Ouster, Inc. (OUST)

The highest-unit-volume Western digital-lidar pure-play, repositioning from automotive supplier to a physical-AI sensing platform — revenue is inflecting but the balance sheet is thin.

Investment Research · Photoncap-style deep dive · v1 of "Ouster" · May 14, 2026

What Ouster physically does

Ouster builds digital lidar — laser sensors that measure distance by timing how long light pulses take to return, then assemble those returns into a 3D point cloud the host system uses to perceive its surroundings. The distinguishing word is "digital." Where legacy lidar uses discrete analog components — individual lasers, individual photodetectors, hand-aligned optics — Ouster's architecture puts the laser array and the detector array onto two custom CMOS chips. The light source is a VCSEL (vertical-cavity surface-emitting laser) array; the receiver is a SPAD (single-photon avalanche diode) array. Because the expensive, fiddly parts are silicon, the sensor rides the semiconductor cost-and-scaling curve rather than the optics-assembly curve. That is the entire investment premise: a sensor whose bill of materials and manufacturability improve like a chip, not like a camera module.

Functionally, the core product is the OS series — a family of spinning sensors (OS0 short-range wide-field, OS1 mid-range, OS2 long-range) that rotate to give a 360-degree field of view, plus the DF series of solid-state digital flash sensors aimed at automotive forward-facing applications. Each sensor is built around an Ouster-designed system-on-chip; the current generation is the L3 chip, and the just-launched Rev8 generation runs on next-generation "L4 Ouster Silicon." A lidar sensor is the binding constraint for any robot or vehicle that has to operate in unstructured 3D space in the dark, in glare, or where a camera's depth estimate is too noisy to trust — warehouses, mining sites, construction equipment, last-mile delivery robots, and the smart-infrastructure cameras that meter traffic and count people. Ouster's bet is that "physical AI" — the current industry shorthand for robots and autonomous machines that act in the real world — needs a cheap, reliable, mass-manufacturable depth sensor, and that digital lidar is it.

Ouster reached its current scale partly through consolidation. It merged with Velodyne in 2023, absorbing the company that effectively invented automotive spinning lidar, and in early 2026 it folded in Stereolabs, a stereo-camera and spatial-perception company. The Stereolabs deal matters because it moves Ouster from "we sell you a lidar" toward "we sell you the perception stack" — lidar plus cameras plus the software that fuses them — which is a higher-value, stickier position if it executes.

Product roadmap

The legacy backbone is the OS sensor family — OS0, OS1, OS2 — which has shipped in successive revisions for years and remains the volume product, cumulatively well past 12,600 units in a single quarter as of Q1 2026. The DF series of solid-state digital flash sensors is the automotive forward-facing line, positioned against solid-state competitors but a smaller revenue contributor than the spinning OS family.

The headline 2026 event is the Rev8 OS family, released May 4, 2026 and billed by Ouster as "the world's first native color lidar" — sensors that capture per-point color alongside range, eliminating a separate camera-to-lidar calibration step. Rev8 runs on the next-generation L4 Ouster Silicon and Ouster claims up to double the range and resolution of the prior generation, with an explicit emphasis on functional safety and automotive-grade reliability. Crucially for this theme, Ouster announced on May 5, 2026 that Rev8 is being brought to "the robotics and edge AI ecosystem" with native integration across the NVIDIA Jetson platform — the dominant compute module for autonomous mobile robots, drones, and cobots — and on May 12, 2026 that Rev8 has been qualified to run on NVIDIA's DRIVE Hyperion platform for L4 autonomous-vehicle development. The NVIDIA tie-ins are the cleanest evidence that the robotics pivot is real and not just marketing.

What Ouster does not make: it does not make FMCW (frequency-modulated continuous-wave) lidar — the coherent, velocity-sensing approach that Aeva sells — nor does it make 4D imaging radar. It also does not make the host compute or the perception software's final decision layer; with Stereolabs it now sells more of the perception stack, but it remains a sensor-and-middleware company, not an autonomy company.

