Seeing Machines ($SEE): Attention Economic Situation of Transport

The straightforward tale

Seeing Machines develops AI-based operator monitoring: driver-monitoring systems (DMS) and wider resident monitoring (OMS) for vehicle; Guardian for business fleets; and pilot monitoring for aeronautics. Profits today is a mix of aristocracies from ingrained auto programs and aftermarket hardware/services from Guardian; the business also invests in aviation and off-road (mining, commercial). Manufacturing quantities of vehicles delivering with Seeing Equipments software program have been compounding, and the company has actually safeguarded Tier- 1 and system cooperations with Qualcomm, Magna, Valeo, and Collins Aerospace, while growing go-to-market in Japan through Mitsubishi Electric Mobility (and a North American press via Mitsubishi Electric Automotive America).

Policy is the demand flywheel: the EU’s General Safety and security Law needs chauffeur drowsiness/attention (DDAW) for brand-new kind approvals already and rolls through the park with Advanced Driver Interruption Warning (ADDW) requireds, tightening again in July 2026 for all new car registrations, which indicates a multi-year retrofit of product plans and fitment ramp across mainstream OEMs. In the united state, NHTSA has several, adjacent tasks– rulemaking intros on damaged driving prevention technology and useful safety research for in-cabin DMS– signifying a regulatory trend, not a one-off wave

The most recent KPI disclosures show the ramp is actual, not hypothetical. For FY 2025, auto quarterly manufacturing using Seeing Makers software application struck ~ 488 k in Q 4 (up 36 % sequentially), bringing the FY overall to > > 1 5 million units (up ~ 35 % vs FY 2024, with Guardian Gen 3 getting in complete production and accelerating system sales in fleets. Before that, monitoring had actually flagged 2 2 + million cars and trucks when driving since June 30, 2024, minimal volume warranties starting 2025 (with material impact in FY 2026, and the ramp of a VW program late 2025– all constant with regulatory “pull-through” and growing supply chain dedications. Monetarily, the FY 2024 statements reveal OEM revenue stability through a hard auto background, while Guardian connections and equipment sales increased (12 % and 30 %, respectively), and collective set up cars increased year-over-year right into mid- 2024– proof that royalty return each isn’t the only lever; set up base and connection density matter more for long-run business economics.

That’s the “seen” picture. The deeper thesis starts where most DMS write-ups stop.

The real asset: an “attention layer” for autonomy’s messy middle

The biggest misconception in this room is treating DMS as a conformity checkbox. It isn’t. It’s a safety and security primitive for the untidy middle in between human driving and full freedom. 2 forces make this unavoidable:

1. The responsibility bridge for L 2 + and conditional L 3: As automakers deliver more ADAS features (hands-off, eyes-on; automated lane adjustments; traffic jam pilots), the system has to constantly gauge that is in the loophole and just how they’re doing. Not in a common method, but with robust category of look, eyelid dynamics, head pose, cognitive work proxies, and context (e.g., infotainment communication or phone usage). The point is not merely to ding a driver– it’s to arbitrate control. Without a relied on dimension of vehicle driver state, L 2 + functions end up being uninsurable at scale. Seeing Machines’ enduring research lineage in “driver state” evaluation and its deployment at production volumes places it as a control-relevance provider, not a warning light vendor.
2. Regulatory cog + platform assimilation: The EU’s stepped requireds (DDAW to ADDW) are the classic regulative “ratchet”– when in place, they seldom reverse; they tighten with each collision data cycle. In parallel, system silicon settles the cockpit/ADAS compute domain (e.g., Qualcomm Snapdragon Ride/Cockpit combination). That suggests the expense of not incorporating high-quality DMS/OMS climbs gradually due to the fact that the “port” inside a combined software pile ends up being limited real estate. Seeing Machines’ very early cooperations with Qualcomm produce optionality: the a lot more the pile is standard, the more pull-through you get from being the de facto DMS/OMS module on that stack.

If DMS is security plumbing, then the business’s worth hinges on two more challenging inquiries: (i) exactly how strong is the data/AI moat; (ii) how much utilize does assimilation give them vs. Tier- 1 s who possess hardware and system contracts? That’s where mirror business economics, Valeo/Magna, and Guardian issue.

