Robotics VCs on what’s real, what’s coming, and what to keep in mind

Last week, at TechCrunch’s robotics event at UC Berkeley, we sat down with four VCs who are making a range of bets on robotics companies, from drone technologies to robots whose immediate applications aren’t yet clear. Featuring Peter Barrett of Playground Global, Helen Liang of FoundersX Ventures, Eric Migicovsky of Y Combinator and Andy Wheeler of GV (pictured above), we covered a lot of terrain (no pun intended), including whether last-mile delivery robots make sense and how much robots should be expected to do without human intervention.

We also discussed climate change and how it factors into their bets, and why the many private enterprises focused on creating fully automated vehicles may need to do much more to empower the cities in which they plan to operate. You can find excerpts of our talk below. And for access to the full transcript, become a member of Extra Crunch. Learn more and try it for free.


TC: How do you think about investing in the here and now, versus the future (which is complicated for VCs, given that venture funds need to produce returns within a ten-year window, typically):

PB: One of the challenges with investing in robotics is that robotics companies do tend to take a lot longer to mature than your average enterprise SaaS company. There are some classes of investments that we know the technology works; it’s just a question of commercializing it and bringing it to market, and Canvas [a Playground-backed company that makes autonomous warehouse carts and was just acquired by Amazon] did an extraordinary job of finding a market that existed and had technology in hand that would solve that problem.

There’s other stuff like the amazing work that the folks are doing at Agility [Robotics] with a biped that can operate for many hours in unstructured human environments that today is really, candidly, a research robot, and to reach its long-term aspirations, there’s a whole other set of technologies that we’ll need to develop as the company matures.

We think about blending the stuff that’s very impactful but is going to take a long time because it’s fundamentally a new science and technology that needs to be created, [with] immediate applications of technologies that are proven today, that we’re deploying against real markets.

AW: As for whether we try to build a portfolio where there are exits at different stages, generally, when I’m looking to invest in a robotics thing, I understand that the timeframes can be fairly long, and so what we’re looking for are things that really are going to be very large opportunities — that can generate billion-dollar-plus exits.

TC: A growing number of small last-mile delivery robots has attracted funding. Helen, your firm is an investor in one of these startups, Robby. What’s the appeal?

HL: We look at where we see a pain point in the market. During our team meetings on Fridays, we always use DoorDash. It feels awkward when we order a $100 meal, and the delivery person has driven a long way. We’ll give him a $15, but it’s still [tricky for that person] in terms of economics. If you have a central station for the food delivery, and robots can handle that last-mile delivery, we think that’s a more cost-effective approach.

Robby has partnered with PepsiCo [to delivering snacks to students attending the University of the Pacific in Stockton, Ca.] that makes it more like a vending machine, and we think that’s an interesting market, too. We’ll see how fast adoption will happen.

EM: YC is an investor in Robby as well, and we think of this as kind of the perfect example of how hackers can get into a fairly complex industry. When you look at some robotics and specifically autonomous vehicles, you see extremely large investments going into some of the some of the big players, but then at the same time, you see groups and hackers that are able to use off-the-shelf technology to solve real problems that affect businesses or people, and build services or products that that are valuable. We’ve seen this over and over.

You don’t have to be looking for a large VC investment to compete in the space. It is possible to stay frugal stay nimble and build something on a small scale to demonstrate that you found a problem that people are willing to pay money to solve. Then, if you’re interested, [you can] pursue larger VC investment or not. It’s kind of open right now.

TC: VCs we’ve talked with in the past have suggested that in robotics, they often see cool ideas for which there isn’t necessarily a market or big market need. Is this also your experience?

PB: This is a common pattern where there was some mechanism, some capability of the robot, some feat of dexterity or something [and founders think, ‘That’s really cool, I’m going to make a company out of it.’ But we think about it in terms of, what do you want from the robots? What’s the outcome that everybody agrees is worthwhile? And then, how do you find and build companies to achieve those goals?

