Duos Technologies (NASDAQ:DUOT) produces inspection portals called rip (rail inspection portal) based on machine vision and AI, which it sells to rail companies.
These rips mostly generate one-off sales, but there is a stream of recurring revenue attached in the form of service, consultancy, and algorithms.
They have advanced plans to develop a subscriber model which would broaden the market significantly and produce more recurring revenues.
At present the company has three class one rail customers and 13 rip deployments across the US, Canada, and Mexico, with two more scheduled for H2, but the aim is to end the year with 20 rips.
The company is also deploying machine vision to its ALIS (Automated Logistics Information System), where it has 19 deployed at 10 different sites at a major retailer.
At present, these ALIS mostly function as an automated gate processing system but when they are upgraded with machine vision and algorithms they become truck inspection portals, the equivalent of rips for trucks.
Spectacular train derailments with dire consequences earlier this year have brought back the focus on technologies that can help here and triggered a bipartisan effort to get the proposed Rail Safety Act 2023 through Congress.
There is some progress here in the Senate (the House will likely take months, and then more months to write the actual regulations), but the sector might be spurred into action even without new laws given the seriousness of the accidents.
The company has a decent market position already (Q1CC):
our rips performed over 1.7 million comprehensive railcar scans, of which more than 238,000 were unique railcars. This metric encompasses all of the railcars scanned at locations across the US, Canada and Mexico, representing approximately 15% of the total freight core population in North America.
Machine Vision and AI
The core of Duos’ competitive advantage lies in the algorithms it has developed to detect faults, but there is something more general at play here, and we have to quote at some length (Q1CC):
we control all four of the control knobs that are required to actually produce AI and then actually deploy it correctly. And that involves being able to make adjustments on the hardware side, the IT infrastructure side, the software side. That all supports good artificial intelligence. Just about everybody else that I’m aware of, somebody else produces the hardware. They might do the hardware, IT, and software, but then they’re almost always sending the AI production out to a third party who doesn’t understand the other three components of the solution. And so we’ve seen some of our rail customers or potential customers struggle with that model. We do all four of those components in house. What our goal is to get to is where we can cover the entire FRA inspection checklist with artificial intelligence.
Some manifestation of this is the fact that they can now inspect railcars running at speeds of up to 125Mph, which is up from the 40Mph just a few years ago. Controlling all four of the inputs (hardware, IT, software, AI) has its advantages (Q1CC):
culmination of more than a year’s worth of development effort across several disciplines, including hardware engineering, IT, and software.
Another example is the case of one of their class one customers which suffered from two derailments, which were caused by defects on the end of car cushion (basically a shock absorber) it was determined.
The company developed and deployed a new algorithm (called Head of Cushioning Unit Condition detection model) at the customer’s request, and they can now scan the end of the car cushion. Within 5 days the rip found 19 validated defects which were immediately addressed, avoiding these causing more derailments.
They are continuously increasing the number of algorithms, they have 37 now but plan to have more than 50 by yearend, which also increases the stream of recurring revenue.
Recurring Revenues
Q1 produced $838K in recurring revenues (out of a total revenue of $2.64M, an increase of 84%). The company gets recurring revenues out of:
- Services and consultancy
- The number of AI models (and hence detections)
- Duos Zone portals, its upcoming subscription model offering
with the increased number of AI models we continue to roll out, we have increased the portfolio of detections we’re able to offer, which has led to a steady increase in our recurring revenue base. Our customers are actively using the additional algorithms as they become available, and we are developing new AI with their direct input.
But recurring revenues could get a serious lift when their subscription-based model gets traction.
Duos Zone Portals
Instead of taxing CapEx budgets of prospective customers, these portals would be available to anyone with a subscription. We think this is a very smart move:
- It is likely to significantly (management argues “dramatically”) broaden the potential customer base.
- This de-risks the dependence on a few large customers (they have three Class One customers).
- It creates a predictable stream of recurring revenue, rather than one-off sales (with a smaller stream of recurring revenues for services and algorithms).