The financial print

Ouster reported Q1 2026 results on May 5, 2026. Revenue was $48.6 million, up 49% year over year, with product revenue of $48.2 million on more than 12,600 lidar and camera units shipped. GAAP gross margin was 43%, down sharply from 60% in Q4 2025 — a swing management attributed to mix and the integration of lower-margin Stereolabs camera revenue, and a number to watch closely. The company posted a GAAP net loss of $17.5 million; adjusted EBITDA was a loss of $6.9 million, an improvement year over year but still not break-even. Management guided Q2 2026 revenue to $49.5–$52.5 million, including a full quarter of Stereolabs. Cash, cash equivalents and restricted cash stood at $80.5 million as of March 31, 2026 — a real number but not a large one for a company still burning, which makes the path to adjusted-EBITDA break-even the central financial question. There is no meaningful forward P/E; the company is loss-making, so the valuation rests on revenue growth and the gross-margin trajectory. The roughly $2.1 billion market cap implies the market is paying about 10x a run-rate revenue of ~$200 million for a sensor company that is not yet profitable — a momentum multiple, not a value one. The next binary is Q2 2026 earnings, expected on or around August 5, 2026.

Customer mix today

Ouster's customer base is unusually fragmented for a company this size, which is both a strength (no single-customer cliff) and a weakness (no anchor program). Management does not break revenue out by named customer the way a single-OEM supplier would, and as of Q1 2026 reporting no single customer is disclosed as more than ~10% of revenue. By vertical, the structural shift is the story: in earlier years Ouster's narrative was automotive-and-trucking heavy, leaning on the Velodyne legacy; by Q1 2026 management explicitly called out smart infrastructure and industrial automation as the key demand drivers of the quarter, citing new million-dollar contracts for the Ouster BlueCity smart-intersection product and several million-dollar industrial-automation deals. Robotics and edge AI is the fastest-growing slice — management flagged "strong demand from companies building foundational AI models and advanced robotics platforms" — but it is growing off a small base and is not yet separately quantified. Automotive is now the long-tail vertical rather than the headline. The Rev8 launch customer list Ouster published on May 5, 2026 is the best available proxy for direction of travel: it names Google, Volvo Autonomous Solutions, Liebherr, Epiroc, Field AI, Flyability, Skydio, Plus, Seegrid, Gecko Robotics, Burro, Third Wave Automation and roughly a dozen others — a mix heavily weighted toward robotics, mining, construction and warehouse automation, with autonomous passenger vehicles a minority. Treat that list as design-in intent, not booked revenue.

What's actually happening at the robotics / NVIDIA end-market

The most concrete development for the robotics thesis is Ouster's two-step integration with NVIDIA. On May 5, 2026 Ouster announced that the Rev8 OS family is natively integrated across the NVIDIA Jetson platform — the Jetson Orin and Thor modules are the de facto compute standard for autonomous mobile robots, inspection drones and cobots, so being a plug-and-play sensor in that ecosystem lowers the integration cost for every robotics OEM building on Jetson. On May 12, 2026 Ouster added that Rev8 is qualified on NVIDIA DRIVE Hyperion, NVIDIA's reference platform for L4 autonomous-vehicle development, with integration into NVIDIA DriveWorks. The mechanism that matters: NVIDIA does not pick a sensor lightly, and "qualified on the reference platform" means every robotaxi or AV developer prototyping on Hyperion can specify Ouster without a custom bring-up. The named robotics customers — Field AI (foundation models for robots), Skydio (autonomous drones), Gecko Robotics (industrial inspection robots), Seegrid and Third Wave Automation (warehouse AMRs and forklifts), Burro (agricultural robots) — are exactly the early-cycle physical-AI cohort the theme is built on. The honest caveat: these are design-ins and ecosystem placements, not multi-year production POs with disclosed unit volumes, and the robotics revenue line is still small enough that Ouster does not break it out.

