Hugo Navarro returns to Yet Another Value Podcast to pitch Seeing Machines, the London-listed half of a two-company duopoly in driver monitoring, weeks after Europe made the technology compulsory on new cars. Andrew Walker spends most of the hour pushing back — on the convertible that matures in October, on receivables growing faster than revenue, on a fleet product whose trials keep not converting, and on a management team that long-term shareholders have learned to distrust.
👤 Guest: Hugo Navarro, who writes the Undercovered and Undervalued newsletter
🎙️ Host: Andrew Walker, who writes Yet Another Value Blog
📰 Published: 30 August 2026
🔴 YouTube | 🟢 Spotify | 🟣 Apple Podcasts | 🔗 Show notes | ⏱️ 52 min | ✅ Time saved: 31 min
Key Takeaways
Europe has made a product two companies can build compulsory on every new car
"Seeing Machines is the leader of a two-player duopoly in what's called DMS technology"
Both players took twenty years and hundreds of millions of dollars to get there
The valuation rests on a fixed cost base, not on growth
"Market cap is 330, so on mid-range it trades at around 11 times free cash flow"
The operating expense base is around 55 million US dollars a year and largely fixed
The second leg — Japan and the United States — converts revenue into cash almost one for one
"Instead of 70 extra on revenues from Europe, only 20 in free cash flow, 70 extra from Japan and America's is 70 extra in free cash flow. That's massive."
The moat is footage, not code
Mitsubishi Electric trained on synthetic data and it "worked really, really badly" outside the lab
Seeing Machines has billions of hours of truck and mining drivers on camera
The convertible maturing in October is the reason the stock is cheap
Walker: letting it get "within two months of expiration is lunacy"
Navarro says the company is in an exclusive period with a final lender and expects it done before October
The fleet business is the swing factor and it is not converting
Guardian 3 trials are running and not closing, in what he calls a worldwide trucking recession
The fix he expects is licensing the software to telematics providers rather than selling hardware against them
Robotics is worth roughly nothing in his numbers
"I assign practically zero value to it, although it's exciting over the long term"
Mitsubishi Electric paid 40 million pounds for its stake and is paying Seeing Machines to build prototypes
Receivables are growing faster than revenue, and he says that is how royalties work
"accounts receivable goes from $11 million to $25.3 million. So revenue up 45%? Accounts receivable up 120%."
Royalties are reported at quarter end and paid 60 to 90 days later, by car manufacturers
On Europe alone the automotive business gets to roughly 20 million dollars of free cash flow
Vehicles covered go from about 5 to 6 million this fiscal year to 9 or 10 million next
Full autonomy is a threat to the second leg, not the first
"even if we get L5 in five years, regulation will be slow to keep up with that"
The CEO's remaining option tranches only pay if the whole thesis works
A four-year plan of 50 million dollars of free cash flow at 20 times would treble the share price
He is not pretending this is safe
"it's a huge upside, but it's definitely one of the riskiest stocks in my portfolio"
A Duopoly Europe Has Just Made Compulsory
Navarro is on the show for the second time. His first pitch, NCR Atleos, was acquired within about two months of the episode — "I hope this one goes like the first time", he said. This one is Seeing Machines, ticker SEE.L, traded in London. Walker noted at the top that Navarro had been asking for the episode since April or May, and that both men were reminding listeners nothing on the podcast is investment advice, with the extra warning that this is a foreign security.
The company found him rather than the other way round. It was not on a screener; it appeared on his desk, and what he concluded was that the market had completely mispriced the ramp in European volumes. "It's been a 50% return since I entered into the company", he said, and he thinks the second leg of the thesis is still ahead.
Driver monitoring software watches the driver's face to confirm they are looking at the road
"Seeing Machines is the leader of a two-player duopoly in what's called DMS technology"Only two companies have the technology at the level required"both have taken 20 years and hundreds of millions of dollars to develop this""I think Seeing Machines has spent half a billion in research and development only over the last two decades"
Both companies have been listed for decades and have lost money for decades
The regulatory change is the whole reason the arithmetic works now: Europe requires the technology on every new car
The Whole Thesis Is Operating Leverage on a Fixed Cost Base
Navarro's estimate is for the fiscal year ending June 2027, which he flagged carefully because the company's year runs from June to June. On his numbers the company makes 20 to 40 million dollars of free cash flow that year against a market capitalization of 330 million — "Market cap is 330, so on mid-range it trades at around 11 times free cash flow" — and he thinks it can grow at high double digits from there.
The reason is the cost base. Operating expense is around 55 million US dollars a year and is largely fixed, so extra revenue does not bring extra cost. Europe pays for the programs already in place; the next regulatory leg is close to pure margin.
