The Dynamo Show: a Dynamo Ventures Podcast

Saman Farid argues deployment complexity, not intelligence, is the main barrier to robotics adoption at scale

27 min Listen to the original Episode page
RoboticsPhysical AIAsset FinancingDeploymentFormicVenture CapitalCloud ComputingFleet Management

Summary

Saman Farid, founder of Formic, contends that while robot intelligence has improved, deployment complexity remains the primary bottleneck to widespread adoption. He describes Formic as a hardware-agnostic commercialization platform that manages installation, maintenance, and financing to offer end-to-end outcomes to customers. Farid outlines three pillars for successful deployment: operational infrastructure, software infrastructure, and financial infrastructure. He highlights Formic's use of asset-level financing structures and a hybrid control approach combining deterministic systems with AI for edge cases.

Key takeaways

  1. 1
    Physical AI commercialization requires managing peripheral activities like installation, simulation, maintenance, and financing, which are fundamental frictions for adoption.

    Unidentified speaker (speaker_2) · 4:36

Episode sections

  1. 0:38 – 2:16

    Saman's Journey Into Robotics

    Saman Farid describes his background as an engineer and investor, noting that he initially believed robot intelligence was the main barrier to adoption. He later concluded that deployment complexity is the true bottleneck, a realization that led him to leave venture capital and found Formic five years ago.

  2. 2:16 – 4:07

    Why Deployment Complexity Is the Robotics Bottleneck

    Farid explains that while many robotics companies build excellent products, they hit scaling bottlenecks related to deployment. He notes that successful deployment requires a rare combination of technical, operational, software, hardware, and financial skills that are difficult to assemble in a single team.

  3. 4:07 – 5:43

    Formic as a Commercialization Platform for Physical AI

    Formic is described as a hardware- and model-agnostic commercialization platform for physical AI. It provides the peripheral capabilities necessary for adoption, including installation, task scoping, simulation testing, maintenance, spare parts management, and financing, allowing it to deploy a wide variety of robots.

  4. 5:43 – 8:40

    Why Customers Need Fully Managed Robotics

    Farid compares robotics to cloud computing, arguing that most end customers prefer to outsource the operational burden to specialists rather than building in-house expertise. He states that Formic offers an end-to-end solution where they take responsibility for performance, allowing customers to simply buy the outcome.

  5. 8:40 – 13:50

    The Three Pillars of Robotics Deployment

    Successful robotics deployment requires three pillars: operational infrastructure (physical teams for installation and maintenance), software infrastructure (codifying human knowledge for pre-sale validation, deployment, and fleet management), and financial infrastructure (asset-level financing structures that protect lenders).

  6. 13:50 – 16:27

    Robot Financing, Unit Economics, and Fleet Performance

    Formic uses a quilt of funding sources similar to mortgage securitization, tailored to contract length and customer credit quality. The company highlights a 98% renewal rate and 99.6% fleet uptime as critical track records for securing institutional asset financing, contrasting their positive unit economics with many competitors who have negative economics.

  7. 16:27 – 19:27

    What Makes a Robotics Deployment Successful

    Farid asserts that 80% of success comes from pre-deployment scoping and validation, requiring the robot to solve 100% of the task to be financially valuable. He advises against selling to highly opinionated corporate innovation teams, preferring customers who pay for performance and trust Formic's technical approach.

  8. 19:27 – 26:39

    Combining Deterministic Controls and AI Models

    Formic employs a hybrid approach, using deterministic controls for routine, predictable subtasks and probabilistic AI models (VLAs or LLMs) for edge cases. Complex tasks are broken into sequences where each subtask uses the most appropriate technology, such as computer vision for identification and deterministic controls for precise placement.

Claims worth verifying

Formic was founded approximately five years ago.

Unidentified speaker (speaker_2) ·3:54

How to verify: Check Formic's official website 'About' page or Crunchbase for founding date. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Formic has raised $50 million in its most recent round.

Unidentified speaker (speaker_1) ·22:17

How to verify: Check Formic's press release or SEC Form D filing for the raise announcement. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Primary source

Formic has over 100 live facilities.