Mirror economics and the “every-trim” unlock

A subtle yet crucial architectural option is the rear-view mirror assimilation By moving video camera + electronic devices into the mirror component, you get three non-obvious advantages:

• Cross-trim scalability: Mirrors are basic; control panels and pillars vary by trim and version. An interior-mirror DMS/OMS decreases BOM intricacy for OEMs that want a consistent conformity solution across A/B/C/ D segments. That’s critical when the required turns from “brand-new kind approvals” to “all brand-new registrations” in July 2026, since the long tail of trims need to be covered cheaply. Seeing Makers and Magna have clearly developed and demonstrated this mirror-based DMS/OMS combination, co-marketing it to OEMs as a modular, cost-contained path to mandate conformity plus cabin attributes.
• Field of view for resident analytics: From the mirror, a single cam can “see” both vehicle driver and much of the cabin, enabling OMS functions (kid visibility discovery, seat belt logic, air bag tuning, and later on, customization). Seeing Devices has verbalized this as a calculated reason the mirror is “the” position for interior picking up– since it breaks down DMS and OMS right into one viewpoint, without a sensor zoo spread around the cabin.
• Retrofit course for fleets and combined systems: The exact same industrial design that works for embedded programs can inspire scalable aftermarket kits (Guardian Gen 3 that fleets can take on promptly, speeding up the comments loophole of real-world information (exhaustion, distraction, phone use) under harsher responsibility cycles. Q 4 FY 2025 Guardian hardware devices were up 120 % quarter-on-quarter; that’s not simply income– it’s fuel for information benefit and design hardening.

Mirror assimilation appears ordinary, however it’s actually a go-to-market wedge: it decreases OEM re-engineering drag, and it converts a regulative expense line right into an upgradable attribute node for cross-selling OMS/future services. That’s where Valeo is available in.

The Valeo transfer and Asaphus acquisition: take advantage of the Rate- 1 scale, maintain the software application brain

In 2024– 2025, Valeo and Seeing Machines revealed a calculated partnership that includes moving Valeo’s driver-monitoring understanding software program task (Asaphus Vision) to Seeing Devices , efficiently combining the assumption software IP and group under the software-first company, while leveraging Valeo’s Rate- 1 scale in video cameras, ECUs, and system assimilation. This is a nuanced but effective structural step:

• It hardens the moat where it matters (the understanding formulas, datasets, and learned priors throughout physiology, lighting, eyewear, ethnic backgrounds, headgear, face hair, masks, etc), while outsourcing the capital-heavy assimilation to a worldwide Rate- 1 with OEM relationships.
• It aligns rewards : Valeo take advantage of best-in-class software program as a system vendor; Seeing Machines gain from Valeo’s scale to win even more car nameplates without bloating its very own annual report.
• It increases distribution into Stellantis-scale rollouts : While the specifics of programs are usually under NDA, market reporting and specialist coverage suggest Valeo + Seeing Machines task across huge OEM groups preparing for the EU 2026 step-up. The macro point: the partnership establishes the business approximately be the software program conventional ingrained behind numerous Rate- 1 front doors, not a single Tier- 1 -secured captive.

This mirrors just how various other flexibility layers combined: Mobileye offered systems through Rate- 1 s for a decade while worsening software/data. In indoor noticing, Seeing Machines is shaping a comparable structure– however with two flywheels rather than one.

2 flywheels: Automotive ingrained & & Fleet Guardian– each feeds the various other

1. Automotive ingrained flywheel

• Input: program wins, hardware standardization (mirror), system ports (Qualcomm cockpit/Ride), Tier- 1 distribution (Magna/Valeo).
• Output: climbing manufacturing quantities, incredibly high-margin aristocracies, installed base for OTA function opens (e.g., diversion analytics, youngster existence, customization), and– crucially– in-the-wild datasets under diverse societies, lights, and actions.
• Seeing Makers’ disclosures reveal consecutive production development and a pathway to minimal volume assurances in 2025– 2026, which need to improve profits presence and cash money conversion at scale as per the H 1 FY 2025 and FY 2024 decks.

