Open in interactive viewer → charts, metric popovers & call review

SES AI Ships Its First Real Drone Cells — and a 4.9% Gross Margin

The Korean factory finally converts a four-quarter promise into purchase orders, but clearing obsolete ESS inventory breaks the margin story management told just three months ago.
SES · Earnings Call · 2026-11-12

SES AI: The Flywheel Turns, the Gross Margin Doesn't

For four straight quarters, SES AI has told the same story: a Korean factory built for GM's lithium-metal dreams gets retooled into one of the only NDAA-compliant drone-cell plants outside China, and revenue follows. This quarter the line is actually running, the orders are real, and the income statement took the bill for getting there. Revenue rose 31% sequentially to $6.6M, but gross margin collapsed to 4.9% from 22.6% — the cost of pushing the last of an obsolete 5-kilowatt-hour ESS product out the door before a 16-kilowatt-hour replacement ships into 2027.

Our GAAP gross margin was 4.9% in the quarter, compared to 22.6% in the second quarter, primarily due to lower margin sales of older ESS hardware inventory due to line and product upgrades.

Yi Liu, Chief Financial Officer · 2026-11-12

That is a defensible decision on its own. It lands badly against what the same management team promised three months ago.

The margin promise that broke

In August, an analyst asked directly about gross-margin improvement. The answer was unambiguous: “as the drone revenue picks up, we'll see margin increase because a drone sale typically carries a higher margin compared to the ESS business. So we'll see margin improvements in the second half.” — Yi Liu, Chief Financial Officer · 2026-08-11 The second half delivered the opposite. And Q4 does not rescue it: the CFO said the company will keep clearing old ESS inventory at lower prices right through the seasonally strongest quarter of the year. So the guide embeds a near-doubling of revenue into a worse mix.

Nor is the cost side being shown yet. Asked for per-cell cost and yield on the new line, the CEO deferred: “we are going to share those things starting in Q1 next year.” — Qichao Hu, Founder and Chief Executive Officer · 2026-11-12 The first quarter of a line's life is by definition its worst economics, so the deferral is understandable — but it is also the quarter investors most need to see.

Fly: from promise to purchase orders

The operating news is genuine, and it is the reason this report is more than a margin miss. Chungju reached an installed capacity of 1 million NDAA-compliant drone cells a year, has been producing since early October, and earned an ISO certification for drone cells. More than 50 drone makers and pack integrators sit in the pipeline; about half a dozen have audited the line. Critically, orders have started arriving from winners of the Drone Dominance program's Gauntlet Challenge — “we have received orders from several of the participants and winners of the Drone Dominance program's Gauntlet Challenge.” — Qichao Hu, Founder and Chief Executive Officer · 2026-11-12 The volumes are no longer sample-sized: “We are talking about in the quantity of several thousand and then to tens of thousands.” — Qichao Hu, Founder and Chief Executive Officer · 2026-11-12 Roughly $5M of incremental capex takes the plant to 3 million cells a year by the end of Q2 2027, with a 2027 delivery target of “at least 1 million to probably 1.5 million qualified cells to customers.” — Qichao Hu, Founder and Chief Executive Officer · 2026-11-12

But the sequencing has slipped, repeatedly. In April the CEO said “we expect revenue in Q2 for the NDA-compliant cells and then to start to pick up in Q3 and then Q4.” — Qichao Hu, Founder and Chief Executive Officer · 2026-04-23 In August, customers were “waiting for that 1 million cell capacity to be operational in September.” — Qichao Hu, Founder and Chief Executive Officer · 2026-08-11 It ran in October. Meaningful drone revenue is now a Q4 event — for the third consecutive quarter. It is fair to say this time the mass production purchase orders exist and the Chungju production line is physically producing; that is the difference between a promise and a shipment. It is not yet the difference between a shipment and a business.

The mix tells the same story. Energy storage was 47% of the quarter, with the balance from materials and, to a lesser extent, drone cell samples. The line receiving the incremental capital is still the smallest revenue contributor — the flywheel is an elegant metaphor, but Charge and the electrolyte business still pay the bills. And the allied drone supply-chain theme is a cross-market wave, not a SES invention: the Defence Drone Initiative surfaced across other recent reporters in this cycle. Useful corroboration, though the public tape on defense-drone names has been weak for months, so the theme is live without being hot.

Discover, and the lab that pays for itself

Molecular Universe 4.0 — branded M-Labs — is the genuinely differentiated asset: a closed-loop system integrating simulated and experimental data, running fully on-premise and air-gapped at customer sites, now in evaluation with what management calls two of the largest battery makers in the world. The security argument is the whole product: no battery maker will let proprietary formulation data train someone else's model. The problem is that this revenue has been "in the coming quarters" since 2025 — first as SaaS tiers, then as the Hisun JV electrolyte materials business. Materials is now evidently the second-largest revenue bucket, which is progress, but it arrives with no disclosure of contribution, and the first M-Labs system sale is still ahead.

The cost structure shows the pivot clearly. R&D of $11M, down from a $21M peak through the last filed quarter, while R&D stands at just 1.2x SG&A. That is capital discipline, and it is also a company that has stopped buying optionality in the lab to fund the plant.

Balance sheet, buyback, and the listing

SES ended the quarter with roughly $137M of liquidity, no debt, and an operating loss of $19M in the last filed quarter. Strip out the buyback and net cash used was $16.5M — call it eight quarters of runway if nothing changes. Which makes the capital allocation argument the most interesting tension in the release: $9.2M was spent repurchasing 16.2M shares at $0.57, a 4.9%-margin company burning $16.5M a quarter also filed a proxy for a reverse split to protect its NYSE listing, and for the first time management is talking about tax credit capture by evolving part of manufacturing into the US.

The tape is not sympathetic. The stock is down 94.6% from its 2021 peak, off 27% over 90 days, and after spiking to $1.00 on October 13 gave back 42% into this print. A sub-$1 share price plus a reverse split plus a buyback is not a confidence signal the market is buying; it is a company managing a listing while it manages a pivot.

What has to be true by spring

Q4 revenue of $11.6M to $16.6M would be nearly double Q3 — the largest quarter in the company's history by a wide margin, built on ESS seasonality and the first purchase orders that are more than samples. Then unit economics in Q1, the first M-Labs system sale, and repeat orders from the Gauntlet cohort. If Q4 lands and the first quarter of 2027 shows drone gross margin above the corporate average, the flywheel is real. If it shows 4.9% again, then this was not an inventory flush but the actual economics of a company trying to fund a factory, a lab, and a buyback out of $137M.