The headline is the volume: 206 GWh over five years, 1.7 times everything EVE Energy shipped last year. The part that matters for the next cycle is who is buying — an integrator that has made AI data centres its growth story.
EVE Energy (300014.SZ) said on September 18 that its subsidiary Hubei EVE Power had signed a supply framework agreement with Fluence, the US storage integrator. Under the terms, the EVE Energy Fluence supply framework covers 206 GWh of battery production and delivery between 2027 and 2031, with 16 GWh committed for 2027 and 190 GWh of reserved capacity across 2028 to 2031.
What the EVE Energy Fluence Agreement Actually Says
The structure is standard for this industry and worth reading closely. It is a framework, not an order book. EVE’s filing states that the committed and reserved volumes apply to all product types, models, configurations and categories Fluence purchases, counted cumulatively in GWh equivalents — in other words, the total is a ceiling on future purchase orders, not a delivery schedule. Specific matters are subject to purchase orders signed separately, and the company said the financial impact for the current year is uncertain.
The counterparty is one of the three largest storage system integrators in the world. Fluence was formed in 2017 as a joint venture between Siemens and AES, is headquartered in Arlington, Virginia, and has been listed on Nasdaq since October 2021. It reported about $4.2bn of revenue in its 2025 financial year, up 15%, has deployed or contracted more than 22 GW of storage across 47 markets, and was rated a Tier-1 storage supplier for 2026 by S&P Global. It does not make cells — its business model is to buy them, which is what makes a Chinese cell maker its natural supplier.
Scale, for EVE, is the point. The 206 GWh in the framework is 1.7 times the company’s total power and storage shipments in 2025, which came to 121.2 GWh, and — measured against its storage shipments in the first half of 2026 — equivalent to more than two years of that business.
Why 206 GWh Is a Statement About EVE Energy’s Position
EVE’s storage business crossed a line this year. In the first half of 2026 the company shipped 44.46 GWh of storage cells, up 54.88% year on year, and storage revenue reached RMB 15.09bn (about $2.08bn), or 33.03% of group revenue. For the first time, storage overtook power batteries as its largest shipment segment, and SNE Research ranks EVE second globally in storage cell shipments.
The framework locks in the base of that business for five years and does so with a Western integrator, which matters in a market where buyers increasingly want non-Chinese supply chains for grid assets. For Fluence, the calculation is the reverse: securing cell capacity from a top-two global supplier at a moment when AI data-centre demand is pulling on the same factories that serve utility-scale storage.
The AI Data-Centre Door Fluence Opens
The reason this deal is more than a volume story is Fluence’s own pivot. Chinese trade coverage of the agreement points to Fluence’s June 2026 work with Siemens and NVIDIA — a power architecture for AI data centres, presented as a reference design for NVIDIA’s DSX Vera Rubin AI clusters — and to management guidance that the company has signed supply agreements with two hyperscale cloud customers.
Data-centre power is becoming the storage industry’s most demanding new buyer: it wants firm capacity, fast response, high cycle counts and a decarbonisation story, and it is willing to sign long contracts to get them. Any cell maker that becomes the supplier behind those contracts is inside the AI build-out, not adjacent to it. That is what EVE is buying with the volume discount it is implicitly granting — a route into the AI power supply chain through an integrator that already sits on it. The same dynamic is visible elsewhere in the power equipment market, where transformers and grid hardware are being ordered specifically for AI campuses.
What EVE Energy Gives Up for Volume
The margin line is where the deal gets uncomfortable. EVE’s storage battery gross margin was 12.51% in the first half of 2026 — roughly four percentage points below its power battery business and about half the level CATL reports. That gap is the price of a business model built on standardised large-format cells sold in bulk to integrators: high utilisation, limited pricing power.
A five-year framework reduces demand risk. It does not improve the margin, because the pricing is set order by order, and the framework itself contains no price. EVE also carries the working-capital load of building the capacity ahead of the orders. The company’s own filing is careful on this point, noting that the agreement does not affect its operational independence and that the specific financial impact depends on the purchase orders that follow.
