ECARX Promised Memory Costs Would Pass Through. Gross Margin Fell to 7%.
Three months after the CEO said memory inflation would have "no impact" on margin, the bill arrived — and now a $400-million-plus fourth quarter has to carry the whole year.
ECX · Earnings Call · 2026-11-04
Three months ago, the CEO said this couldn't happen
On the August call, Deutsche Bank's Wei Huang pressed ECARX's founder and CEO Ziyu Shen on whether the memory price spiral would eventually hit the P&L. The answer was unusually categorical:
Also principally, I want to say confidently that most of -- I think most of the increase actually pass over to the customer. So no impact on our gross margin, that's for sure.
It was not, in fact, for sure. On Wednesday, CFO Dylan Jeng reported that gross margin had collapsed to “a gross margin of 7% compared to a gross margin of 22% in the same period last year” — Dylan Jeng, Chief Financial Officer · 2026-11-04, with adjusted EBITDA swinging to a $37.7 million loss from a $0.9 million loss, and “approximately $30 million of which came from lower gross margin from memory costs” — Dylan Jeng, Chief Financial Officer · 2026-11-04 versus the prior quarter. Revenue grew 10% year over year to $241 million and shipments rose 31% sequentially to roughly 721,000 units. The top line held. The spread didn't.
The nuance worth keeping: the CFO saw it coming even as the CEO waved it off. Jeng had already conceded in August that margin pressure would bite in the second half. What broke this quarter is the pass-through mechanism itself — ECARX chose to eat cost to hold volume and protect programs, which is a defensible industrial decision and a terrible near-term optics decision.
This is also not a company-specific demand failure. The input shock is market-wide: memory pricing sat among the market's top themes earlier this year, and the 30-day tape now shows the memory-exposed complex selling off hard — "high bandwidth memory" down roughly 36% across fifteen decliners spanning ASM, LRCX, AMAT and Micron. Peers are scrambling for supply rather than pricing: HLIT this week talked about securing memory supply, and Jeng's own remedy list is long-term memory supply agreements. ECARX is riding a genuine global wave — just from the wrong side of it.
The company-specific sin is the forecast. In February, then-CFO Phil Zhou told the same analyst that full-year 2026 gross margin would land in a range “about 15% to 18%” — Wei Huang, Analyst · 2026-02-12. A 7% quarter is not a rounding error against that. Tellingly, the company's own keyword signal has gone generic: memory cost was ECARX's single highest-ranked theme last quarter with the sharpest momentum of any phrase in its vocabulary; this quarter it has been quietly replaced by the flatter cost increase. When the specific risk becomes unnamed, that is usually management fatigue, not resolution.
The quarter that has to be four times normal
ECARX reaffirmed $1.0–1.1 billion of 2026 revenue against a $241 million third quarter. Wei did the arithmetic out loud — an implied fourth quarter of roughly $400–500 million, or a step-up of 66% to 107% sequentially — and asked how much was volume versus ASP, and how many units. He got “we don't provide our quarterly guidance” — Dylan Jeng, Chief Financial Officer · 2026-11-04, plus a reference to program launches and order backlog. Asked about 2027 breakeven, Jeng deferred to a budget process not yet approved by the board.
There is a real argument for the ramp: Peter Cirino noted the model launch cadence was always back-half weighted, with 21 new models across five brands hitting mass production in Q3 and more scheduled for Q4, and the company explicitly asks investors to smooth the lumpiness — “we continue to encourage you to look at the trailing 4 quarters rather than any single one” — Peter W. Cirino, Chief Operating Officer · 2026-11-04. But a company that just watched gross margin fall fifteen points and won't decompose its own recovery guidance is asking for an act of faith, not analysis.
Paying cash for software while software halves
Buried under the memory story is a second, quieter one. ECARX closed the Flyme software business acquisition at the end of September, wiring $191.5 million of consideration funded by cash reserves and $67.8 million of fresh bank financing, with a further $30.5 million drawn for working capital. Total cash fell to $53.7 million from $165.5 million at June 30.
The strategic logic is sound — owning the application layer under Flyme Auto (now in more than 3.5 million vehicles) converts a hardware-tied business into a licensing one. The results, so far, point the other way: software license and services revenue was $18 million, down 53% year over year, attributed to the "seasonal nature of production launches." Compare that with how the company used to frame the line: “it moves up and down by a big percentage every quarter, but it's a very, very small number... generally around, say, $1 million to $2 million a quarter” — Mark Hankinson, Investor Relations / Conference Call Host · 2026-08-11. Software is now a materially larger line that just halved — and because Flyme was a common-control acquisition, prior periods were retroactively restated, obscuring the comparison further.
To management's credit, the convertible note facility was terminated, capping dilution at the cost of a $10 million termination fee and full repayment of the remaining balance. That is shareholder-friendly housekeeping. It also leaves a $309 million market-cap company with roughly $54 million of cash and a fresh bank line as its liquidity cushion heading into its most back-loaded quarter.
Silicon, robots, and a number IR didn't want to give
The most interesting forward-looking thread is the widening of the technology stack beyond the vehicle: an MOU with Arm on next-generation physical AI platforms, a 5-nanometer SE2000 chip in development with SiEngine (which raised $200 million in the first half and in which ECARX remains the largest shareholder), and a joint cockpit Innovation Lab with a European luxury automaker — its first such partner outside the Geely orbit. Cirino's framing: “the same stack can bring AI from the cloud to the edge across a much wider range of intelligent devices” — Peter W. Cirino, Chief Operating Officer · 2026-11-04.
Worth flagging: this "beyond the car" story is an escalation, not a debut. Robotic lawn-mower LiDAR, humanoid robots and non-automotive applications have been asked about and answered on calls going back to 2025. Likewise the overseas target — 30% of revenue by 2028 and 50% by 2030 — has now been repeated essentially verbatim for four quarters, which is what a target becomes when it is not yet a plan.
And then there is the concentration. IR declined to break out Geely, and the CEO promptly did it himself: “usually, Geely revenue right now like 75% and 25% is beyond Geely” — Ziyu Shen, Founder and Chief Executive Officer · 2026-11-04. Against the 40–50% Geely auto plus 30–40% Geely ecosystem disclosure from mid-2025, 75% is a reminder that the globalization story is still mostly a slide.
The bottom line: ECARX executed on volume and shipped more units into a shrinking China market — that is real. But the quarter's defining fact is that a low-margin hardware business absorbed a commodity shock it had publicly promised to pass through, and now must deliver roughly a billion-dollar run-rate quarter with a damaged margin and $54 million of cash. The memory wave is the market's; the forecast that ignored it is the company's.