On what the second leg does to the cash flow: "Instead of 70 extra on revenues from Europe, only 20 in free cash flow, 70 extra from Japan and America's is 70 extra in free cash flow. That's massive."
Japan and the United States are the next mandates, which he put at around 2030
The fleet segment — the same technology sold to truck fleets, with Amazon and Caterpillar named as customers — is the optionality in between, and it is recurring
Walker played the thesis back: a duopoly on a low multiple where "the free cash flow kind of explodes because it's all operating leverage"
Navarro was candid that the discount exists for reasons: "there are some risks. And there's some reasons why this is cheap, we could say"
Buying the Seatbelt Maker the Year Before Seatbelts Were Required
The framing Walker liked best came from Navarro's own write-up, and he flagged that he might be paraphrasing it: buying a company before "16 million cars are mandated to get it" sounds mad, but so would buying a seatbelt manufacturer before seatbelts were required, a year ahead of most cars rolling off the line with them.
Navarro's answer is that the product is not only a trade. It reduces the risk of a catastrophe on the road by about 90%, in his telling, and that matters most for truck drivers.
On why insurers should eventually pay for this: he pointed to an insurance-technology company he thought was Lemonade, which offered a large premium reduction to US drivers using self-driving systems
Japan is pushing its regulation into 2029 or 2030 largely because accidents caused by distracted drivers looking at their phones are rising, with associations campaigning for it
He expects the insurance argument to become clearer over time as the cost savings show up
Why a Well-Funded Newcomer Cannot Simply Build This
Walker's first pushback was competitive. A duopoly in a niche is one thing; a duopoly in a market that regulation has just expanded to 16 million vehicles is another. Cumulative capital spending of roughly 200 million dollars over ten years is a lot, he said, but much of what was spent five to ten years ago is probably wasted — so why could a competitor not build a state-of-the-art product now for 40 or 50 million? And why would a manufacturer outsource something that is suddenly required, or a customer like Amazon hand over all that in-car data rather than building it in-house?
Navarro said the first leg is already locked. Seeing Machines and Smart Eye are in those 16 million vehicles, and the contracts run three to five years. He does expect a third or fourth player eventually, because that is what happens in the sector — but the technical bar is the reason it has not happened yet.
The example he keeps returning to is a customer that tried and failed: Mitsubishi Electric is on the cap table and is a client because it could not develop its own solutionIt trained on synthetic data, which tests well and then fails in the field"they got great results in, we could say, in lab test, but when they took the solution to a naturalistic environment, it worked really, really badly"
The advantage came from the company's history in mining and trucking, long before automotive mattered"So they have billions of hours of footage of truck drivers and mining employees using this technology."On his read, that is why its accuracy beats Smart Eye's
Walker pressed the point with Uber's in-car cameras and with Tesla, both of which collect data
On Tesla's own system: Navarro said it has its own driver monitoring, and pointed at a video circulating online of someone fooling the self-driving system "with like a plastic head""Like that's the current level of Tesla accuracy regarding DMS."
The Convertible Due in October Is the Elephant in the Room
The balance sheet is where Walker was hardest. The company has, in his words, "a 55 million dollar convertible that is due in October", and in its August 11 earnings deck said it was in late stages of renegotiating it, with the loan held by a customer that is fully supportive.
He did not buy the calm. "I've done markets a long time", he said, and letting a convertible loan of that size get "within two months of expiration is lunacy, I'm going to say" for a company of this market value. "Look at the balance sheet and say, this is the balance sheet of a company that's distressed or there's kind of something I'm missing."
Navarro agreed the refinancing is the gate: "I think a lot of investors are waiting for the convertible to get refinanced before investing into this."
The timing was awkward by design. The company started the process around April or May because it needed to report the volume numbers first, to show lenders the European ramp was real
Signing has slipped more than once, which he attributes to due diligence taking longer than expected
Where it stands now, on his account: "They are in an exclusive period, meaning they are now with a final lender, and they are in the final due diligence process"The company is also in contact with Magna about an extension if there is a short delay"So I think this will get solved before October. It's a risk, to be honest, but I don't think there's a high probability of this being resolved with dilution."
The Footnote That Reads Worse Than It Is
Walker had been reading the semi-annual statement to December 31, and stopped at footnote 21: an amendment with a major customer that accelerated a royalty payment, improving near-term liquidity but giving rise to future payment obligations to the customer. Read cold, that sounds like money the company will have to hand back.
Navarro said that is not what happened. The contracts with manufacturers carry minimum volume warranties. One vehicle program fell below its threshold, which triggered the clause and made the minimum royalties payable immediately.