2 Fleet Guardian flywheel

• Input: Guardian Gen 3 equipment production (now scaling), enterprise sales (straight and network), and regulatory/macro tailwinds in commercial transport security (insurance coverage, ESG, and lawsuits pressure).
• Outcome: connected subscriptions, behavioral threat decrease, and high-fidelity tags for fatigue/distraction events under durable obligation cycles; this is an AI training goldmine. Q 4 FY 2025 showed 120 % quarter-on-quarter acceleration in hardware systems and proceeded growth in connected systems and driver-monitoring profits (the high-margin persisting component) as reported earlier in FY 2024

Since both flywheels observe people moving (vehicles, vehicles, mining equipment, airplane), renovations in one domain name frequently transfer to the other. That’s greater than “synergy” talk: it’s information regimen overlap You can think about it as developing a structure model for human attentional states in vehicles– one that maintains improving as the distribution changes.

Aviation is not a side pursuit– it’s a high-signal sandbox for “driver state” scientific research

Aviation has more stringent certification, well-instrumented simulators, and more structured treatments. The Collins Aerospace collaboration (exclusive agreement to jointly market eye-tracking fatigue/alertness across aeronautics, with a multi-decade TAM estimate in their products) offers Seeing Machines a lab for a few of the hardest facets of operator tracking: lengthy obligation cycles, circadian disruption, workload spikes, and blended automation. When those models improve in the cabin, they transfer back to roadway cars where edge situations look remarkably comparable (graveyard shift, micro-sleeps, job overload, automation complacency).

To put it simply, air travel work raises the content credibility of the formulas. It isn’t immediately revenue-dominant, however it hones the blade that cuts across the bigger automobile market. This is a defensibility loop that rivals without cross-domain direct exposure battle to reproduce.

Freedom’s “backup driver” is an existing strained organization

Among the even more under-appreciated 2025 datapoints: Seeing Equipments announced a $ 1 2 M arrangement to supply additional Guardian backup-driver tracking for an international leader in self-governing driving technology (name concealed). That’s an inform. Also companies chasing after driverless procedures have to prove that safety and security motorists or backup plans are checked to a high standard as they repeat. It’s both compliance and risk monitoring. And it places Seeing Machines as an essential ingredient in the regulatory scaffolding of freedom , regardless of which ADS supplier wins. Think of it as offering shovels for the autonomy gold rush– today, not someday.

NHTSA’s identical AV oversight propositions and research study structures strengthen that in-cabin tracking is an item of any kind of reliable ADS test and release strategy. The factor is not whether freedom is “quickly.” The factor is that earns money as autonomy is presented and supervised — and Seeing Makers is locating means to monetize those stages.

Japan as a tactical movie theater: Mitsubishi’s equity and circulation signal

Mitsubishi Electric Wheelchair’s investment (targeting up to ~ 20 % with an initial 15 % using key issuance and secondary purchases) and collaborations with Mitsubishi Electric Automotive America show a critical beachhead in Japan , where conservative safety and security societies and keiretsu-style supply chains prefer companion reputation. Funding plus network accessibility can equate right into program wins that compound throughout Toyota team suppliers, Nissan, Honda, and heavy market. The North American cooperation press (MEAA advertising Guardian Gen 3 trials) extends that go-to-market muscle mass right into fleets and OEMs alike.

This is not just capital– it’s an endorsement that reduces the “occupation risk” for Japanese decision makers to pick Seeing Machines for mandate conformity and past.

The regulative “accordion”: Europe leads, the united state harmonizes in different ways– and both are great

The EU sets crisp vehicle-type authorization turning points (2022 kind authorizations; 2024 brand-new types under ADDW; 2026 all registrations). The U.S. hardly ever mirrors that framework however rather layers functional safety and security study, ANPRMs, and insurance-driven pressures that functionally press OEMs towards comparable abilities (specifically for L 2 + attributes). The result is an accordion result : Europe presses timelines and develops scale; the U.S. broadens use-cases (e.g., incorporating DMS with impaired driving prevention tech) and monetizable functions, especially when OTA ends up being a standard.