Unidentified speaker (UNKNOWN) ·16:27

How to verify: Check Formic's website or press materials for deployment statistics. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Formic reports 99.6% fleet uptime.

Unidentified speaker (speaker_2) ·15:08

How to verify: Check Formic's technical whitepapers or investor presentations for operational metrics. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Formic reports a 98% customer renewal rate.

Unidentified speaker (speaker_2) ·15:08

How to verify: Check Formic's investor materials or case studies for contract renewal data. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Formic is hardware-agnostic and model-agnostic.

Unidentified speaker (speaker_2) ·5:35

How to verify: Check Formic's website or technical documentation for supported robot brands and AI models. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Formic operates robots in Iowa, Oklahoma, North Dakota, and South Carolina.

Unidentified speaker (speaker_2) ·9:31

How to verify: Check Formic's case studies or press releases for specific deployment locations. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Formic calculates that 20% of Americans have something in their house made by one of its robots.

Unidentified speaker (speaker_2) ·25:33

How to verify: Check Formic's internal calculations or public statements for this specific statistic. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Formic uses a hybrid approach combining deterministic controls and AI models (VLAs/LLMs).

Unidentified speaker (speaker_2) ·20:27

How to verify: Check Formic's technical blog posts or whitepapers describing their control architecture. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Formic's financial infrastructure includes bankruptcy remoteness and backup servicing provisions.

Unidentified speaker (speaker_2) ·12:48

How to verify: Check Formic's financing documentation or legal disclosures regarding asset securitization structures. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify Formic's customer renewal rate and fleet uptime metrics against independent industry reports or Formic's own published case studies.
  • Check Formic's funding structure and lender relationships to assess the stability of their asset-level financing model versus traditional corporate debt.
  • Review the technical specifications of Formic's hybrid control systems to validate the claim that deterministic controls handle routine tasks while AI handles edge cases.
  • Research Formic's competitors in the managed robotics space to compare unit economics, particularly regarding negative vs. positive economics claims.
  • Investigate Formic's customer base to determine if they are primarily highly opinionated corporate innovation teams or performance-based clients as claimed.
  • Analyze the 'bankruptcy remoteness' provisions in Formic's financing contracts to understand the risk profile for lenders and investors.
  • Look for third-party validation of Formic's deployment success rates and the financial value delivered to customers.
  • Check for any conflicts of interest or venture capital ties that might influence Formic's claims about market bottlenecks.
Full transcriptexpand

0:06Unidentified speaker (speaker_0): Capital is abundant. Understanding is scarce.

0:10Unidentified speaker (UNKNOWN): The

0:10Unidentified speaker (speaker_0): Dynamo Show goes deep with the new industrialists rejuvenating industry and commerce, brought to you by Dynamo Ventures. Learn more at dynamo.bc/slash podcast. Now, here's your host, Madeline O'Farrell.

0:28Unidentified speaker (UNKNOWN): And

0:29Unidentified speaker (speaker_1): welcome back to the Dynamo Show. I'm your host, Madeline O'Farrell, and I'm excited to be joined today by Simon Farid. Simon, thank you so much for being here.

0:36Unidentified speaker (speaker_2): Thanks for having me. I'm excited to be here.

0:38Unidentified speaker (speaker_1): Really excited to dive into yourself and what you're building at Formic. Before we do that, give us a little bit about your journey in the world of robotics.

0:47Unidentified speaker (speaker_2): Yeah. So I was an engineer by training, but you

0:49Unidentified speaker (UNKNOWN): know,

0:50Unidentified speaker (speaker_2): as a kid with been was been obsessed with the intersection of computers and the physical world. I wanted to see the code that I wrote turn into things moving, and robotics seemed to be the best place for that. So I spent a lot of my career in the manufacturing industry. And then I built a software company and sold it. And then I spent some time in Venture as an investor and got to work with a lot of really amazing robotics companies. And