"It's because one customer went below the volumes, basically canceled a program."As he put it, they planned to build 100 of something and built 20
The payment was pulled forward under a legal right the company already had, not lent by the customer
"that's definitely like it was poorly explained", he said, and he had raised the wording with management
Walker's complaint was as much about disclosure as substance — the company does not hold many calls, so the footnote is what an outside reader has to work from
Europe Is Locked for Three to Five Years; Japan and the US Are Not
Regulatory theses can be delayed or watered down, Walker said, and this company knows it. If manufacturers complain the requirement is too onerous, the acceleration the whole case depends on could stall.
Navarro does not see that risk in Europe, because of how cars are built. "when you design a vehicle and you put it with some certain type of software or some certain type of camera until you stop producing that vehicle, it will be with that software" — which fixes the next three to five years.
The pushback risk sits in the second leg, in Japan and the United States, not in the European ramp
He reads the complaints already coming from customers as a demand for higher accuracy rather than an attempt to remove the requirement, which he thinks favors Seeing Machines
"It's very difficult to take this away now."
Robotics Is Optionality He Values at Practically Zero
Walker put the robotics story to him skeptically. The pitch is that the company's software runs at the edge on a 20-dollar piece of silicon instead of on 20,000-dollar Nvidia chips in a data center — low latency, low power, far cheaper. That sounds like a lot for a car-industry supplier to claim in a sector this crowded.
Navarro did not defend it hard, because he does not pay for it. "I assign practically zero value to it, although it's exciting over the long term", he said.
Where it came from: Mitsubishi Electric took a 20% stake about two years ago — the industrial arm, not the car businessThe two built a joint plan covering fleets and adjacent markets, among them smart factories, robotics and humanoids
Seeing Machines is being paid to develop the solutions, and would earn a royalty if any of them reaches production
There has been a pilot and there are prototypes Mitsubishi likes, but nothing commercial
The one hard number attached to it: Mitsubishi paid 40 million pounds for the stake, at a price he believes is close to today's
Systems Versus Pure Software, and Why the Expensive One Can Be Cheaper
Walker — who noted he does about half a day of preparation for these episodes — had read that Smart Eye sells cheap software and leaves the manufacturer to source the camera and the infrared, while Seeing Machines charges double to triple and delivers the whole package.
"Almost. You know, not exactly", Navarro said. Smart Eye is a pure software approach, which the market likes. Seeing Machines runs teams on software, on optics and on the camera internals, and sells a system.
The point is total cost, not unit price: a four-dollar piece of software with a 25-dollar camera can cost more than eight-dollar software with a 20-dollar camera built for it
Building the camera for the software is also why he thinks the accuracy is better
"We could say Android versus Apple, like Apple builds their hardware for their own system."
Nobody in the Industry Can Tell You Whose Software Is in the Car
Walker asked the obvious verification question: has anyone outside the company's management confirmed that the systems approach really saves manufacturers money?
"I have not been able to confirm", Navarro said, and explained why the check is hard to run. People working in the automotive industry deal with Valeo or Magna or another tier one supplier, not with Seeing Machines.
"They are like a tier 3, we could say. So they don't really know what's going in their car. They just know that Valeo makes it work."
On why no tier one ever bought the company: neither Seeing Machines nor Smart Eye belongs to one, and Navarro thinks two decades of losses is the reasonCurrent royalties are small against what it cost to develop the technologyAn exclusivity agreement, as Magna has, a stake, as Mitsubishi took, or a partnership was the cheaper routeValeo sold its research and development team to Seeing Machines and partnered that wayNobody bought outright, he suspects, because it was never certain the regulation would arrive
Fleet Is the Swing Factor, and the Trials Keep Not Converting
Automotive, Navarro said, is now the low-risk part: the ramp is here, more royalties are coming, and he expects volumes to settle at around 2.5 million vehicles a quarter. Some of that is not even European demand — cars are built on global platforms, so a vehicle developed in Europe and sold in the United States or Japan carries the technology anyway.
The uncertain part is fleet. Guardian 3 is a camera costing around 500 dollars plus a recurring annual monitoring fee, installed in a truck to detect whether the driver is asleep or distracted. Guardian 2 worked; the new product has stalled.
