Robotics
A genuine FMCW technology lead and a watershed exclusive-OEM win, priced at ~28x forward sales against ~24 months of runway and a 2028 start-of-production — the technology is real, the chasm between here and revenue is the trade.
Research
The verdict
A genuine FMCW technology lead and a watershed exclusive-OEM win, priced at ~28x forward sales against ~24 months of runway and a 2028 start-of-production — the technology is real, the chasm between here and revenue is the trade.
Aeva designs a 4D LiDAR-on-chip built on Frequency-Modulated Continuous-Wave (FMCW) sensing — the differentiator is that it measures instant velocity for every pixel (Doppler), plus depth, reflectivity and inertial motion, where legacy Time-of-Flight (ToF) LiDAR measures depth/reflectivity only. The architecture integrates the sensing onto silicon photonics (the same tech base proven in telecom), targets ranges up to ~500m, and is immune to crosstalk from other LiDAR and sunlight.
The business model is a classic pre-commercialization deep-tech ramp: today's revenue is product sales (evaluation/pilot sensor units) + non-recurring engineering (NRE) services to customers running R&D and development programs; the company explicitly describes itself as a "development stage company" operating in one reportable segment. The prize is the transition from these low-volume dev programs to series-production design wins that pay per-unit at scale.
Product line: Atlas (high-performance automotive-grade 4D LiDAR), Atlas Ultra (~3x the resolution, passenger-car packaging), Atlas Orion (industrial/short-range), and the newly introduced Omni — a wide-view short-range 4D LiDAR for robotics / "Physical AI", co-designed with LG Innotek.
End markets span automotive (ADAS/AD), industrial automation, consumer electronics, consumer health, and security. Founded Dec 2016/2017 by ex-Apple engineers; HQ Mountain View, CA; Nasdaq Global Select (AEVA / AEVAW warrants).
Key payment-term structure: fixed-price delivery of a specified number of sensing systems under customer agreements, plus NRE arrangements to customize the perception solution — i.e., milestone/development revenue today, not recurring or take-or-pay.
Upstream: Aeva is fabless on the sensor — it designs the LiDAR-on-chip and relies on third-party manufacturers to build production capacity ("we are expanding our manufacturing capacity through third-party manufacturers to meet our customers' anticipated demand"). The core IP is the FMCW silicon-photonics module + custom silicon; inputs are wafers/photonic components from foundry partners (not named in the filing). Inventory is tiny ($6.0M at Mar-2026), consistent with a pre-volume model.
The company → end customer, with named stakeholders:
Chokepoints / single-source dependencies: (1) Aeva itself is the single source of FMCW-on-chip for its customers — the bull's moat is the bear's "single seller" risk. (2) Production depends on un-named third-party foundries/contract manufacturers — execution risk Aeva does not fully control. (3) Customer concentration is the dominant chain risk: FY2025 top-3 customers = 64% of revenue; FY2024 top-2 = 72%; at Mar-2026, 3 customers = 51% of accounts receivable.
Technology moat (real but contested): FMCW + on-chip integration is a genuine architectural differentiator — per-pixel instant velocity, long range, sunlight/crosstalk immunity, and a path to silicon-photonics cost-down at scale. Aeva is the most credible pure-play FMCW name; most rivals ship ToF. The Dec-2025 exclusive passenger-OEM win and the NVIDIA DRIVE Hyperion selection are third-party validations that the tech clears automotive-grade bars.
Switching costs / design-in lock: automotive LiDAR is a multi-year design-in; once a sensor is locked to a series-production platform (Level 3, mid-2030s), it is extremely sticky — this is the source of the bull case's durable cash flows if SOP is reached. The lengthy design-win-to-implementation cycle cuts both ways (Aeva flags it as a risk).
IP: silicon-photonics + FMCW patent estate (patents/ dir on the research shelf is empty — count not sourced here; n/a).
Bargaining power — weak today. As a pre-revenue supplier dependent on a handful of large OEMs/strategics for both revenue and financing (Sylebra, LGIT, Apollo), Aeva needs its counterparties more than they need it. The OEM win improves this materially on a forward basis, but near-term the leverage sits with customers and capital providers. The provision for anticipated losses on the LGIT JDA contract — Aeva expects to lose money fulfilling that development contract — is direct evidence of weak near-term pricing power.