1:13Unidentified speaker (UNKNOWN): I

1:13Unidentified speaker (speaker_2): think that entire journey was really eye-opening for me because when I first started my journey in robotics, I really felt like there was this challenge around robot adoption that was that came down to the robot. Like we just need to make the robots smarter and then we'll see adoption at scale. And ultimately came away with the belief that deployment complexity is really where the biggest friction and bottleneck is. And so yeah, I've been setting out to solve that problem for the last five years. I also was lucky to spend a good chunk of my career in China, where I worked in manufacturing and also worked with a lot of robotics and technology companies in China. So I got to see a very different ecosystem with a bunch of different primitives and a very different ecosystem of companies that was built in that world. And I think that there was a lot of learnings coming out of observing the Chinese startup ecosystem and the Chinese robotics ecosystem that informed a lot of my thinking for Formic as well. So yeah, been passionate about it for a long time and excited to get more robots out into the world. I hope we get millions and millions of robots deployed as soon as we can.

2:16Unidentified speaker (speaker_1): Amazing. Well, thank you for the background. And I'm curious, especially because you had a stint in the venture world, what was it that you saw in particular that made you feel like you didn't want to tackle this problem from the investor seat, but rather from the founder seat?

2:30Unidentified speaker (speaker_2): Yeah. So as an investor, I funded a bunch of robotics companies that were building amazing products. And each of them had a different take and would build different kinds of technology to make these robots do new tasks, whether it's picking apples or, you know, moving around a grocery store or sorting garbage or other things. And a lot of these teams were incredibly technical and really product minded and product oriented. But all of them would reach bottlenecks to scaling when it came to d figuring out how to deploy. And when I recognized that this was a problem, initially I actually tried to incubate this company as the investor, as you know, life as an investor is a lot more cushy than life as a family. And so I got to, I, you know, I tried a few times to put together a team and give them some initial seed funding to get them going, but ultimately realized that it was a really difficult business to build. It required both technical

3:34capabilities, but also really heavy operational capabilities, required software. knowledge, hardware knowledge of the customers and manufacturing, and also required actually quite a lot of financial sophistication for like asset financing and how to capitalize this type of company correctly. And I realized like those that combination of skills was really hard to put together. And I also recon realized that if I don't start this myself, I'll probably regret it for the rest of my life. So I left my role in venture about five years ago and started Formic and haven't looked back since.

4:07Unidentified speaker (speaker_1): Love that. Tell us a little bit about what Formic actually is. You've recently started describing it as the commercialization platform for physical AI. What does that practically mean?

4:16Unidentified speaker (speaker_2): Yeah. So physical AI, I think I kind of define it as these new models that allow for more direct control of robots. And as robots get more smarter and more capable, I think there are a lot of companies building products around those robots. But to actually deliver those successfully to end customers requires like a whole bunch of kind of peripheral capabilities, not just getting the robot to work, but also how do you successfully install it? How do you successfully scope out the job and analyze the task? How do you successfully test out the models and simulation? How do you successfully maintain those robots once they're in the field? What do you do about spare parts? What do you do about remote operations for those robots? And what do you do about financing those robots to get positive unit economics? And all of those kind of what I think of a little bit as the peripheral activities around robotics are a fundamental friction for adoption. And so as we have built out those capabilities across the hundreds of robots that we've deployed, we realize that we're able to use those to deploy a wide variety of different kinds of robots. And this has been, yeah. So, so, so the commercialization platform really is we can take any kind of robot, we're or hardware agnostic, we're also

5:35model agnostic. So any kind of hardware, any kind of model. and we can successfully deploy it and make it useful for an end customer.

5:43Unidentified speaker (speaker_1): And you've talked about these kind of ancillary things around robotics being actually the difficult piece of automation deployment, right? A lot of manufacturers will buy a robot and then they discover it's really the integration, the actual running of it, the programming it that's really the hard part. And so you're solving this through kind of that full service model. Can you pull on that thread a little bit about what full service really means and why just selling a robot point blank typically ends badly?

6:10Unidentified speaker (speaker_2): Yeah. I mean, I think you you gotta really think about the end customer profile here and what they're like. Most of these users of robots are not experts in robotics. And I think the analogy that is really useful for thinking about this is cloud computing. Right. So 90%, 95% of companies in the world have decided that even though they want to use cloud services, they are better suited not building their own data centers in-house. Right. Most of us don't have a bunch of refrigerated rooms with, you know, water cooling and a bunch of servers and power connectivity in in in our offices. We all have collectively decided it's much more efficient to hand the operations of these data centers to these hyperscalers and neo clouds and let them professionalize around managing these assets at scale. And by managing these assets on behalf of all of us, they're able to drastically reduce their cost of operations, their cost of procurement, their operational efficiency goes up, their utilization goes up because I might use it on the weekends and you might use it on the weekdays. And so they're able to dynamically reallocate those servers to us. And that drives efficiencies across the board.