"they are doing many trials, but the problem is those trials are not converting"
His explanation is the market: a trucking recession he described as practically worldwide, with higher insurance costs and higher diesel prices, and discretionary capital spending pushed out as far as possible
The sales argument is about liability, not running costs: "if you have one of these devices and your truck driver is asleep and the truck crashes or something like that, you can show that it's not your fault as a company"Walker pushed back that a product which saves money is the easiest thing to sell in a recession, and asked whether a sleeping driver really shifts liability off the operatorMr. Navarro's answer was that it turns on proving negligence, as it does in aviation
The same product exists for aviation, sold with Collins, and that is also going slowly
Competing telematics systems offer much of the same package with weak driver monitoring, which limits Seeing Machines to operators who care specifically about distraction
On the company's own diagnosis: the upfront hardware fee does not appear to be working, and management, which spent the last couple of years focused on automotive, is now trying a more recurring modelThe conversion problem is concentrated in the largest corporate customersThere are large pipeline deals; if any of them lands, he said the outcome for the fiscal year changes
Walker was unmoved by the framing: "as I've gotten more jaded, like management excuses have fallen a little bit soft"
Receivables Are Growing Faster Than Revenue
Working capital ballooned in the first half, and Walker walked through why that worried him: "accounts receivable goes from $11 million to $25.3 million. So revenue up 45%? Accounts receivable up 120%." Put next to a company that is growing fast, burning cash and carrying a near-dated maturity, he said, "if I was like a forensic accountant with a Z-score, I'd say, uh-oh, things are getting pretty crazy over here."
Navarro's answer is that this is what a royalty business looks like on the way up. Volumes are reported after the quarter closes, and payment follows.
"they report end of quarter and they get paid like 60 to 90 days after the quarter, but the risk is low"The payers are car manufacturers, so he treats default risk as very low
The company recently arranged a receivables financing facility and has not drawn on itHe reads that as a sign it does not need immediate cash, and as a positive signal about the refinancing talks
He did not dispute the optics: "I understand that the balance it looks ugly right now, especially in a company that's just in the brink of free cash flow generation and growth inflection"
How Much More the European Ramp Alone Is Worth
Setting robotics and fleet aside as cherries on top, Walker asked the core question: on Europe alone, how much further does automotive grow, and when does it reach the 40 million dollar cash flow figure? He had the ramp in front of him — production going from 488,000 vehicles in the fourth quarter of 2025 to 2.1 million in the fourth quarter of 2026, revenue up 45% in 2026 and automotive revenue up 135%, with adjusted earnings before interest, taxes, depreciation and amortization close to breakeven but still negative.
Navarro's answer separated the two legs of the number.
Next fiscal year is probably about 10 million cars, against roughly 5 to 6 million in fiscal 2026
The 2.1 million quarter covered March to June, before the rule took effect, so he expects a further step up in the July to September quarter and a range of about 9 to 10 million
On automotive alone, without fleet: "On automotive, I think it's probably 20 million in free cash flow."
A small additional lift comes from Level 2 driver assistance systems, which drive demand for monitoring outside the regulation and are growing mainly in the United States
Does Full Autonomy Kill the Business?
If a car drives itself, nobody needs to watch the driver. Walker asked whether an investor who genuinely believes the autonomy story should read Seeing Machines as a short rather than a long.
Navarro thinks the timing protects him. "even if we get L5 in five years, regulation will be slow to keep up with that, especially in Europe, and we will probably see requirements for people in the driver's seat for a long time", he said — and public confidence would lag the technology as well.
Management is working on the same risk, he said, and the answer is to sell what the company is actually good at, which is vision systems
The long-term alternative is replacing single-purpose sensors with cameras — one camera detecting whether a seatbelt is fastened, rather than a sensor doing it
Walker was unenthusiastic about that as a fallback
China Mandates It Too, but Chinese Manufacturers Do Not Pay
Chinese electric vehicles are taking share in Europe, Walker said, and if they arrive with their own in-house driver monitoring, the European market may be smaller than it looks.
Driver monitoring is also mandatory in China, Navarro said — "it's also mandatory. It's a huge market there." — and the suppliers there are local.
"there's been lots of complaints regarding bad performance of DMS in China and in Chinese cars in Europe"
He has been told Seeing Machines is looking at putting its software on Chinese cars sold into Europe
The catch is price: "the problem is that Chinese OEMs don't pay a lot, so maybe they just will not pay up"
If Chinese manufacturers take share from European ones, he expects the European market to shrink somewhat
Licensing to the Telematics Players Instead of Fighting Them
The part of the story Navarro wanted to add himself was fleet licensing. His base case treats the European automotive ramp as priced in and as the margin of safety; fleet is where he thinks the upside is, and not only through selling hardware.
The plan is to license the software to hardware developers, which changes the competitive position entirely: instead of competing with telematics providers whose driver monitoring is weak, the company gets integrated into them.
