Thesis. Record $1.44B of orders and a $6.4B backlog including ~$850M of datacentre business — and it still cut guidance, because it cannot manufacture fast enough. The cleanest case of real demand meeting failed execution on the watchlist.
Reviewed 19 Sep 2026, after the most recent quarterly results and the July datacentre awards. Next review after Q4 FY2026 earnings on 24 Nov 2026.
What it does
Fluence builds grid-scale and behind-the-meter battery energy storage systems — containers of batteries that store electricity and release it later. It is an integrator: it buys cells from others and assembles complete systems. That distinction sets the economics, because an integrator's margin is the spread between cells it does not make and systems it does not operate, and it is thin by design.
For datacentres the appeal is the same as fuel cells at BE: power on site, without waiting in an interconnection queue — here by buffering rather than generating.
Orders nearly tripled and the company cut its own revenue guidance by $400M in the same breath. When a business fails on a quarter like that, the problem is not the market.
Bull case
- Record order intake $1.44B — nearly triple the prior year. Total orders +80% year to date
- Record backlog $6.4B, +14% sequentially and +30% YoY, against a total pipeline of $33.1B. $2.2B is scheduled to convert to revenue in FY2027
- Datacentre business ~$850M secured through July 2026 — a $300M behind-the-meter deal with a developer plus $550M of awards from a hyperscaler in July
- Datacentre pipeline 16 GWh, +35% sequentially, on a fast three-month lead-to-order cycle
- SmartStack is 75% of year-to-date orders; SmartStack 10 raises density to 10 MWh per unit from 7.5
- ARR guidance of $180M maintained
Bear case
- GAAP gross margin collapsed to 5.1% from 14.8% (adjusted 5.9% from 15.4%). Net loss $44.3M against $6.9M of net income a year earlier; adjusted EBITDA −$29.3M against +$27.4M
- FY26 guidance cut: revenue to $2.9–3.1B, down $400M at the midpoint, and adjusted EBITDA to −$30M to +$10M from $40–60M. The stated cause is production delays at new contract manufacturing facilities
- It needs $300–500M of additional working capital within a year to support the order intake. Liquidity $863M with cash around $365M — the gap is real, and it will be closed with debt or dilution
- Stock fell 24.2% on the print and sits ~57% below its 52-week high of $33.51
- Competes against Chinese cell suppliers with structural cost advantages, into a business model where integrator margins are already thin
- Customer concentration is not disclosed, and the two datacentre counterparties are unnamed
Major customers
- Utilities and independent power producers — the core grid-scale storage base
- An unnamed hyperscaler — $550M of awards in July 2026
- An unnamed datacentre developer — $300M behind-the-meter deal
- Renewable project developers — solar-plus-storage pairings
Individual customer concentration is not disclosed.
Read-through. Fluence is the cleanest case in this file of demand being real while execution fails — orders tripled and guidance was still cut, because it cannot build fast enough. That is fixable, but it costs money the company does not currently have, which is why the financing question matters more than the order book. Its datacentre business puts it alongside BE, VRT and the grid names serving the same buildout, and its $550M hyperscaler award is independent corroboration of the same demand FPS and ETN are reporting — three companies, three order books, one conclusion. Its cell supply exposes it to the same Chinese competition that pressures EOSE.
What would change the view
- Gross margin recovery from 5.1% — the number that decides this name
- How the $300–500M working capital gap gets funded — debt or dilution, and on what terms
- Whether the manufacturing delays at the new contract facilities resolve
- Backlog conversion, particularly the $2.2B scheduled for FY2027
- Datacentre segment growth beyond the ~$850M secured
- Whether guidance is cut again — a second cut would make this an execution problem rather than a timing one
Update log
19 Sep 2026 — Converted to the standard template. Current through the last print; no new disclosures since. The working capital gap moved up in the bear case, since it, rather than demand, is what determines whether the backlog converts.
Jul 2026 — $550M of datacentre awards from an unnamed hyperscaler, taking datacentre business secured to ~$850M.
Most recent quarter — Record order intake $1.44B (nearly 3x); record backlog $6.4B (+30% YoY); pipeline $33.1B. GAAP gross margin 5.1% (from 14.8%); net loss $44.3M; adjusted EBITDA −$29.3M. FY26 guidance cut to $2.9–3.1B revenue and −$30M to +$10M adjusted EBITDA on contract-manufacturing production delays. Stock −24.2%.
Research and education only — not investment advice.
Made with Bullet