One thing we’re struggling with right now is that there’s no real hardware or software platforms. You think about 10 years hence [and] the kinds of things we’ll be investing in, [and it’s] robotics applications that are aggregates of neural networks and some explicit software bound together in some form that can be delivered, so a large enterprise can use an application and not have everybody start from first principles. Because right now, when you built a robotics application, you make all the hardware, you make all the software. All the intellectual and actual capital [money] gets dissipated, building and rebuilding those same things. So robotics applications over time will be investable, much more like the way we invest in software, and that will allow smaller units of creativity to produce useful products.

TC: Andy, how long do you think it’s going to take until we get there?

AW: I think I think we’re making we’re making steady progress on that front. To your earlier question, this space has a lot of folks that are building technology a bit in search of a problem. That’s a common thing in startups generally. I would encourage everybody who’s looking to build a startup in the space is to really find a burning business problem. In the course of solving those [problems], people will build these platforms that Peter was talking about, and we’ll eventually get there in terms of [founders] just having to focus on the application layer.

TC: There are so many buckets: delivery robots, self-driving trucks. Both relate in ways to the overarching problem for our age, which is climate change. How much do you factor climate change into the investing decisions that you make?

PB: When we look at applications and robotics in agricultural, a lot of [our questions are] around how do you deal with a minimum carbon footprint, [and] how you replace workers who are missing. And dealing with climate change will be increasingly be a central thought in what we want from our robots. [After all] what we want from them is the ability to maintain or improve the lifestyles we have without further unwinding the environment.

TC: We talked backstage, and you think we are over-indexing on autonomy as the answer.

PB: When we think about autonomy, it’s not clear how autonomy helps cities. . . There are absolutely applications for autonomy, [including] on a farm or in a logistics environment. I think we still really don’t know how to do Level 5 [which is complete automation, requiring zero human assistance]. And I don’t think we know whether it’s exponentially hard or asymptotically. I think it’s decades before there’s any significant Level 5.

[In the meantime, if] we cared about safety, we’d install roundabouts or lower the blood alcohol limit and not try and make a sentient vehicle that drives on the road the way we do, right?

I’d much rather see having the city collaborate with the vehicles and instrument the city to collaborate with clever vehicles for the benefit of everybody who lives there. But that’s not Level 5 autonomy as the way we think of it

EM: It’s slightly interesting that autonomous vehicles, specifically the individual passenger car, evolved in America, because it’s one of the countries that has the least public transport per capita. And that that’s one of the things that the industry has to acknowledge — that there are other options that can be blended into the transport solutions for cities.

It seems like it might be happening because it’s something that an individual can take somewhat control over. You can’t own a bus, but you can own or [rent] a self-driving car.

PB: Or [an electric] scooter or a bike, right. The future of mobility is going to be a blending of all of these things. But not taking advantage of a logistics platform in a city means you’re kind of doing it the hard way, trying to make a robot to have all the human priors required to drive safely. And it’s just not clear that we know how to do that yet.

TC: Andy, GV is a big investor in Uber. What what’s your thinking? Does the city need to be a kind of central brain in order for these private enterprises to work effectively?

AW: I don’t think it’s a strict requirement at all. We’ve seen success with with self-driving trials where the city is not super involved from an infrastructure perspective, I do think it makes it a lot easier if that’s the case, though.

Note: EC readers, you can find the entire transcript below, though it’s sometimes difficult to make out as we had audio issues the day of the event.

Connie Loizos: It wasn’t that long ago that investing in robotics was considered fringe but obviously no longer. In February alone, I think something like $4.3 billion was invested in robotics companies. And that was a big climb from January where $1.4 billion was invested in them. And it’s not really a surprise, given that robotics companies sent more of their robots to North American companies last year than ever before.

We’re also seeing great exits like the $3.4 billion sale of Auris Health to Johnson & Johnson earlier this year, and that’s not including milestone payments. So lots of exciting things going on, but obviously there are lots of challenges as well, which is why we are so thankful to have this esteemed panel of robotics venture capitalists with us to talk to us about these things. Eric, Peter, Helen, Andy thank you so much for coming.

