsets cost pressure in anodes, copper foil and current collectors.
• Pre-stocking ahead of the Chinese VAT change. Global lithium-ion shipments among profiled suppliers reached approximately 230 GWh in Q2 2026, up 18% quarter-over-quarter, largely as manufacturers pulled volume forward ahead of China's elimination of the remaining battery export VAT rebate on 1 January 2027.

That third point cuts both ways and is the reason the 2030 numbers are not simply an extrapolation of 2026 cost curves. The State Council Information Office's January 2026 announcement confirmed the rebate fell from 9% to 6% on 1 April 2026 and disappears entirely on 1 January 2027. ESS News reported at the time that this raises export costs for manufacturers. CEA's Q3 forecast explicitly notes that the impending removal is expected to soften the pace of cost deflation heading into 2027. In other words: the steepest part of the import-price decline may already be in the rear-view mirror, and the pre-2027 pre-stocking is borrowing from 2027 shipments.

The 2026 price stack, in one picture

The vertical spread in that chart is the whole story. A 587 Ah LFP cell lands near US$45/kWh. Put it in a container with modules, racking, BMS, HVAC, fire suppression and containerised DC balance-of-system and the China ex-works DC block sits around US$75/kWh. Cross a border into a jurisdiction with tariff exposure and compliance overhead and the same electrochemical content is quoted at US$110–195/kWh. The electrochemistry did not change. The institutional wrapper did.

Why the U.S. number has a floor

The U.S. domestic number is not simply the import number plus a tariff. Three structural costs keep it elevated:

1. Section 45X compliance is now a design input

Advanced Manufacturing Production Credits under Section 45X require careful traceability of cell, module and, critically, cathode active material. CEA notes that cathode active material is a primary barrier to 45X compliance, which is precisely why South Korean CAM producers — L&F, POSCO and EcoPro — have announced roughly 23 GWh of LFP cathode capacity for 2026. If fully realised, South Korean material could meet nearly 40% of U.S. domestic demand. That creates a third price tier: compliant but not fully domestic.

Projects using components tied to China-linked entities face what CEA describes as heightened vulnerability, because stringent definitions of covered foreign entities cast uncertainty over long-term bankability and tax-credit qualification. For a developer this is not a procurement footnote — a 45X credit is worth roughly US$10/kWh on the module/cell side alone over the credit period, which can exceed the entire projected 2026–2030 deflation. Losing eligibility to save US$20/kWh on the block is a losing trade.

2. Domestic capacity is slipping, not scaling

Roughly 11% of announced 2026 U.S. production capacity is at risk of slipping into 2027, per CEA — driven by line conversions, qualification hurdles and capital constraints among newer entrants. LGES is adding 33 GWh across three factories and Samsung SDI plans to start its Indiana LFP line, but those are anchor volumes, not a competitive merchant market. Domestic supply that is thin and late does not price like supply that is abundant and early.

3. Trade policy is still in motion

New U.S. trade rules, compounded by potential tariffs targeting overcapacity and forced-labour risk, complicate imported supply and add compliance hurdles. A tariff regime that can change inside a two-year procurement cycle is not modelled as a price — it is modelled as a risk premium, and risk premiums are what push a bank's required DSCR up and a project's effective cost of capital with it.

What this does to a capex model — and to LCOE

Take a 100 MW / 400 MWh (4-hour) standalone BESS. DC block is only one line in the cost stack. A representative U.S. build for 2026 looks roughly like this, using CEA's Q3-2026 DC block anchor plus typical non-block costs:

At US$315/kWh installed, the DC block is 62% of capex. So the CEA forecast is not a 26% capex reduction — it is a 16% installed-cost reduction if the full mid-case DC block decline is realised. Regenerate the same table with a 2030 mid-case block at US$145/kWh and the total lands near US$260/kWh. Run the import-market path at US$110/kWh and it lands near US$230/kWh — but only if 45X eligibility is not required for the revenue stack. That caveat is doing an enormous amount of work.