The competitive threat / Hesai and RoboSense

The terminal competitive fact about lidar is that the volume and the cost curve now sit in China. Per industry tallies cited at CES 2026, Chinese players account for roughly 60% of automotive lidar revenue; Hesai shipped over 1.2 million units in 2025, reported full-year non-GAAP profitability, and told reporters at CES 2026 it would double production again on "accelerating demand" in automotive and robotics, with lidar now in roughly 25% of new EVs sold in China. RoboSense delivered on a similar order of magnitude. Against that, Ouster's ~$200 million revenue run-rate and 12,600 units a quarter is a niche position. The bull rebuttal is that Hesai and RoboSense are automotive-volume players whose Western robotics and infrastructure penetration is limited by procurement caution and, increasingly, by US policy friction around Chinese sensors in critical infrastructure and defense-adjacent applications — which is precisely where Ouster's BlueCity and industrial business sits. On the Western side, the competitive field has actually thinned: Luminar filed for Chapter 11 and is winding down or selling its lidar business, removing a long-time bear-case comparison and a money-losing price competitor. Aeva (AEVA) competes on a different technology (FMCW) and a different segment (long-range automotive and industrial), so it is more an adjacent name than a head-to-head rival. The net read: Ouster is not the cost leader and never will be, so its survival case rests on owning the Western robotics/infrastructure niche where Chinese supply is structurally disadvantaged.

The terminal risk

Two structural transitions constrain the multiple. The first is the Chinese cost curve: if Hesai and RoboSense drive automotive-grade lidar ASPs low enough fast enough, and if Western procurement caution erodes, Ouster's addressable margin pool shrinks regardless of unit growth — you can win the robotics niche and still not earn an acceptable return on it. The second is sensor-architecture displacement: a meaningful camp in robotics argues that camera-only or camera-plus-radar perception, with enough neural-net compute, makes lidar optional for many indoor and structured-environment robots — the same "vision-only" argument Tesla made in automotive. If that view wins in warehouse AMRs and cobots, Ouster's fastest-growing vertical is also its most exposed. Ouster's hedge against both is real but unproven: the digital-silicon architecture is the credible answer to the cost curve (it scales like a chip), and the Stereolabs acquisition plus native-color Rev8 is the answer to vision-only (sell the fused lidar-plus-camera stack rather than fighting cameras). Whether those hedges are enough is exactly what the next two years of gross margin and robotics revenue disclosure will reveal.

Bull / Gap / Optionality (Photoncap framing)

Bull

1. Revenue is genuinely inflecting, not just guided to inflect. Q1 2026 revenue of $48.6 million was up 49% year over year, and Q2 guidance of $49.5–$52.5 million keeps the run-rate above $200 million annualized. For a company that spent years stuck near break-even revenue scale, two consecutive quarters of ~50% growth driven by infrastructure and industrial demand — not a single fragile OEM program — is the most important fact in the file.

2. The NVIDIA double-integration is a real moat-builder. Native Jetson integration (May 5, 2026) plus DRIVE Hyperion qualification (May 12, 2026) means Ouster is the default depth sensor in the two NVIDIA ecosystems that matter for robotics and AV development. Every Jetson-based robot OEM and every Hyperion-based AV developer can now specify Ouster without custom bring-up — that lowers Ouster's customer-acquisition cost across the entire physical-AI cohort.

3. The Western competitive field is thinning in Ouster's favor. Luminar's Chapter 11 removes a chronic Western price competitor and a bear-case comparison. With Hesai and RoboSense facing procurement and policy friction in US infrastructure, industrial and defense-adjacent applications, Ouster has a cleaner run at the exact niche — BlueCity smart infrastructure, mining, construction, warehouse robotics — where it is winning million-dollar contracts today.

4. Rev8 plus Stereolabs moves Ouster up the value stack. Native-color lidar eliminates a calibration step; the Stereolabs acquisition adds stereo cameras and fusion software. Together they let Ouster sell a perception bundle rather than a commodity sensor — a stickier, higher-ASP position if the integration delivers, and a direct answer to the "vision-only" bear case.

5. No customer-concentration cliff. With no disclosed customer above ~10% of revenue and demand spread across infrastructure, industrial, robotics and automotive, Ouster lacks the single-program risk that sinks most small sensor companies. The flip side is in the Gap section — but the diversification itself is a genuine durability feature.

