Fluence Energy

Refresh delta (since 2026-06-17)

What changed

1. Fiscal Q3 missed and margins collapsed. Revenue $649.8M (+8% YoY) but roughly $90M below plan; GAAP gross margin 5.1% vs 14.8% a year earlier; adj. EBITDA $(29.3)M vs +$27.4M; net loss $(44.3)M. The 10-QForm 10-QThe quarterly version of the annual report. Lighter, and not audited. names three causes: liquidated damages for late projects, cost overruns deploying Gridstack Pro and Smartstack (including US-built projects), and higher battery prices on fixed-price contracts, partly offset by IEEPA tariff refunds.
2. FY26 guidance was cut twice. On 2026-08-05: revenue $3.2–3.6B → $2.9–3.1B, adj. EBITDA $40–60M → $(30)M to +$10M, with ~$400M of deliveries pushed into FY27. On 2026-09-16: revenue to ~$2.4B and adj. EBITDA to ~$(200)M loss, blamed on the Houston contract-manufacturing ramp. Nine-month actuals are $1.59B revenue and $(90.8)M adj. EBITDA, so the new guide implies a fiscal Q4 of about $0.81B revenue and a ~$(109)M adj. EBITDA loss.
3. Demand held, and the data-center buyer arrived. Order intake $1.44B in the quarter (≈3× the year before), backlog a record $6.4B (48–53% to recognize in the next 12 months), and ~$850M of data-center business through July: a $300M behind-the-meter order from a developer plus ~$550M of awards from a hyperscaler, the latter not yet firm POs.
4. Cash is now the constraint. Cash fell from $690.8M (Sept 2025) to $339.3M (plus $25.6M restricted); nine-month operating cash flow was $(366.5)M; inventory rose to $783.0M from $455.0M. On the call the CFO said Fluence "may require an additional $300 million to $500 million of working capital over the coming year". Six weeks later the stated FY27 objective became neutral-to-positive operating cash flow "without external capital".
5. Operations leadership was replaced from the owners' benches. On the August call, Roman Loosen (ex-Siemens) took over supply chain and manufacturing. On 2026-09-15 Bernerd Da Santos, a 26-year AES executive (AES COO 2017–2023) and until that day an AES-designated Fluence director, became EVP and COO.
6. The founders are selling down. On 2026-05-15 AES Grid Stability, SPT Holding and Qatar Holding sold 23.0M Class A shares in a secondary, and AES redeemed 10.07M LLC units; Fluence received nothing. Shares outstanding are 143.2M Class A + 41.4M Class B-1 (2026-07-31). The 10-Q also discloses $3.8M of nine-month "legal and consulting fees related to potential strategic transactions."
7. Stock: ~$23.5 → ~$7.46 (2026-09-24 close, market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth. ~$1.38B), including a ~15% one-day drop after the September cut. Targets were cut across the Street (Baird to Sell at $3, RBC $4, Mizuho Underperform at $6, BMO $7, JPMorgan $8); UBS went the other way, upgrading to Neutral at $12 on a likely FY26 trough.

What held. The structural read stands: an integrator with no cell IP, a policy-shaped US moat, and chronic execution risk. Demand, the part the June bull case leaned on, actually came in stronger than expected. What broke is the part the bear case named: execution concentrated in the second half. The June pre-mortem ("H2-FY26 slipped (again)… the stock round-tripped to the low teens") happened within three months, and went further than it said.

Phase A — Understand the business

Company Overview — refreshed (new product, new buyer)

Unchanged at the core: a battery-storage integrator that buys cells, assembles and controls systems, and sells services and software on the fleet (see Lens 1, the previous dossier). Two things moved:

• Product. 75% of year-to-date orders are for Smartstack, and Smartstack 10 raises density from 7.5 MWh to 10 MWh per unit. The margin damage this year comes from launching new platforms (Gridstack Pro, Smartstack) on new contract-manufacturing lines at the same time, not from the legacy Gen6 fleet.
• Buyer. Data centers are now a real segment: a 16 GWh data-center pipeline (+35% on the prior quarter) and a sales cycle management calls "much faster than our traditional market segment, converting from lead to order in 3 months".
• Operating metrics (2026-06-30 vs 2025-09-30): deployed 7.4 GW (19.3 GWh); contracted backlog 12.6 GW (+38%); pipeline 45.6 GW / 163.7 GWh (+34%); services AUM 6.3 GW.

Supply Chain — carried, with one correction

See the previous dossier, Lens 2. Correction: the contract-manufacturing risk named there as "Arizona" now centres on Houston (15 GWh/yr, fully automated, limited production from fiscal Q3, full production targeted for fiscal Q1 2027, running on temporary generators while it waited for a grid connection) and on a new international facility (China, per the call) whose "initial production did not meet our quality expectations" and needed rework. Fluence also booked a ~$15M upfront cost for a planned long-term international cell-supply agreement, which buys price and supply certainty.