The LCOE sensitivity nobody models correctly

Storage LCOE is dominated by three terms: capex per kWh, annual cycled throughput, and the discount rate. A 16% capex reduction on a 4-hour asset drops LCOE by roughly 12–14% if throughput, augmentation schedule and financing cost are held constant. But they are not held constant:

• Augmentation capex is indexed to the same block price. A system planned for a 10% augmentation at year 8 to year 12 prices that top-up at the then-current block cost. Cheaper block prices reduce the augmentation reserve too — but only if the operator models augmentation as a future-price purchase rather than a fixed-contractual one.
• Larger cells change the degradation assumption, not just the price. A 587 Ah cell has a different thermal and current-density profile from a 280–320 Ah cell, and different cycle-life data. Assuming both degrade identically while assuming only the cost changes gets you a number that looks conservative and is not.
• The discount rate moves against you. A tariff-exposed supply chain attracts a higher required return. If the import path saves US$30/kWh of capex but adds 50 basis points to the project discount rate over 20 years, the NPV advantage largely evaporates. This is why the "just buy the cheap block" reasoning fails at the board level.

Sodium-ion: real, but not yet a price lever

CEA's report gives sodium-ion a market-focus section and a sober verdict. Hithium's containerised sodium-ion solution is cited as an example with extended service life and minimal augmentation requirements — genuinely attractive for high-cycle, long-duration duty where the LFP cycle-life penalty bites. But sodium carbonate raw material cost being a fraction of lithium carbonate does not translate into a cheap cell: processing the required precursor inputs remains constrained to smaller production scales. CEA's conclusion is that until domestic cell manufacturing depth for sodium-ion matures, the technology is unlikely to meaningfully alter near-term U.S. storage economics.

That is the correct read for 2026–2028 planning. Sodium-ion is a hedge against lithium and a hedge against augmentation cost, not a route to a lower first-cost block. Modelling it as a drop-in substitute at a lower $/kWh is currently an error.

AI data centres: the demand-side change that prices the block

One number in the CEA release deserves more attention than it got. U.S. lithium-ion imports for rack-level battery backup units designed for AI data centres reached 5,529 metric tons between January and July 2026, up 131% year-over-year. Imports for centralised UPS systems hit 10,723 metric tons, up 77% over the same period, driven by lead-acid replacement and the high-power demands of AI computing clusters.

This matters for grid-scale storage economics because data-centre backup competes for the same cells, and it pays on reliability and lead time rather than $/kWh. When a hyperscaler will take allocation at a premium to guarantee delivery, merchant availability for energy arbitrage projects tightens. The 2026 price stack is being set at the margin by a buyer whose objective function is not LCOE.

What to change in your model this quarter

Four adjustments, in order of impact:

1. Split the capex input. Stop using one BESS capex scalar. Carry block price, PCS price and BOS price separately, each with its own trajectory, so a 26% block decline does not get applied to interconnection costs that are rising.
2. Model the supply tier explicitly. Run 45X-eligible domestic, South-Korean-CAM-hybrid and import scenarios as separate cases with separate discount rates. The winning case is rarely the lowest-capex case.
3. Index the augmentation reserve. Tie the year-N top-up to a forward block-price curve rather than today's price. On a 26-year horizon this is a multi-million-dollar line item and it is frequently set at a flat rate.
4. Re-check cell-format degradation assumptions. If you moved from a 280–320 Ah to a 500–600 Ah cell for cost reasons, re-run SOH and RTE through the degradation table rather than holding the old curve.

None of this changes the direction of the market. Storage keeps getting cheaper on a levelised basis, and the U.S. decline of 15–30% by 2030 is a real, forecastable tailwind. What has changed is that "the price of storage" is no longer a single number you can look up. It is a function of where the cell was made, who the cathode supplier is, which credit the project qualifies for, and which buyer is standing in the allocation queue ahead of you.

Sources