For a software provider with Rate- 1 companions and platform integrations, harmonization isn’t a requirement. Weight of possibility is enough: if the majority of vehicles in Europe have to have decent DMS/OMS, many united state trims that share designs will certainly include it also– even if the U.S. rule lags. That’s how the air bag and ESC stories unravelled historically. Enjoying NHTSA’s DMS-related study docket is hence a leading sign of income slope, not simply conformity checklists.

Concealed choice value: insurance coverage, ESG, and the courtroom

Guardian’s fleet data– and, in time, OEM telematics– produce credible, defensible evidence for threat scoring connected to interruption and exhaustion. The convergence of in-cabin sensing + telematics + insurance policy can unlock repeating information services profits, either directly or by means of collaborations. Fleets already get Guardian to reduce events and cases; insurance firms might underwrite discount rates conditioned on connected Guardian states. Over a longer arc, consumer insurance prices can incorporate DMS-derived attributes (even if privacy-preserving aggregates are made use of). The worth below is not just cost earnings; it’s calculated stickiness that maintains Seeing Equipments inside the loop of wheelchair risk markets. The more powerful the empirical web link between DMS treatments and crash reduction– and the much more this turns up in lawsuits end results– the a lot more inelastic the need ends up being. (For context, the business has actually emphasized driver-monitoring income as “high-margin recurring” from linked Guardian units; connections grew to ~ 62 k by June 2024 and maintained broadening right into FY 2025’s Gen 3 ramp.).

What the numbers are silently informing us

• Production energy is intensifying: FY 2025 device manufacturing of vehicles shipping with Seeing Devices software program > > 1 5 M, with Q 4 at ~ 488 k devices (up 36 % Q/Q). This is the oxygen for future nobilities, because high-margin OEM nobilities are a function of devices in the area and attribute take-rates , not simply bookings. The business’s Q 4 FY 2025 KPI release is just one of the cleanest photos of that fad.
• Guardian Gen 3 is a margin mix shift with time: Q 4 FY 2025 hardware systems up 120 % Q/Q to ~ 2, 536 signals not only deliveries yet likewise an action towards higher-margin software/service profits as those systems attach and stay attached. Earlier FY 2024 disclosures already showed the persisting “motorist monitoring” slice (from connections) climbing 12 % alongside 30 % growth in hardware/installation, with 19 % growth in connected units to ~ 62 k. The comp result in FY 2026 -FY 2027 is non-linear if spin stays reduced and connect rates climb up.
• The balance of “few, deep cooperations” is calculated: Monitoring’s FY 2024 outcomes discussion highlights “fewer, deep cooperations,” naming Rate- 1 s like Magna, Valeo, Collins, and a renewed Caterpillar relationship for off-road. That’s the playbook for a software core that intends to compound by means of network range without blowing up of IP. It’s precisely just how the moat remains software-centric also as hardware quantities explode.
• Minimum volume guarantees + VW ramp are stimulants: H 1 FY 2025 flagged minimum quantity assurances starting in 2025 with a product effect in FY 2026 and a late- 2025 VW ramp. Guarantees are uncommon in automobile unless the mandate/regulatory timeline obliges forward exposure. It indicates OEMs are securing the port and underwriting minimums, which can lower drawback volatility on the royalty line in FY 2026– FY 2027
• Japan funding + network lowers calculated risk: Mitsubishi Electric Flexibility’s equity (and MEAA promo of Guardian in The United States and Canada) offer the business both money path and illuminated pathways right into or else conservative procurement societies. It’s a non-obvious derisking of the worldwide pipeline, priced at a costs to current standards at the time of issuance. That’s rare signaling in small-cap security technology.