Peter, I thought we would start with you if you don’t mind, I was looking at your portfolio this week. One of your companies just sold to Amazon, Canvas. Congratulations. Your portfolio is so interesting and diverse. You have warehouse robotics, but you also have this very sort of sci-fi company called Agility Robotics which has these kind of bipedal robots whose application isn’t immediately clear. So I was wondering, how do you think about investing in the present and investing for the future?

Peter Barrett: One of the challenges with investing in robotics is that these companies do tend to take a lot longer to mature than your average enterprise SaaS company. And I think that there are some classes of investments where we know the technology works, it’s just a question of commercializing it and bringing it to market and the Canvas guys did an extraordinary job of finding a market that existed and it had technology in hand that would solve that problem.

I think we have some others like FarmWise who are doing agriculture. There’s a class of things which are mature markets and mature technologies, and there’s other stuff like the amazing work that the folks are doing at Agility with a bipedal that can operate for many hours in unstructured human environments that today is candidly really a research robot to reach its long term aspirations.

There’s a whole other set of technologies that we’ll need to develop as a company matures, and we think about it as blending the stuff which is very impactful and is going to take a long time because fundamentally a new science and technology needs to be created, balanced against immediate applications of technologies that are proven today. We’re playing against real markets.

Connie: Andy, backstage we were talking a little bit about this and I asked how do you think about things time wise? And you sort of seem to shake your head like you don’t think of it necessarily in such discrete ways. Is that accurate?

Andy Wheeler: Generally when I’m looking to invest in robotics as Peter said, I understand that the timeframes can be fairly long. And so what we’re looking for are things that really are going to be very large opportunities right? Things that can generate kind of billion dollar plus exits.

Connie: Do you have a more flexible timeline than some other firms or traditional venture firms might?

Andy: I mean, we still use a traditional venture funding structure so our funds are still 10-year life vehicles, the way that other people do.

We do start a new one every couple years. So maybe slightly faster than others but we’ve always got 10 years.

Connie: Helen, talking about things that are out in the world, you’re an investor in Robby. There’s another company out here I saw on campus called KiwiBot that seems sort of similar, maybe a little bit smaller, but these little delivery robots. There’s Robby there’s KiwiBot, there’s Starship, there’s Marble. I think even Postmates is coming out with something called Serve that is a similar vein and they can roll them out in Los Angeles.

Tell us a little bit about why those investments?

Helen Liang: So when we look at last mile delivery. A lot of times it begins with economics right, whether you use humans to make a last minute – so I’ve had experiences at our office when we have team meetings on Friday, and we always do DoorDash. So, I find it sometimes awkward when we order maybe a $100 meal for the delivery guys who actually drove a long way, and you give them maybe $15 tip. That still can be very hard in terms of the economics, how does that work out?

Then we say if you have some kind of central station for the food delivery. Then you have the last mile delivery done by the robots I think that’d be maybe a more cost-effective way. That’s one angle we look at this, right and a campus for like, Robbie right now they’re doing a pilot with PepsiCo or Coke’s delivery and it makes more like a vending machine, and of course it can do pizza delivery and other things. So, we do think that’s a very interesting market, and we will see how fast adoption will happen.

Connie: So I have to say I think they’re so great, but I’m also a little bit skeptical because your stories about people tipping them over and blocking things on the sidewalk. Eric, how do you think about this? Y Combinator obviously has a huge portfolio of startups but it hasn’t made a bet on this type of delivery robots.

Eric: We’re invested in Robbie as well. I think of the space as one of the perfect examples of how hackers can get into a fairly complex industry.

When you look at some robotics and specifically the autonomous vehicle investments, you see extremely large investments going into some of the big players but then at the same time you see groups and hackers that are able to use off the shelf technology to solve real problems that affect businesses or people, and build services or products that are valuable.

We’ve seen this over and over again I think one thing that I want to caution is that in a sense you have to be looking for a large VC investment to compete in the space. It is possible to stay frugal, stay nimble and build something on a small scale to demonstrate that you found a problem that people are willing to pay money to solve. And then if you’re interested, pursue larger VC investment or not, it’s kind of open right now.

Connie: That’s great. A year ago some of the VCs I talked with said one issue that they see in robotics is often hackers coming up with these cool ideas for which there’s not necessarily a market need. So I wonder, a year has passed and I don’t know if much has changed. Peter, you’re nodding your head, what do you think about this?