- To our knowledge, it’s not said or implied anywhere, but we expect increased recurring revenues to boost gross margins (at present around 20%) and this could help lower the break-even point of the company.
The company raised $4M this quarter for the buildout of rip systems on a subscription basis. The company is on track to build the first subscription rips in the Southeast in H2/23.
Management is also in negotiations to repurchase existing portals to deploy them on a subscription basis. It is finalizing negotiations with the first subscription customer and expects this to be operational by the end of Q2. From the Q1CC:
we’re in discussion with probably about 20 legitimate car owning companies or companies that, chemical companies, we’ve got a couple of those that we’re talking with and some other shippers, if you will. There’s serious interest from all of them. They can see the benefits of it. By and large, the subscriptions are definitely at the top of the list. I would say that there are probably about a third of those.
Finances
- The company managed to expand a $9.4M Master Services agreement with a major national passenger carrier by 40% to $13.5M with the completion of the first two high-speed passenger portals later this year.
- The company completed two other rips for two other customers (in Georgia and Texas), taking the number to 13 with an additional two rips later this year.
- Backlog was $9.4M at the end of Q1 (40% services and 6% project revenues), of which $7.7M is expected to be recognized this year.
- FY23 revenue is guided at $20M-$21M (+33% to 40% from $15M in FY22).
- The completion of the first two high-speed passenger portals is slated for later this year.
- Revenue +84% to $2.64M ($838K of which are recurring) down a lot from the $5.93M in Q4/22, but revenue is lumpy quarter-to-quarter.
- Gross profit +142% to $537K or a gross margin of 20.3%.
- OpEx -5% to $2.72M.
- Cash was $4.3M (after a $4M financing this quarter).
Q1 figures are not yet in the graph:
The company is still quite a bit away from being profitable as OpEx is virtually as large as revenues, so the latter has to increase five-fold (on a gross margin of 20.2%) to become profitable.
At an $11M OpEx run rate and 30% gross margin (gross margin was 31.6% in FY22 and 36.1% in Q4/22), revenues need to be $36.6M for the company to break even.
The revenue level at which the company stops losing cash is probably a few million below that but still well ahead of the guided FY23 revenue ($20M-$21M).
Things look better on a 40% gross margin as it becomes a more palatable proposition with $27.5M as the break-even revenue level and this is within grasp in 2024.
We think that gross margins have a shot at 40% as they are quite variable with output (reaching 36% in Q4/22 on $5.9M and sinking to 20% on $2.64M in Q1/23) and hopefully a gross margin boost from a shift towards recurring revenues when the subscription portals take off.
Valuation
There are 80K warrants and 924K options, 433K common shares issuable upon conversion of Series D Convertible Preferred Stock, and (iv) 1,33M common shares issuable upon conversion of Series E Convertible Preferred Stock.
So fully diluted that’s 9.84M shares at $4.6 or a market cap of $45.2M and an EV of $40.9M or 2x FY23 expected sales.
Risks and Conclusion
The company has a really interesting technology that has wider applicability beyond rips, where it already has a considerable market position. It keeps improving the technology, adding algorithms and recurring revenues through an innovative subscription model.
This could give the company a boost, addressing a much larger part of the market, and further growth impulses might come from the industry’s reaction to recent derailments and legislative efforts around this.
However, at present, the company is still far away from breakeven and it has considerable CapEx needs in building out the rips for the subscription model while they have fairly limited cash.
We expect that the subscription rip model will boost gross margins and lower the break-even revenue level, but there is not much investors know about the economics of the model, so we can take little for granted here.
It also remains to be seen whether the expected market expansion effect of the shift to a subscription model outweighs the spreading out of revenues over time.
However, we think that in 2024, the company has at least a shot at $27.5M in revenue and 40% gross margin, which would be enough to reach breakeven, which would make the shares pretty interesting at these levels but not one of our top picks.
Editor’s Note: This article covers one or more microcap stocks. Please be aware of the risks associated with these stocks.
Read the full article here