Why the moat is much more durable than it looks

1. Physiology is messy; datasets compound defensibility. Any kind of team can train a CNN on “eyes closed” vs. “eyes open.” However durable attentional state estimate throughout sunglasses, eyelash extensions, deal with masks, hats, night glare, and occlusions– while additionally keeping reduced incorrect positives/negatives appropriate to regulatory authorities and consumers– is a multi-year, multi-domain information build-up game. Seeing Machines has actually harvested this for years throughout fleets, mining, air travel, and auto. The circulation change strength is the moat.
2. Side restrictions produce “winners by optimization.” Interior picking up must run on constricted power/thermal budget plans, on SoCs that are busy with various other jobs. Algorithms that are both accurate and compute-efficient win out of proportion ports in standard pile combinations (e.g., Qualcomm). The groups that co-optimized early obtain welcomed back when next-gen silicon lands. That’s path dependancy at work.
3. Mirror + Tier- 1 partnerships compress time-to-compliance. When July 2026 shows up in the EU, OEMs will prioritize time-to-compliance at scale Mirror-based assimilation with a Tier- 1 that already provides mirrors/cameras (Magna) or broader indoor components (Valeo) lowers the rubbing. The vendor that helps an OEM stay clear of a recall or qualification hold-up makes repeat organization (and minimums).
4. Cross-domain trustworthiness matters to auditors and insurance providers. Air travel collaborations with Collins are a signal to safety culture stakeholders: your operator surveillance isn’t simply a dashcam; it’s medically grounded. That trustworthiness oils purchase and governing approval throughout domain names.

Themes most financiers underweight

• From conformity to capacity: The minute DMS/OMS beings in the calculate material, item groups will certainly utilize it for convenience (auto-logins, customized HVAC/audio), security (chauffeur recognition), and security analytics (kid existence, belt interlocks). These are revenue-multipliers that transform a regulative price into function ARPU. Seeing Machines, by being the understanding spine, is levered to this “capacity creep,” not simply the first conformity sale.
• The courtroom is a growth channel: Sadly, sidetracked driving lawsuits and fleet liability are broadening. Tried and tested treatments (audio/seat vibration/alerts) linked to quantifiable risk reduction become criterion of care. When that happens, pulling out ends up being legally hazardous. Guardian’s event-based proof trail is an unpriced possession in adoption contours.
• Autonomy’s lengthy tail needs a human-state oracle: Even as robotaxis increase geofences, mixed-traffic, mixed-weather, and edge-case handling will certainly maintain human beings in the managerial loophole (literally existing or remote) for many years. Somebody should continuously adjudicate human attention– prior to handover, during backup, and in “supervised automation” programs. Seeing Devices is offering that adjudication currently (the backup-driver surveillance agreement) and will certainly market it in richer kinds later (remote operator surveillance, attention-aware HMI).
• Aviation overflow right into “security UX”: Lessons from pilot workload monitoring can motivate vehicle HMIs that regulate signals, limit infotainment communications under high load, and change ADAS actions based upon measured cognitive state. That is a design room– not a checkbox– and the provider holding the measurement secret gets to influence that design.

Dangers (and how they’re mitigated)

• Competitors (notably various other DMS suppliers) and OEM verticalization: DMS is a warm room. Some OEMs will certainly attempt to in-house. Mitigation: (i) standard silicon stacks incentive independent, well-optimized modules; (ii) Valeo/Magna circulation plus mirror economics press time-to-compliance; (iii) cross-domain datasets are not trivial to reproduce.
• Price stress and aristocracy dilution: As fitment becomes global, OEMs lower rate. Mitigation: (i) volumes increase faster than rate decrease in early mandate years; (ii) OMS functions and solutions include ASP; (iii) minimal volume assurances and system reuse support unit business economics.
• Regulative slippage, particularly in the united state: If united state timelines stretch, fostering may delay. Reduction: EU/UK and other markets lead; shared systems splash abilities throughout areas anyhow; united state litigation/insurance stress create a de facto adoption course (NHTSA study and ANPRMs are already topping the pump).
• Implementation across too many fronts: Automotive programs are long and complex; fleet calls for a different muscle; aeronautics includes certification problem. Reduction: management’s “less, deep cooperations” position and the tactical handoff of system integration to Tier- 1 s is exactly just how you prevent organizational stretch.