Peter: Well I, there was a common pattern where there was some mechanism, some capability of the robot, some dexterity or something where I was like “that’s really cool I’m going to make a company out of it. What’s it for?”

And we’d like to think about it in terms of what do you want from the robots, right? What’s the outcome that everybody agrees is worthwhile? And then how do you find and build companies to achieve those goals? And I think one thing we’re struggling with right now is there’s no real hardware or software platforms, right?

You think about 10 years, hence, the kinds of things we’ll be investing in robotics applications, which are aggregates of neural networks and some explicit software bound together in some form that can be delivered. So Mercedes console seats or some other large enterprise can use an application and not have everybody suffer from first principles. Because right now, when you build robotics applications, you make all the hardware, you make all the software, all the intellectual property, and actual capital gets dissipated, building and rebuilding those same things. So I think robotics applications, over time, will be investable, much more like the way we invest in software, and will allow smaller units of creativity to produce useful products.

Connie: Yeah, you actually shared a really surprising stat with me earlier that something like $120 billion dollars was spent just on software iteration last year.

Peter: Yeah, I mean, you look at the spend in automation, and robotics, and the vast majority of it evaporates on people doing the same thing over and over again. So three running software, it’s doing the systems integration of the same class of thing over and over again, it’s not accruing common value to a platform. And once we get good hardware and software platforms, six degrees of freedom, kinematic trains, where you can buy that from different companies and run the same applications on it. Having applications operating highly at the perception and sort of neural net manipulation and not worrying about deployment and operations and security. Right, all of that platform stuff is entirely missing in the robotics space. It exists in the mobile application space. And we have a huge ecology that comes out of that, we need to build that up for robotics.

Connie: Andy, how long do you think it’s going to take until we get there?

Andy: I mean, I think I think we’re making steady progress on that front. You know, to take a slightly different tack, I’d say for sure this space that has a lot of folks that are building technology a bit in search of a problem. And I think that’s a common thing in startups generally.

But the thing I would encourage everybody who’s looking to build a startup in the space is to really find a business problem that is burning. And in the course of solving those, people will build these platforms that Peter was talking about, and we’ll eventually get there, you just have to focus on the application.

Connie: And when you’re talking to young founders here who have these great ideas, what are you looking for in terms of the team composition? Do you want to see an advising Professor involved? Do you want to see somebody who’s an MBA?

Helen: So most often, at very early stages or seed stage, you see techy founders, they’re out of a MIT Lab or Berkeley Lab, that’s very common. What do we typically do is ask a lot of questions to think through how big a problem you’re solving, right? Whether the market is ready. So a lot of times we do see the techie founder, and it’s just because they know how to do it. But that’s where we actually lost a lot of question.

Of course, very often they have an advisor, maybe the professor at a school, that may not be the most important thing, the most important thing is that you define a product market fit, and how you take the prototype to go there and find that the right value proposition. That might be the most challenging part. So we want to see how the founders think about that right from the beginning, even though they may not have an answer right away, you will help them in terms of product design. How they position their startup right from the beginning. That’s where I see we ask them more questions.

Connie: Thank you. Eric, back to Y Combinator, which has so many companies that its baffling sometimes. You are an investor and somebody robotics one is, but some of them seem to compete somewhat directly, like Starsky and Embark. So how do you think about that? Does it matter to you if they compete directly?

Eric: YC’s philosophy is that more startups are better for the world. So we’ve taken a really broad stance supporting founders, starting from even our free programs like start school, which is a free online course that anyone can pursue. So we have supported numerous companies that compete with each other, sometimes in the same batch more often than not separated by batches.

As an investor, we start at the early stage and we tend to help more on the mechanics of starting a business. And the actual technical nature of the startup, we aren’t the experts, like one of the reasons why we invest in the companies that we do is because they know more. They’re super smart founders, they know what they’re doing. They know what they’re working on. And we’re here to support them. I think we can do our job pretty well, with competing companies, other work extremely careful to make sure that the same partner is not necessarily the person who’s working on them. But in the past, we haven’t seen too many problems there.