One reportable segment (LiDAR sensing + perception software). So the meaningful split is revenue type and end-market/geography:
Revenue by type — Q1 2026 vs Q1 2025:
| Line | Q1 2026 | Q1 2025 | YoY |
|---|---|---|---|
| Product | $2.427M | $2.481M | −2% (flat/declining) |
| Professional service (NRE) | $3.835M | $0.887M | +332% |
| Total revenue | $6.262M | $3.368M | +86% |
The signal the headline "+90%" hides: product (hardware) revenue is flat-to-down; the entire beat is NRE/engineering-services revenue. That is normal for a company funding development programs, but it means the "record revenue" is not yet evidence of scaling unit sales. Watch for the product line to inflect as Daimler SOP (2026) and the OEM platform (2028) approach.
Annual revenue trajectory (full-year):
| FY | Revenue | YoY | Gross profit/(loss) |
|---|---|---|---|
| FY2023 | $4.312M | — | $(5.886)M |
| FY2024 | $9.065M | +110% | $(3.790)M |
| FY2025 | $18.079M | +99% | $(0.660)M |
Revenue has roughly doubled each year, and gross loss has narrowed from −$5.9M (FY23) to −$0.66M (FY25), with gross margin inflecting positive in Q1 2026 (+$1.94M gross profit, ~31% GM). Geographically, revenue is sourced from customers in North America, Asia and Europe; automotive applications are the primary driver.
The print:
Unusual vs its own history: the positive gross margin and the move to a Top-10 OEM exclusive are step-changes; the negative book equity and the rising SBC are the offsetting deteriorations.
No transcripts/ on the research shelf (transcripts=0); sentiment is from coverage of the calls.
Pure-play Western LiDAR peers. Multiples are `` with source/date; where a clean current multiple is not sourceable I mark n/a rather than fabricate.
| Company | Ticker | Mkt cap (Jun-2026) | EV/Sales | P/E | Div yield | 5-yr avg ROE | Note |
|---|---|---|---|---|---|---|---|
| Aeva | AEVA | ~$1.52B | ~76x TTM / ~28–29x FY26e | n/m (loss-making) | 0% | deeply negative | FMCW pure-play; richest multiple |
| Ouster | OUST | ~$1.23B (Apr-2026) | ~9.6x | n/m | 0% | negative | ToF; diversified "physical AI", bear's preferred name |
| Innoviz | INVZ | ~$161M (Jun-2026) | n/a | n/m | 0% | negative | has secured NREs; cheaper |
| Luminar | LAZR | ~$15M (Jan-2026) | n/a | n/m | 0% | negative | collapsed ~99% from peak; cautionary tale |
| Hesai | HSAI | n/a (Jun-2026) | n/a | profitable | 0% | positive (turned profit) | Chinese leader; ~37% share, 195,818 units Q1'25 |
Read: Aeva trades at a 3–8x premium to Ouster on EV/sales and an enormous premium to Innoviz/Luminar — the market is paying up for the FMCW story + the OEM win. The cleanest scaled, profitable comp (Hesai) is Chinese and structurally cheaper. Bears use the P/S ~44x vs peer-group ~10x gap (and an industry average ~2.7x) as the core overvaluation argument. Aeva's premium is a bet that FMCW + the Level-3 exclusive justify a different multiple regime; that is the crux of the trade.
Pattern: AEVA is a catalyst-and-validation stock — it reacts violently to design wins, marquee-partner endorsements (Daimler, NVIDIA, the OEM), and anchor-investor buying, far more than to the quarterly P&L itself. With short interest ~11.5% of shares / ~17.9% of float, positive catalysts get squeeze amplification.
Accounting risk map:
Regulatory findings (required):
This is a pre-profit story; EPS stays negative across the projection window. The useful forward lens is revenue ramp + cash runway-to-SOP, with EPS as the scoreable output. Anchored to FY2026 guidance ($30–36M) and the disclosed program timeline. All outputs ``; inputs labeled.
Base case:
Bull case: European OEM economics disclosed favorably + a second passenger OEM + NVIDIA-platform design-ins convert → FY2028 revenue $120M+, gross margin clears 40%, the market re-rates on a credible path to cash-flow breakeven by ~2029–2030. EPS still negative but the multiple expands on de-risked SOP.