7:27And most of us don't need to then become experts in data centers and servers. We just use the outcomes of that to run our web apps or our our kind of in-house programs. Most of us have decided

7:38Unidentified speaker (UNKNOWN): to.

7:39Unidentified speaker (speaker_2): you know, to let these cloud services companies manage the data centers. And AWS does it more efficiently than I ever could on my own. Most end customers, I think, have a very similar mentality around robotics. If I run a factory that makes chocolate chip cookies or I run a factory that makes aerospace parts, usually I'm going to be an expert in cookies or aerospace parts. I'm not also going to be an expert in robotics. And so for these end users, selling them some hardware or selling them some software or selling them some services really is not desirable. They want typically an end to end solution, a fully managed solution. Somebody else take the responsibility of performance, somebody else to take the burden of operations. And they just want to buy the outcomes from that. And that's what we're able to offer. And I think that's one of the reasons that we've been able to scale so quickly is we take on end to end responsibility for performance. And in turn, we get robots deployed in places that previously thought robots were inaccessible to

8:39Unidentified speaker (UNKNOWN): them. I

8:40Unidentified speaker (speaker_1): think that makes a ton of sense and really aligns with where we're seeing a startup activity. Like the value proposition is no longer let us be this point solution or this toolkit. It's how can we handle things end to end-to-end end. and actually sell you the outcome? And it's just such a different equation from the customer buyer perspective. Touching a little bit on what you think is required to actually do robotics as a service well, you've been pretty explicit that there's three kind of types of infrastructure that's required. Can you tell our listeners what exactly those are and also maybe which one you think is the hardest to build?

9:12Unidentified speaker (speaker_2): Yeah. So I think the first one. is operational infrastructure. The second one is software infrastructure. And the third one is financial infrastructure. So I think you need all three in order to successfully deploy a robot. The operational infrastructure is easiest to understand, but hardest to build because you basically need physical teams that can install, manage, and maintain robots across the country or eventually across the globe. So we have robots deployed, for example, in Iowa or in Oklahoma or in, you know, North Dakota or South Carolina. And regardless of where that robot is deployed, we have to have rapid response times, rapid r resolution of errors, and a rapid deployment. And those are things that we have had to. build over time. You know, the first time we deployed in Iowa, it was expensive. The second time it was more efficient. The fifth time or the tenth time, now that we have a fleet of robots that are operating in all these regions, we have all this operational infrastructure that's paid for by the existing fleet. The second piece is software infrastructure. So the way that I think about software infrastructure is we're codifying the kind of human knowledge. knowledge for evaluating and deploying and managing robots. So it comes, there's

10:31a few buckets to that. One is pre-sale, we need to evaluate a factory, create it in simulation, create the task in simulation, and validate that the robot will work, both in terms of the physical characteristics, will will it it reach, re reach,

10:48Unidentified speaker (UNKNOWN): you

10:48Unidentified speaker (speaker_2): know, will it reach long enough? Will it be able to pick up the weight? Will it be able to perk pick up that surface texture? Also, operational things, like will it hit the speed? Will it be able to do 10 cycles a minute or 15 cycles a minute or eight cycles a minute? So depending on what operational characteristics you need to validate for, we we do that through software. The second piece is the actual deployment of the robot. So once you install it, how quickly can you get it up and running? And there's hundreds of factors you need to consider. You know, what's the ceiling height? What's the size of the door? What's the footprint available? What's the power utilization? What's the height of the table that you might be working on? Is there direct sunlight or not? You know, how many samples of data collection do you need in order to train the model to work effectively? What are the kinds of error modes that you're going to encounter? How do you train the operators who are there to interact with this robot? How do you resolve different kinds of errors and issues? So there's a whole bunch of deployment related challenges that we built software to address. And then the third is post-deployment, running the robot, what are the interventions that are necessary? What are the safety things that are necessary? What are the times where you need to supplement with additional data? How do you do over the air upgrades? How do you use, how do you do cross-training from models, you know, at site A where you've encountered some error errors and issues, how do you use that to improve the performance at site B. So this kind of