There are now over 2000 companies have been funded by Watson,

Peter: It’s tricky not to have competing companies.

Connie: But there’s only so many self-driving truck companies, and although it’s amazing how many there are. These companies are good for the environment, there’s no shortage of people to drive trucks. Hopefully we’ll have more electric vehicles.

Generally speaking, I just wonder how you think about different buckets you can invest in with the overarching problem of our age being climate change. And I wonder how much you factor in your decision making that reality and how it influences the decisions that you make?

Peter: Yeah, well, I think when we look at applications like robotics in agricultural, we look around how you deal with a minimum carbon footprint, how you replace workers who are missing. And I think dealing with climate change will be an increasingly central thought in what you want from our robots. What we want from them is the ability to maintain or improve lifestyles we have without further unwinding the environment.

And I think when we think about our economy, it’s not clear how economy helps. It’s not clear how autonomy helps cities, it’s not clear how autonomy helps equity for transportation within cities. And so I think there are other lenses that you can look through, climate change being one of them, that makes you make different choices about the kinds of things you invest in.

Connie: So we’re over-indexing on self-driving cars, self-driving trucks/

Peter: You know, I may get in trouble for some of this. But I think that I’m not sure what problem it really solves. There are absolutely applications for autonomy, small autonomy on a farm or in a logistics environment. I think we still really don’t know how to do level five. And I don’t think we know whether it’s exponentially harder. And I said I think it’s decades before there’s any significant level five. If we cared about safety, we would lower the blood alcohol limit, not try and make a sentient vehicle that drives on the road the way we do, right?

I mean, I much rather see having the city collaborate with the vehicles and instrument the city to allow the city to operate with clever vehicles for the benefit of everybody who lives there. But that’s not the level five economies we think of. It’s a collaborative framework that these robots, and the city can interact in a useful way that makes it better for everybody.

Eric: It’s slightly interesting that autonomous vehicles, specifically the individual passenger car, evolved in America, because it’s one of the countries that has the least public transport per capita. And I think that that’s one of the things that the industry has to kind of acknowledge. That there are other options that can be blended into the transport solutions for cities.

Now, it seems like it might be happening, because it’s something that an individual can kind of take somewhat control over. You can’t own a bus. And you certainly can’t own a centrally a self-driving car. But you can own a scooter or a bike, right.

Peter: You know, I think the future of mobility is going to be a blending of all of these things. But I think not taking advantage of a logistics platform in a city means you’re kind of doing it the hard way. Trying to make a robot have all the human priors required to drive safely. And it’s just not clear we know how to do that yet.

Connie: Andy, GV is a big investor in Uber. What’s your thinking, and your colleagues thinking, about this? Do you think that cities have to be the central brain working with these private enterprises? Or that’s not necessarily the future that you imagine?

Andy: With respect to kind of self-driving cars, right?

Connie: How do they work? If they’re not working with a municipality?

Andy: I would say, I don’t think it’s a strict requirement at all. And I think we’ve seen success with self-driving trials where the city is not super involved from an infrastructure perspective. I do think it makes it a lot easier if that’s the case, though.

Connie: Going back to my first question – how long can it take for a company to exit? Like surgical robotics companies, Auris did very well, but Auris is 12 years old and it was founded by somebody who had a very established reputation, and it found another publicly traded company. Is anyone here investing in surgical robots?

Helen: We have been looking into that space.

Connie: What are some of the areas that you think are most interesting? So Eric, you for example, have invested in drone companies. Tell me a little bit about where you see the opportunities there. It seems like there was so much hype over drones.

Eric: There certainly was, on the consumer side, I think the more interesting story now is using the consumer level technology that’s been developed and now fits in your pocket for specific business applications that have an immediate ROI for the business or enterprise. We’re seeing this over and over again, both in small scale robotics, in the warehouse and logistics world, to agriculture, technology, to indoor farming, there’s just a wealth of opportunities. They start out as an issue. In fact, some of the best ideas look really small at this stage. But they’re solving something with instead of building the entire stack themselves.