The context is an industry where cell prices have been the battleground rather than a stable input — Chinese storage cell pricing has moved sharply this year, and pricing pressure has shifted between makers and buyers more than once. Signing 206 GWh gives EVE a floor on volume, not a floor on price. Competitors are racing on the same terms: storage is also where sodium-ion is arriving as a cheaper alternative chemistry, as HiTHIUM’s 4 MWh sodium-ion container showed this month. Localisation is the other pressure: Chinese battery makers are being pulled into US supply chains as much by customer requirements as by tariffs, the pattern behind CATL’s licensing arrangement for Ford’s US plant.
What the Fluence Deal Does Not Tell Us
Four gaps are worth holding on to.
There is no price. Not per GWh, not per year, not per cell format. Without it, the 206 GWh figure says nothing about revenue, and a framework can be renegotiated or allowed to lapse if the market turns.
There is no product specification. EVE has not said which cell platform the agreement covers, which matters for whether this is existing capacity or new lines. Nor has it said where the cells will be made, which matters for tariffs and local-content rules.
The 2027 commitment is small. Sixteen gigawatt-hours in the first year is roughly a third of EVE’s storage shipments in the first half of 2026 — a real but not transformational start. The other 190 GWh is reserved, not ordered.
The data-centre link is inferred. Fluence’s AI power work and its hyperscaler agreements are facts about Fluence; EVE’s filing does not say any of this volume is destined for data centres. That reading comes from the counterparty’s strategy, not from a disclosure by either company.
Accuracy note: The agreement was announced by EVE Energy in a filing dated September 18, 2026, and the structure — 206 GWh from 2027 to 2031, 16 GWh committed for 2027, 190 GWh reserved for 2028–2031, subject to separate purchase orders — is as the company disclosed it. The 121.2 GWh comparison, the 44.46 GWh and 54.88% figures, the 12.51% storage gross margin, the RMB 15.09bn storage revenue and the 33.03% revenue share are from EVE’s first-half 2026 results as reported in Chinese financial coverage; the 1.7-times comparison and the “more than two years of shipments” framing are our calculations from those numbers. Fluence’s revenue, project portfolio, market count and Tier-1 rating are as reported by the company and by Chinese financial media, which describe it as one of the world’s three largest integrators. Currency conversion uses RMB 7.25 per USD. The description of Fluence’s June 2026 work with Siemens and NVIDIA and its hyperscaler supply agreements comes from Chinese financial and trade coverage rather than from a document we could inspect directly, and we have not independently verified it. No pricing, product specification, manufacturing location or exclusivity was disclosed, and neither company has said any of the volume is destined for data centres — that link is our reading of Fluence’s strategy.
Sourcing note: The primary source is EVE Energy’s September 18, 2026 filing as reported by Chinese financial media, which carries the agreement structure and the company’s own characterisation of it. Chinese business coverage supplied the counterparty profile, the half-year comparisons and the analysis of EVE’s margin position, and trade coverage of the deal supplied the AI data-centre context around Fluence. EVsays did not speak to either company and had no access to the underlying agreement, so every figure here is either a disclosed number or a calculation from disclosed numbers, and the interpretation is ours. Our editorial standards are set out in our editorial policy, and corrections are handled under our correction policy.
Sources & Further Reading
- Sina Finance — “206 GWh: EVE Energy signs a super-sized order” (2026-09-18) — the filing summary: the 2027–2031 window, the 16 GWh commitment, the 190 GWh reservation, and the wording on how committed volumes are counted.
- Sina — “EVE Energy wins a 206 GWh storage order; nearly half of annual shipments pre-booked” (2026-09-18) — the scale comparisons against EVE’s 2025 shipments, the Fluence profile, the half-year storage figures, and the margin analysis.
- Chinese financial media coverage of the same filing — the AI data-centre context, including Fluence’s June 2026 power architecture work with Siemens and NVIDIA and its reported hyperscaler supply agreements.
- EVsays — related coverage: storage cell pricing in China, HiTHIUM’s sodium-ion storage container, AI data-centre power hardware and CATL’s US licensing route.
