Eos Had Its Best Quarter Ever. The Stock Is Down 66%.
Record cube shipments, an eighth straight margin improvement, and a backlog that went the wrong way — the widening gap between the factory floor and the tape.
EOSE · Earnings Call · 2026-11-05
An operating quarter the tape refuses to underwrite
Eos shipped more batteries in the September quarter than in any quarter in its history, posted its eighth consecutive quarter of gross-margin improvement, and cut unit costs across every major line. It also watched its equity fall 66% over the prior ninety trading days, a seventeen-week slide from a $5.55 close on July 1 to the low of the range. That contradiction is the entire story. The factory is finally behaving; the balance sheet and the customer mix are what investors are now trading. The headline numbers were genuinely good on a sequential basis. “Revenue closed at $82.7 million. That's up 171% year-over-year and 20% sequentially.” — Alessandro Lagi, Chief Financial Officer · 2026-11-05 Shipments rose 23% quarter-over-quarter, adjusted EBITDA loss narrowed to $63.8 million with margin improving 27 points sequentially and 96 points year-over-year, and operating expenses were flat against a 20% revenue increase. Against a $69 million quarter in Q2, that is a business line inflecting. But the tape is not arguing about the quarter. A $680 million market cap with a 90-day drawdown of 66% and a whole-history drawdown of 94% from the 2021 peak tells you the market has moved on from "can they build it" to "who buys it and with whose money."Where the cost-out is real
Strip out the noise and the most defensible thing Eos said this quarter was about cost per cube. Material cost per cube fell 9% sequentially, direct labor per cube fell 17%, and manufacturing overhead per cube fell 8% — driven by higher facility utilization and better fixed-cost absorption. Management explicitly called the material savings structural rather than one-time, and the conversion cost step-down as Thorn Hill absorbs volume is the mechanism behind the Q4 guide of negative 25% to negative 20% adjusted gross margin, a further 19 to 24 points of improvement. The physical evidence came from operations, not the finance deck. “On average, Thorn Hill is producing about 20% more batteries per shift than Turtle Creek.” — John Mahaz, Chief Operating Officer · 2026-11-05 Line 2 reached the same first-pass yield as Line 1 in three months, against 13 months for the bipolar line and 26 months for the battery line at the older site. That is the sort of learning-curve compression that makes a second site worth building — and it is why the consolidation into Thornhill is being accelerated rather than deferred. One year ago, on the August call, the CEO pre-empted the obvious objection: “This is not about "Oh Turtle Creek doesn't work." It's a story of Thorn Hill works better.” — Joseph Mastrangelo, Chief Executive Officer · 2026-08-07Three things the market is actually pricing
One: the backlog went backwards. It is the sentence management put third in the prepared remarks, which is itself telling.Maybe. But for a company whose entire equity story is a scaling curve, a sequential backlog decline in the quarter you ship your most cubes is exactly the wrong sign, and it lands against a stock already in freefall. Two: the counterparty concentration has not been solved. Related-party revenue was 83% of the quarter, and Frontier Power USA remains roughly 43% of backlog. An analyst asked the same question in August — “how should we think about that from a nonaffiliated entity in the backlog” — Stephen Gengaro, Analyst · 2026-08-07 — and the answer then was aspirational: “What I'd like to see is less percentages change and more of the size of the pie grow.” — Joseph Mastrangelo, Chief Executive Officer · 2026-08-07 The pie did not grow this quarter. Six orders closed with four new customers, which is progress at the margin, but 83% is 83%. Three: cash. Total cash of $344 million including $271 million unrestricted, down $20 million sequentially after $106 million of net cash use offset by $87 million of DOE reimbursement. The effective net cash position of -$584 million is the number that explains a 66% drawdown better than any margin line. Management is leaning hard on the story that operating cash now tracks operating results — and to be fair, working capital was essentially flat on a 20% sequential revenue increase, which is real discipline. But the keyword set this quarter flagged cash burning in the top five themes of the transcript for a reason. And then there was NYSERDA. Management volunteered that no Eos projects were selected in the first round of New York's bulk storage solicitation, a round that awarded less than a third of the program's 3-gigawatt target. The pivot to "other states' programs are larger" may be true, but the NYSERDA miss is a concrete data point where the prior calls offered only pipeline language.We ended the quarter at $765 million, about 3.2 gigawatt hours. That's up 19% from a year ago and down from $807 million in Q2. We shipped more than we booked last quarter. This is project timing, not a fundamental change in demand.
The promise that keeps recurring, and the one thing that is genuinely new
Here is the tell. Eos is working 3.5 gigawatt hours of late stage opportunity — more than its entire backlog — with more than half of it waiting on financing. That framing has been in the company's keyword set before, and it was the same framing a year ago when a global utility was "testing in the United States." The conversion rate, not the opportunity count, is the honest scoreboard, and the scoreboard moved the wrong way this quarter. What is new is the Google/MN8 ten-hour project in West Virginia. A named hyperscaler, through a named developer, validating long-duration zinc is a different category of proof than a national lab test.The problem is timing: the AI data centers cluster in the global tape is one of the worst-performing themes over both 90 and 30 days, with 57 declining constituents against 7 advancing on the 90-day screen. Eos is leaning into a narrative the market has already begun to de-rate — which may be the opportunity, or may simply be a lag. One smaller thing worth flagging: warranty costs surfaced as a fresh theme, and the CEO pre-empted it directly — “our warranty reserve goes up with volume doesn't mean there's warranty costs underneath it” — Joseph Mastrangelo, Chief Executive Officer · 2026-11-05 — noting that utilization of the reserve has not changed. That is the right answer to a question investors were clearly asking. It is also the kind of question that gets asked when a fleet of first-generation systems is finally running at scale.Out of that 3.5 gigawatt of near-term opportunities, we're going to see orders closing that meet this use case.