12:07fleet management aspect and 24-7 operations and predictive maintenance of the components and predictive failures and all of these kinds of things in order to achieve really high uptime are important. So that's the that's the software infrastructure. And then the third is the financial infrastructure. The financial infrastructure is also extremely complicated to build because robots are essentially a revenue generating asset. So once you deploy a robot successfully, funding it with corporate equity, especially venture equity, is really inefficient. And what we in order to get access to institutional scale asset financing, there's a lot of additional work that's necessary. For example, we need to build in place bankruptcy. remote servicing so that if formic goes out of business, that asset can continue to generate revenue and the lenders on that asset can continue to get paid. Or we have to build backup servicing. So if format goes out of business, are there backup maintenance teams that can show up and

13:07have enough documentation to be able to address maintenance issues on these assets? So there's a bunch of things like that need to be put in place in order to make the financing parties comfortable enough to fund these assets on their own merits and not as debt against formic the company. So the financial infrastructure that's necessary to deploy hundreds of robots as kind of asset level financing also requires years of training and experience and track record to get access to. So we spent a lot of our time building all three of those so that now we can offer this seamlessly. seemingly simple product, which is robots on a subscription. So customers just pay a flat monthly rate to use these robots. But in practice, behind the scenes, there's a lot of complexity that we've abstracted away from us.

13:50Unidentified speaker (speaker_1): I'm curious about this financial infrastructure piece. As you pointed out, robots on a factory floor are effectively cash generating assets and be quite high-yield assets. What actually does that balance sheet looks look like when you work with these kind of financial partners? Who owns the robot? How do you underwrite a fleet? Some of those kind of parameters.

14:09Unidentified speaker (speaker_2): Yeah. So there are different ways to do it, but generally, the way that we've set it up is we have various funding sources that fund different kinds of projects. Some funding sources fund long-term contracts, some funding sources fund. sh short-term contracts. Some funding sources fund high credit quality end customers. Some funding sources fund low credit quality end customers. Just like with mortgage, right? Like a if you have a mortgage, you might ultimately sell it to different sources depending on the characteristics of that mortgage, if it's jumbo or high credit or subprime or whatever. And in a similar way, we've had to pull build this like quilt of various funding sources. that we can fund these deals with. And then we've also had to build in the track record that I was talking about. So because we have a history of deployment, we have very high renewal rates. Generally, 98% of our customers renew their contracts at the end of the term. And we also have very high uptime. We have 99.6% uptime. across our fleet. So these are things that are necessary to show a track record of so that the lenders against these assets feel confident that these will continue to have a really high yield. And we've been able to build that

15:28through through having this kind of history and track record. And then the last one is demonstrating the unit economics of the systems. So there are a lot of robotics companies today that are deploying systems. most of which have negative unit economics. And especially if you account for the total cost of operations and ownership, right? They have hours of teleop, they have maintenance and depreciation and cost of capital. If you account for CAC, you have to account for all of these things in order to act accurately represent the unit economics of a system. And unfortunately, way too many end users, sorry, way way to too many of these robotics companies kind of have have have rose colored glasses on. But these these financial companies that fund these deployments are ultimately extremely practical, right? They want to look at the unit economics and the gross asset returns on these systems and understand what it's going to look like two years, three years, five years, 10 years from now. And again, a lot of that has come down to being able to build a track record to demonstrate that and the history on these assets.

16:27Unidentified speaker (UNKNOWN): I'm

16:27Unidentified speaker (speaker_1): curious about those high renewal rates that you alluded to in there. In addition to that, you also report having over a hundred live facilities with 99.6% uptime, really impressive metrics. In an industry where a lot of automation projects get shelved or get stalled out, what actually separates a deployment from one that gets stuck versus actually goes live and is successful?