What to view

1. Quarterly automotive production volumes making use of SEE software program (already disclosed): the incline matters more than any kind of solitary print. Continual > > 400 k/quarter suggests a > > 1 6 M annualized baseline before 2026’s required widening; the Q 4 FY 2025 datapoint at ~ 488 k is a healthy and balanced marker.
2. Guardian Gen 3 connections (not simply equipment deliveries): connected devices are the annuity; Q 4 FY 2025’s equipment surge should turn up as brand-new links and rising “motorist monitoring revenue” in subsequent updates. Keep an eye on spin and ARPU mix.
3. Program turning points (e.g., VW ramp, various other late- 2025 SOPs, minimum quantity assurances taking effect right into FY 2026: these flip optionality into capital and increase the defensibility of the “port” inside OEM designs.
4. Japan grip (programs affected by Mitsubishi Electric Mobility): search for statements linking SEE’s perception software into Japanese OEM systems and for North American Guardian network wins using MEAA.
5. Air travel commercialization signals with Collins (aircraft/simulator releases, cert pathway markers): these are little at first but high-leverage for algorithm trustworthiness and cross-domain discovering.
6. United state policy breadcrumbs: NHTSA study dockets, ANPRMs/NPRMs around DMS and damaged driving tech. Also “voluntary frameworks” commonly foreshadow purchase and insurance coverage trends that push OEMs to assemble on Euro-level capabilities.

A straightforward scenario framework (qualitative, not a DCF)

• Bear case (execution/price pressure): Mandate compliance drives adoption, yet price compression is extreme, take-rates are uneven, and Guardian delays. End result: slow-moving aristocracy development, uneven money gen. Mitigants: minimal volume guarantees; varied channels through Valeo/Magna; Japan capital/GTMs.
• Base situation (mandate-pull, mirror economics, Guardian intensifying): EU mandates plus platform reuse return multi-million unit annual deliveries; Guardian connections worsen to six numbers; OMS features start to add ASP. OEM nobility base turns into a significant cash money engine by FY 2026– FY 2027; fleet provides steady repeating income.
• Upside case (software standard with services): SEE ends up being the de facto perception standard for interior sensing throughout several Tier- 1 s and platforms; insurance-linked data services emerge; aviation cert opens up training/simulator earnings; U.S. approaches practical demands that benefit robust DMS. In this globe, valuation is about set up understanding endpoints and solutions connect , not simply per-unit royalties.

The crookedness comes from platform standardization : if you possess the perception module on extensively released cockpit/ADAS stacks, every new design year is a choice exercise.

Why this is a buy-and-hold

• Time arbitrage: Most screens miss that we’re mid-curve on a mandate fostering S-curve. The market has a tendency to over-focus on one of the most current quarter’s cash money burn and under-price the regulative ratchet plus platform leverage that make 2026– 2028 inflect.
• Moat maturation throughout scale-up: Every brand-new link, every new OEM program, and every aviation/fleet dataset tightens up the loophole. Rivals can win logo designs, but matching distribution-weighted data diversity is hard. This is the sort of moat that boosts during the very years public markets still dispute earnings timing.
• Stimulant layering: Q 4 FY 2025 KPIs already reveal a healthy and balanced manufacturing ramp; VW program ramp and minimal quantity warranties remain in the pipeline; Japan/MEAA cooperations and U.S. policy breadcrumbs can include boosts without any single large bang news release.
• Disadvantage cushion by means of universality: Also in a globe where autonomy lets down or customer ADAS development slows, governing conformity maintains DMS/OMS as basic equipment. The item comes to be as unavoidable as airbags were in the 1990 s.

A last mental model: Seeing Machines as the “human-state service” inside wheelchair

If you squint, the company isn’t marketing video cameras; it’s offering ground truth concerning human beings to makers that should share control. Over time, that ground reality underpins:

• safety and security adjudication (handover, contingency),
• customization (identification, preferences),
• compliance (DDAW/ADDW, insurance coverage), and
• autonomy hosting (backup vehicle driver tracking, remote operator readiness).

The only way to do this credibly throughout societies, illumination, trim levels, and operating domains is to have the understanding mind and plug into big distribution nodes (Tier- 1 s, system SoCs, mirror components). That is precisely what the Magna/Valeo/Qualcomm/ Collins/Mitsubishi latticework is designed to do. The FY 2024– FY 2025 disclosures show those pipes are full of water, not just air.

The prize is not simply greater nobilities; it’s having a critical input to the focus economy of transportation — the layer that tells makers when humans are really there.