They’re starting from what can I buy off Amazon? Can I buy it? It’s already available on the consumer side, and then they create the software and create the offering that solves that problem. That’s been doing great for us. We’ve seen this work really well. On the agricultural side, we’re seeing like logistics and warehouse solutions, just immediately flooding in and either saving money or making more money for the enterprise.

Andy: We’re seeing companies get a lot of traction on things like building on the DJI platform. Taking a completely off the shelf drone, but then building commercial applications on it. So for example, in our portfolio, SkyCatch is doing that for construction. So yeah, we’re seeing a bunch of people taking advantage of these regular consumer platforms.

Connie: What about Skydio? That’s one of your deals – it’s been called the autonomous selfie drone. It follows you, it knows who you are right?

Peter: So it is the most sophisticated flying autonomous robot, that has this amazing perception system where it can operate in unstructured environments and not fly into things. And it has this beautiful perception and planning system that is increasingly available through API’s to allow people to build commercial applications.

And there’s lots of very useful things have you done with a drone like that? And I think for me, the real dichotomy is between applying something like a SCA do for you know, various commercial applications and various non-consumer applications versus documenting the skies with drones delivering pizzas, which I don’t think anybody,

Connie: Right.

Peter: So I think those flying platforms will get more capable and have better sensing. And I think interesting startups will build software applications using those existing hardware platforms.

Connie: You know I wanted to ask in terms of forming syndicates – I thought, for sure everyone on the panel would have known each other already and that wasn’t the case. I just think that speaks to again, how much interest there is from all over the world. Robots have a lot of money coming in from China and South Africa from everywhere. How do you think about investing in these things – like ideally, does a corporate partner come in at the Series B or Series C? If that doesn’t happen, is the company doomed?

Andy: Most of most of the companies in my portfolio have pretty traditional venture syndicates, all the way through, and we’ve got the first year where that has crossed into late-stage financing. And you’re seeing traditional crossover investors come into those. So I think a pretty traditional investing path is totally possible here. And we do have a handful of companies where it’s been more strategic based, that’s particularly if they occupy you know, an area in the niche or a niche industry.

Helen: Yeah, so I would add on to that. We invest in a few drone companies one is called Aptonomy. You know, a lot of drone companies early on look at the security space. So this company is also looking at space and working with like amount of employees for July 4 firework, etc. And they pivoted into inventory management, so they do right now actual mapping for all the car dealers’ parking lots. If you like a car, you need to find out where they are, this says this car is here, go find it.

So this company, they’ve come out with a solution is really nicely. So they have all the cars going that have like a QR code, and the drone flies over to scan the QR code to take you from where you are to get that car. So this is one of those example. A lot of drone companies today are looking to new ways to solve a problem. Traditionally, you may have never thought about the use of drones for that.

So that’s one regarding the vertical versus horizontal play. A lot of times we tend to think it’s becoming much easier for startups to survive or make some money by going vertical, then when they see the ability or opportunity for them to go, for example, in that case maybe take the more horizontal play. That’s where we’ve seen a couple cases like that. Right from the beginning, if you want to a horizontal play, you will need a very big capital round, so for a product for a long time, you won’t be able to make money.

Connie: So one last thought before we go. I don’t know if it’s come up on stage but we’ve talked online or offline about what happened with these flights and the fact that these the pilots for the Lions Air and the Ethiopian Airlines flights were given very confusing information from the sensors where they were saying the planes were going up when in fact they were going down dangerously fast. Do you feel that we’ve over indexed on how much robotics does versus people? Is there something that they should be think people should be thinking about in terms of design?

Peter: You know, those vehicles are extraordinary complicated, but the emphasis itself was quite simple, it was a safety system, which caused a couple of tragedies.

And an autonomous vehicle is enormously more complicated, the number of sensors, the number of integrations of giant networks with billions of coefficients. We have to get it right. And it’s going to take a long time to get it right and piling on heuristics isn’t going to solve the problem. So it’s a poignant and important example for everybody to understand.

Connie: We are out of time. Alas, I wish we could talk much longer. Thank you so much, guys for coming. Really appreciate it.