16:51Unidentified speaker (speaker_2): Yeah. You know, a lot of what comes down to a successful deployment happens pre-deployment. What I mean by that is not every task is suitable for a robot. And even the the the correct tasks, if you break them down in enough detail, sometimes it's suitable, sometimes it's not. So you really have to do

17:11Unidentified speaker (UNKNOWN): a

17:11Unidentified speaker (speaker_2): good amount of work before you sell something to ensure that you have a high probability of success. That's one of the things where I we see a lot of robotics companies go wrong is they kind of promise the sun and the stars and the moon, but they don't realize what it actually entails to successfully complete that task. And to add a little more clarification, what you need is you need to solve 100% of that task. So not 90% of the task. So 100% of the task, I'll give an example. If you're on a production line, what the person does in that station is maybe they pick something up, maybe they screw on a bottle cap, but then maybe they also have to do an inspection, and maybe they flip it over, maybe they place it in this very specific way. So there may be five or six things that person is doing. And you have to make sure that the robot does all of them before it's actually financially valuable for the end customer. And so

18:02Unidentified speaker (UNKNOWN): if

18:02Unidentified speaker (speaker_2): evaluating that upfront and ensuring that you know really what does it entail to drive success for the customer is very critical. So yeah, I think scoping and validation is I would say like 80% of the work. If you get that right, the rest of it is easy. And then for actual deployment, the thing that we've learned is frankly, like we like working with customers who don't have very strong opinions about how we solve it. It's fine for them to have strong opinions about what we solve, but the how should be our problem, not their problem. And so frankly, a lot of like large corporates with big corporate innovation teams are totally the wrong people to sell to because they are they're very opinionated on the how, despite typically not being experts on these products. And so we find that selling to customers who are willing to kind of only pay for performance, but then trust us to deliver is really important. So picking the right type of customer and having the right relationship with them matters a lot. I think some companies go down the blind alley of customizing everything for the customer to try to make them happy. And they're contorting themselves into these really weird products. with huge amounts of NRE that ultimately are not scalable for the business and also ultimately lead to bad outcomes for the

19:21customer because you've contorted yourself into something that you're not meant to fit into.

19:27Unidentified speaker (speaker_1): Maybe digging a bit deeper onto that how that you mentioned in there, you've described kind of three categories of robotic systems: traditional, rules-based robots, hybrid AI-assisted, and then these newer. kind of visual language action models. Where does Format kind of sit within that landscape? And what are you actually seeing provide results in production work today versus maybe what the prediction is for what's going to get adoption in the coming years?

19:53Unidentified speaker (UNKNOWN): I

19:53Unidentified speaker (speaker_2): think that's a that's a fantastic question. I think there's a lot of hype and excitement today around full end-to-end models. Unfortunately, I think that there's a little bit of a disconnect between reality and and the hype are around them. Not to say that these end-to-end models aren't very useful, but I think that there's a false belief, especially in the startup ecosystem and in the venture community, that end-to-end models are the right way to solve all these problems. And

20:26Unidentified speaker (UNKNOWN): to

20:27Unidentified speaker (speaker_2): answer your question directly, what we've seen to be most successful is this hybrid approach where you have deterministic controls for the robots to solve the routine tasks, especially when there's like known inputs and known outputs and a relatively predictable set of requirements. And then we use VLAs or world models, or even now we're trying out having LLM. LLMs directly control the low the robots, we use those more probabilistic methods only to solve the edge cases or the really unexpected scenarios. And that tends to work much better. So if you what we do in our system is we take any complex task and we break it down

21:14Unidentified speaker (UNKNOWN): into a

21:14Unidentified speaker (speaker_2): sequence. of subtasks. So let's just say the task is loading items into a box. So the there the subtasks may be identifying the object, picking the object, moving the object to the box, placing the object, and then closing the box. So you have these five subtasks. Tasks. Each of those subtasks may require a different level of intelligence, right? Identifying the object. maybe is best done with a traditional computer vision model. Picking the object may be done with a VLA. Moving the object to the box may be done deterministically. Placing the object often should be done deterministically because you want them kind of arranged in a specific way. And then finally closing the box may be done through some kind of VLA or something because it's a little bit hard and you have to adjust and put the flaps down, tape it up. So that's just one very simple example. But most robotics tasks actually are can be broken down in a similar way to many sub-activities. And each of those subactivities requires a slightly different way to resolve.