Bear case: OEM SOP slips from 2028 (LiDAR programs routinely slip), product revenue stays NRE-dependent, and the company must draw the SEPA + raise again into a weak tape → 30–50% dilution before meaningful unit revenue, EPS stays near $(2) on a larger share count, multiple compresses toward Ouster's ~10x. This is the "liquidity chasm" scenario.
Runway-to-catalyst (the lens that matters): $99.5M liquidity (Mar-2026) + undrawn $125M SEPA (all draw conditions met, available to Nov-2026) against ~$115M/yr burn = roughly ~10–12 months on cash alone, extendable to ~24 months with the SEPA, versus a 2028 European-OEM SOP. Management asserts ">12 months" runway. The structural truth: existing capital does not reach 2028 SOP — at least one more meaningful raise (dilutive) is near-certain before the flagship program generates revenue. That gap is the bear thesis and the dominant risk.
(No forecast.ts create run — --watchlist unattended mode.)
Bull case. Aeva owns the most differentiated architecture in automotive LiDAR (FMCW per-pixel velocity, on-chip), and in the span of six months it converted that into the validations that matter most: an exclusive Top-10 European OEM Level-3 series-production platform (mid-2030s tail), NVIDIA DRIVE Hyperion selection, Daimler Truck L4, and SICK industrial. Revenue has doubled three years running, gross margin just turned positive, R&D spend is falling (discipline), and a committed anchor (Sylebra, 25%) plus LGIT and Apollo keep the lights on. If even one flagship program reaches scaled SOP, the design-in lock-up produces years of sticky, high-margin unit revenue that no current multiple captures. Secular tailwinds (UNECE R157, China NCAP 2026 pushing L3) expand the TAM ($960M → ~$6.5B by 2033, 31% CAGR; FMCW the fastest sub-segment at ~49% CAGR).
Bear case (2–3 permanent-impairment risks). (1) The liquidity chasm: ~24 months of runway vs a 2028 SOP guarantees a dilutive raise into the gap; if the tape is weak, dilution is brutal and permanent. (2) Single-architecture, single-customer concentration: 64% of revenue from 3 customers, the whole equity story riding on one undisclosed OEM whose economics are unknown and whose program could slip or be cancelled (LiDAR history is a graveyard of slipped programs). (3) Chinese cost competition: Hesai/RoboSense already ship at scale, profitably, at ToF price points ($200–500/unit target) — if "good-enough" ToF wins the volume L2/L3 market, FMCW's premium becomes a niche, not the standard. Expectations baked into the price (~28x FY26 sales, ~76x TTM) leave no margin for delay.
Pre-mortem (18 months out, thesis broke). The European OEM disclosed underwhelming volumes/economics (or pushed SOP to 2029), product revenue stayed NRE-dependent, Aeva drew the SEPA and did a follow-on at a lower price, the short squeeze unwound, and the stock re-rated from ~$15 toward the bear's $7–13 fair-value zone as the "story premium" evaporated.
Are multiples too high? On any near-term fundamental, yes — 28x forward sales for a company that won't be cash-flow positive before ~2029 is a pure optionality multiple. It is defensible only as a call option on FMCW becoming the L3+ standard.
Contrarian view (what the market refuses to see). Bulls under-weight that the entire +90% "record" quarter was engineering services, not hardware units — the thing that has to scale (product revenue) was flat. Bears under-weight that an exclusive global Level-3 platform at a Top-10 OEM is a genuinely rare asset that, if real, is mispriced low at $1.5B. Both sides are arguing about the same unknown: the undisclosed OEM's volume and economics. The stock is, functionally, a leveraged bet on that one disclosure (promised "early 2026," still pending).
Dismantling the bull case.
The only profitable humanoid maker and #1 by volume — but the float is a Shanghai-only embodied-AI bet whose revenue is 74% lab demos, not labor replacement, and whose Western TAM is being legislated to zero.
A profitable EV maker priced as a solved-autonomy robotics company — the car business is shrinking, the GAAP profit prop (reg credits) is going to zero, and the entire ~190x multiple now rents on robotaxi + Optimus execution that is real but years behind the price.
A genuinely differentiated case-handling robotics moat wrapped around an ungovernable accounting-and-concentration core — adverse ICFR opinion + active SEC whistleblower probe + >90% Walmart revenue means the multiple is pricing a clean compounder that the filings say does not yet exist.