Competitive Advantages — carried

See the previous dossier, Lens 3. The domestic-content argument is unchanged in the 10-Q (US suppliers believed compliant with OBBBA prohibited-foreign-entity rules under the 2026-02-12 Treasury guidance). The new wrinkle: that moat only pays if Fluence can build in the US, and US-built projects are where the overruns sit.

Segments — refreshed (mix moved)

. The June dossier called EMEA the bright spot; it fell by two-thirds this quarter while APAC (Australia) carried growth. Nine-month Americas revenue is +49%. Customer concentration rose: top three customers ≈ 48% of nine-month revenue (top two were 40% a year earlier). Related-party (AES/Siemens) revenue was 7.6% of the quarter.

Phase B — Measure performance

Earnings Result (fiscal Q3 2026, quarter ended 2026-06-30, reported 2026-08-05)

. Nine months: revenue $1,590.0M (+30%), gross margin 6.5% (12.5%), net loss $(136.1)M, adj. EBITDA $(90.8)M.

vs guidance. Revenue came in about $90M short of plan, and the release cut the year the same day (Lens 11). Adjusted gross margin was 5.9%.

Balance sheet. Deferred revenue $956.5M (up from $640.5M), so customers keep prepaying, which is a demand signal. Against that: supplier advances $226.4M (from $126.8M), unbilled receivables $328.2M (from $239.6M), and inventory up 72% since September. Working capital is swallowing the cash. Total liquidity was ~$863M including ~$307M of undrawn revolver capacity, but the revolver's cash-draw sublimit is $150M and the covenants require $150M minimum liquidity through 2026-12-31. The $400M 2030 convertible is the only funded debt.

Earnings Calls — refreshed

First transcript on disk (fiscal Q3, 2026-08-12 per fool.com). The tone moved from May's confidence ("record backlog," "two hyperscaler MSAs") to "demand is not the problem, production is":

• Candid on causes: Nebreda said "initial production did not meet our quality expectations" at the international plant, and described Houston's construction and automation delays.
• Still firm on margin: "we are still very comfortably within the 10% to 15% range… if you take out some of the onetime stuff… around the 12%," and the backlog and new orders are "in line with the 10% to 15%." He waved off a peer's warning about storage-margin pressure.
• Confident on Q4: "roughly half of what we need to for the quarter" was already produced and integrated. Five weeks later that Q4 was cut by about $0.6B. That gap between the call and the September cut is the most useful sentiment reading in this refresh: management's near-term operational visibility has been wrong twice in one quarter.
• New phrase: "working capital," in the $300–500M need. Dropped: the "~70% of revenue in H2" framing.

Comps — refreshed on price

At ~$7.46 × 184.6M shares ≈ $1.38B equity; plus $0.39B convertible minus ~$0.37B cash ≈ $1.4B EV, or ~0.6× EV/Sales on the ~$2.4B FY26 guide (vs ~1.2× in June). The peer table is unchanged and still mostly n/a (see the previous dossier, Lens 7). The stock now trades below the ~$10–11 consensus target, the reverse of June, though the target is falling toward it.

Stock-Price Catalysts — refreshed

• 2026-05-15: 23.0M-share founder secondary (AES Grid Stability, SPT Holding, Qatar Holding).
• 2026-08-05/06: fiscal Q3 miss and first FY26 cut.
• 2026-09-16/17: second cut and COO change; ~−15% on 2026-09-17 ($9.05 → $7.66), 52-week low $7.01.
• Net: ~$23.5 (June dossier) → ~$7.46 (2026-09-24), about −68%. The June lens said "earnings misses are forgiven if backlog/orders are records." That stopped being true once the misses became guidance cuts and the cash began to fall.

Phase C — Judge people & books

Management — refreshed

Nebreda stays CEO and Pasha stays CFO; the operations layer changed twice in five weeks (Loosen from Siemens on supply chain; Da Santos from AES as COO). Da Santos's package: $650K base, 100% target bonus, $700K sign-on cash, $700K one-year RSU, ≥$1.5M FY27 long-term incentive. The read: the owners are putting their own operators into the engine room. That helps execution, and deepens the related-party governance pattern from Lens 10 of the June dossier. The CEO's record now reads as two bad years out of three (FY25 guide cut twice, FY26 guide cut twice). His FY24 turnaround looks more like the exception than the trend.