22:17Unidentified speaker (speaker_1): Absolutely makes sense. Some big news that we're sharing today. You recently announced a $50 million raise. Curious what this means in the form formic of journey and what this capital allows you to do in kind of the next stage of the company.

22:32Unidentified speaker (speaker_2): Yeah. So, you know, we already had a very large install base and we're growing very quickly to new customers. But what we kept encountering was customers coming to us and saying, hey, you solved these tasks for me. Could you also come and solve these 10 other tasks in my factory? And as more and more of those opportunities came up, we realized that it requires us to dedicate more effort towards like diversifying our product suite. So we felt like it was the right time to raise a little bit of additional capital to fund the build out of a couple of new product categories. And so so that's the majority of the use of the capital that we're raising in this round. And then in parallel, we're using some of the capital to fund. you know, sales and develop sales and marketing growth, right? Because we have a product the customers love. We're seeing really high repeatability once we've deployed these robots and high stickiness. And we felt like it was the right time to replicate that across a much wider swath of customers.

23:31Unidentified speaker (speaker_1): Amazing. Taking it maybe a step further, if you were to fast forward 10 years, what does the world look like if Formix Vision comes to fruition?

23:39Unidentified speaker (speaker_2): What really motivates us is living in a world of absolute abundance. There are, you know, if you look at the history of, you know, civilization, human civilization over the last three or four hundred years, the things that used to be reserved for kings and royalty are now available to all of us, right? Like a hot shower, right? We can now easily enjoy a hot shower. This is something that was previously rever reserved for the elite of the elite and maybe once a month, right? And now it's, you know, massively available. Same with, you know, food and fruit. Like we can anything you can imagine that you want in a grocery store, now we have access to. But still there are a lot of things today that are quote unquote reserved for, you know, the elite or the billionaires of the world. or even the developed countries of the world have a very different standard of living from some of the developing countries of the

24:33Unidentified speaker (UNKNOWN): world.

24:33Unidentified speaker (speaker_2): And I believe the thing that will kind of drive a better quality of life for the majority of human society is if we as a society get better at building things. If you look around you in your house, everything is made in a factory, right? Your couch is made in a factory, your clothes are made in a factory. your food is made in a factory, your car is made in a factory. Every single thing that's in your house that you touch on a daily basis was made in a factory. But as as a as a sci society and in particular in America, we've really not gotten better at making things for the last 50 years. So in my view, the biggest counterinflationary Stationary force that we can have as human civilization is to increase our capability of making things. If we can make things better and cheaper, it will create more access for everybody to the things that we know and love, improve the quality of life for people all around the world. And that's really the thing that drives us and motivates us every day. Not to mention, I get to take my kids to Costco and point at all the things that our robots helped make. And it's just so much fun, right? So if you ever go to Chipotle and eat a tortilla, our robots probably help make that. I think we calculated recently twenty percent of Americans probably have something in their house made by one of our robots. So we're starting to really see the impact of the work that we're doing.

25:46Unidentified speaker (speaker_1): Wow. That's incredible. And always excited to be able to end on such an optimistic kind of view of the future. So thank you for that. Saman, thank you so much for all of your perspectives today. It was so wonderful to have you on the show.

25:58Unidentified speaker (speaker_2): Thanks so much for having me. I hope many more people start companies in the robotics space. We need more We need millions and millions of robots deployed. And thank you for putting this together and sharing this with people so that we get more entrepreneurs working in the space.

26:10Unidentified speaker (UNKNOWN): And

26:10Unidentified speaker (speaker_1): if you're an entrepreneur working on this space, give us a call. The team at Dynamo would love to chat. And to our listeners, thank you again for tuning in. We'll catch you next time.

26:19Unidentified speaker (speaker_0): Thanks for listening. If you enjoyed this episode, leave us a five-star review and tell us what you liked. And be sure to head over to podcast. podcast.dynamo.vc to keep up to date with our latest content or subscribe on the podcast platform of your choice. Until next time.

This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.
All insights