Gap

1. The gross-margin reversal is alarming and unexplained-away. GAAP gross margin fell from 60% in Q4 2025 to 43% in Q1 2026. Management attributed it to mix and Stereolabs integration, but a 17-point swing in one quarter undermines the entire "scales like a chip" thesis if it is not promptly reversed. Until margin re-expands, the digital-lidar cost-curve story is a claim, not a result.

2. The balance sheet is thin for a company still burning. Cash of $80.5 million against an adjusted-EBITDA loss of $6.9 million per quarter and a GAAP net loss of $17.5 million is a runway measured in quarters, not years, on GAAP terms. Any stumble in the revenue ramp or the margin recovery raises the specter of a dilutive raise at a momentum-inflated share count.

3. Robotics revenue is real in narrative but invisible in the financials. Management talks up robotics and edge AI as the fastest-growing vertical, but does not quantify it. The investment case for this name within the Robotics theme is therefore largely option value — design-ins and ecosystem placements that have not yet shown up as a disclosed, growing revenue line.

4. Ouster will never be the cost leader. Hesai shipped 1.2 million units in 2025 and is profitable; RoboSense is at similar scale. Ouster ships ~12,600 a quarter. The structural cost and scale advantage sits in China, and Ouster's entire survival case depends on a policy-and-procurement moat around its Western niche holding — a moat that is real today but not guaranteed to last.

Optionality

The trade

Ouster is a Bucket-B, early-cycle option on Western digital lidar becoming the default depth sensor for physical AI — a real revenue inflection wrapped around a thin balance sheet and a single ugly margin print. Initiate at $31.60–$34.90 (current price ± ~5%, the appropriate construction for a name that just gapped up on the NVIDIA news and is mid-melt-up), size at roughly 1.0% of risk capital — speculative tier, because this is a loss-maker whose robotics revenue is still narrative — and set the stop at $26.50, below the post-Rev8 breakout shelf and the rising 50-day EMA cloud. The named catalyst is Q2 2026 earnings around August 5, 2026, which is the binary on whether the 43% gross margin was a one-quarter Stereolabs-integration artifact or a structural problem; a clean margin recovery plus continued ~50% revenue growth would re-rate the name, a second soft margin print would break the thesis. If you want the same physical-AI-sensing exposure with a fortress balance sheet and actual profits, Mobileye (MBLY) is the cleaner expression — slower-growth but self-funding — and serious lidar-volume bulls should note Hesai is the profitable scale player, albeit with China-risk. Conviction: 6 / 10.

Sources referenced inline throughout. Reference v1 of this template format: _Watchlist/hanmi-photoncap-style.md.

\newpage

AEVA — Aeva Technologies, Inc. · WATCH (Tier-2) · Conv 5/10 · Bucket B

ticker: AEVA name: Aeva Technologies, Inc. theme: Robotics bucket: B conviction: 5 entryzonelo: 20.16 entryzonehi: 22.28 currentprice: 21.22 pricedate: 2026-05-14 positionsizepct: 0.75 stoploss: 16.50 thesisoneline: Differentiated 4D FMCW lidar with marquee automotive and industrial wins, but revenue is tiny, losses are deep, and the robotics case is still option value. catalystnext: Q2 26 earnings catalystdate: 2026-08-06 rsi: 69.1 vs50ma: 40.6 forwardpe: 0.0 themecycleposition: early customermixsummary: NRE-heavy; Daimler Truck the anchor automotive program, a top-10 passenger OEM development award, Nikon/SICK/LMI the industrial leg — no single customer at production-volume revenue yet. terminalriskoneline: FMCW's cost and complexity premium never closes versus time-of-flight lidar at Chinese scale, leaving Aeva a niche performance supplier that cannot reach profitable volume. bulldriverscount: 5 gapriskscount: 4 optionalitycount: 6 lastearningsdate: 2026-05-07 nextearnings_date: 2026-08-06