Forensic Red Flags — refreshed (new filings)

• SEC: the 10-Q states "the SEC is conducting a formal investigation" following the February 2024 short report. The June dossier called it an "open inquiry"; the correct word is formal investigation, still unresolved.
• Securities class action: still dismissed without prejudice (2026-03-31) as of the 10-Q; derivative suits stayed. After the September drop several plaintiff firms announced "investigations". No amended or new complaint was found in this run. A new class period around the FY26 guidance is a live risk.
• Estimate risk in over-time revenue: this year's gross-margin hits come partly from "increased estimated total contract costs on certain projects", which are catch-up adjustments on percentage-of-completion contracts. That is exactly the estimate-driven area the 2024 short report attacked. It shows up here as honest bad news, not concealment, but watch for more catch-ups in the FY26 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes..
• Working-capital quality: inventory $783M, supplier advances $226M, unbilled $328M against $339M cash. An inventory write-down is the obvious FY26 10-K risk if Smartstack rework or battery-price moves leave stock mis-costed. The existing provision is $25.2M.
• Strategic-transaction fees ($3.8M nine months) with owners selling down: not a red flag in itself, but it means something is being explored. Do not over-read it.
• Regulatory pre-fetch (regulatory/regulatory-findings.md, run 2026-09-24): no SEC litigation release or AAER.

Phase D — Project & stress-test

Forward Projection — re-run

Anchors: FY26 guide ~$2.4B revenue / ~$(200)M adj. EBITDA; ~$2.2B of backlog scheduled for FY27 revenue as of June, plus the ~$400M (and now more) pushed out of FY26; backlog $6.4B; FY27 plan due with the FY26 print (~Nov 2026).

Brier forecast (June): "FLNC FY26 revenue ≥ $3.4B, p≈0.55." It was not logged (watchlist run), and on the September guide it will resolve NO when FY26 closes on 2026-09-30. Calibration note for next time: backlog coverage said "demand risk low," which was right, but the forecast priced timing risk at 45%, and timing was the whole outcome. Not logged this run either (unattended). If promoted, the scoreable line would be "FLNC FY27 revenue ≥ $3.2B," p≈0.5.

Bull vs Bear — re-run

Bull. Orders, backlog and the data-center pipeline all rose while the stock fell 68%. At ~0.6× EV/Sales on a trough year, the market prices Fluence as if the manufacturing problem were permanent. Houston is a fixable, physical bottleneck (grid hook-up, automation), the China quality issue is reported as resolved, owner-appointed operators are in charge of fixing them, and deferred revenue of ~$1B says customers are still paying ahead. If FY27 simply lands the slipped revenue at ~10% gross margin, the earnings base resets well above FY26.

Bear.

1. Funding before fixing. $339M cash, a $300–500M working-capital need, $150M liquidity covenant, and a promise of "no external capital" that rests on the same operational forecasting that just missed twice. A dilutive raise near the lows is a real path.
2. Margin was the June worry and it got worse. 5.1% gross margin in a +8% revenue quarter. Some of it is one-off (liquidated damages, launch costs), but fixed-price contracts against rising cell prices are structural.
3. Credibility. Guidance cut four times in two fiscal years. The market will not credit the FY27 plan until a quarter prints clean.

Contrarian view. The June contrarian line was "backlog is cancellable and low-margin." The new one runs the other way. The market has swung from pricing flawless execution to pricing failure. The number most likely to surprise is cash, because a heavy Q4 of deliveries plus customer prepayments can release working capital quickly. That flips if Q4 deliveries slip again.

Devil's Advocate — re-run, against the bull case above

• The "fixable bottleneck" story is what management said in May and August. Both times the next update was worse.
• The ~$850M of data-center business includes ~$550M of "awards," not purchase orders. Treat it as pipeline until converted.
• Owners selling 23M shares in May, just before the collapse, while their own executives now run operations: at best neutral, at worst a signal about how they see the equity.
• If FY27 revenue is only ~$2.8B with gross margin under 8%, Fluence still burns cash, and the equity is the funding source. The worst case is a raise at $5–6, not a cut to $3; that is roughly the bear-case targets.
• Dual-class, related-party contracts and an unresolved formal SEC investigation are all unchanged.

Management Questions — carried, re-ordered

See the previous dossier, Lens 14. The top five now:

1. What exactly will FY27 operating cash flow be under the FY27 plan, and at what point would you raise equity rather than draw supply-chain finance?
2. Houston: current daily output versus the 15 GWh/yr nameplate, and the date it runs on grid power at full rate?
3. How many of the ~$550M hyperscaler awards have become firm purchase orders, and on what terms?
4. How much of the FY26 gross-margin shortfall was catch-up on fixed-price contracts, and how much FY27 backlog is priced before the cell-price rise?
5. Why did the August call's view of Q4 ("roughly half… already produced") miss by ~$0.6B in five weeks, and what has changed in planning since?