They are close to a deal with what he believes is a Taiwanese or Japanese company that builds its own hardware but could not make its software work, and wants Seeing Machines white-labeled
"The market it's smaller, but it's also higher margin."
Nothing has been signed yet
On where the idea came from: he had raised the same competitive objection with management himself, that the company is too small to beat the telematics providers on their whole offering, and asked why not license itThe first live deal came from inbound interest rather than outbound selling, and the company is now approaching other players
The CEO's Options Only Pay If the Whole Thesis Works
Walker's last question was on incentives. The chief executive has a first tranche of performance units vesting at the end of the month that is close to in the money, and further tranches above it.
Those, Navarro said, are set high. "Like I think he has some at 20 pence or something like that." Over the four-year window, delivering on both automotive and fleet would mean "50 million in free cash flow, 20x multiple, you get to 3x higher price than this one" — and on that outcome, he said, the CEO deserves to be paid well.
He described a management team that takes calls at weekends and is constantly on the road with investors
The caveat he volunteered himself: shareholders of ten years' standing have a very poor opinion of the team, mainly because it over-promises on timingHis reading is that the deals do land, late: "they tend to be right on what they will achieve, but they tend to be a bit late or a while late on timing"
Walker said stock-price incentives are rare in a European company and that granting them usually means management has a specific route in mind
He then made the opposite case from experience: a company that ten-year holders have given up on is where he has made both his best and his worst investments"It's like, hey, at some point, it's not me. It is actually them, right?"Management teams that are always about to land the big contract are the pattern he has learned to distrust, and Elon Musk is the outlier who makes people believe otherwise
A Retail Shareholder Base That Reacts Slowly
The last exchange was about who owns the stock and how it trades. "there's not a lot of institutions in the stock. Like most of this is traded by retail", Navarro said, and the consequence is that news gets priced in slowly.
He expects the refinancing, when it lands, to move the shares gradually rather than in one jump — which he framed as leaving time for anyone doing the work
The fast move, in his view, comes if the fleet turnaround starts delivering
His summary of his own position: the European ramp is the margin of safety, fleet and the next regulatory leg are the upside, and "it's definitely one of the riskiest stocks in my portfolio"
Walker closed on the irony in that: the safe part of the story is the part already happening, and the exciting part is the one still on the come
Navarro's bottom line is that a mandate has turned a two-company niche into a required component on every new car in Europe, and that at his estimate of 20 to 40 million dollars of free cash flow next fiscal year the market has not repriced it — provided the October refinancing lands and the fleet business stops stalling.
Products, Companies & Tools Mentioned
Seeing Machines (The subject of the pitch: London-listed, the larger half of the driver-monitoring duopoly, on Navarro's numbers at around 11 times his fiscal 2027 free cash flow estimate)
Smart Eye (The other half of the duopoly; sells software alone, cheaper, and on his read with lower accuracy)
Mitsubishi Electric (Took a 20% stake for 40 million pounds; a customer because its own synthetic-data solution failed outside the lab, and the partner behind the smart-factory and humanoid work)
Magna (Holds the convertible loan, has an exclusivity agreement, and is the party the company is talking to about an extension if the refinancing slips)
Valeo (A tier one supplier through which the software reaches manufacturers; sold its research and development team to Seeing Machines)
Tesla (Runs its own driver monitoring; Navarro cited a video of the self-driving system being fooled with a plastic head as evidence of its accuracy)
Guardian 3 (The fleet product — roughly a 500-dollar camera plus an annual monitoring fee — whose trials are not converting)
Amazon and Caterpillar (Fleet customers, with more pilots running; Caterpillar already large)
Collins (Partner for the same technology in aviation, also selling slowly)
NCR Atleos (Navarro's previous pitch on the show, acquired about two months later)
Nvidia (Invoked for contrast — data-center chips at 20,000 dollars against the 20-dollar edge silicon the robotics pitch runs on)
Uber (Raised by Walker as an example of in-car cameras collecting the kind of data that supposedly makes the moat)
Books & Resources Mentioned
Navarro's write-ups on Seeing Machines (Six in the past year on his own Substack; the host linked one in the show notes and it is where the seatbelt-manufacturer framing comes from)
Seeing Machines' semi-annual report to 31 December (Footnote 21, the accelerated royalty payment and the customer amendment that Walker read out)
Seeing Machines' 11 August earnings deck (Where the company said it was in late stages of renegotiating the convertible)
The company's June-quarter trading update (A cash figure rather than a balance sheet, which is why the receivables discussion ran off the half-year numbers)
Navarro's published interview with management (His own recent write-up of where the refinancing stands)
Get the latest market chatter as it happens: