Founded in 1992, Boston Dynamics is arguably the best-known robot company around, in part because its demonstration videos tend to go viral. Now it is attempting to transform from an R&D company to a robotics business, with an eye on profitability for the first time.
When we interviewed Boston Dynamics founder and former CEO Marc Raibert in November 2019, we discussed the company’s customers, potential applications, AI, simulation, and those viral videos. But it turns out Raibert was transitioning out of the CEO role at the time — current CEO Robert Playter told us in an interview this month that he took the helm in November. We sat down to discuss Playter’s first year as CEO; profitability; Spot, Pick, Handle, and Atlas; and the company’s broader roadmap, including which robots are next.
Almost a year as CEO
Boston Dynamics hired Playter in 1994. After 18 years as vice president of engineering, Playter took a director role when the company was acquired by Google. Following Google’s sale of Boston Dynamics to Softbank four years later, he became COO and then a year later, CEO.
“The company is transforming,” Playter said. “There’s a lot of change that’s taking place. And we’re trying to do some things that are pretty hard. Our history is we’re an R&D organization. And really, we’re the best in the world at creating some new concept of a robot and making it work well enough that you could go do a demo with it. And we are trying to hang on to what made us great there: the ability to do advanced development and research and solve fundamental and hard problems but also develop our commercial muscle, learn how to sell a product, manufacture it, support it, and ultimately make a profit. It’s not always easy to have both of those kinds of goals coexist.”
To help Boston Dynamics become a commercial company, Playter has been hiring executives with experience in sales, business development, marketing, HR, and finance. The engineering leadership has come from the R&D realm and will remap how the robots are produced. But the majority of the new work is happening on the business side, where the company is “developing the processes and the discipline to go build a profitable business.”
Spot sales and profitability
In June, Boston Dynamics started selling its quadruped robot Spot in the U.S. for $74,500. Last week, the company expanded Spot sales to Canada, the EU, and the U.K. at the same price point. Playter says Boston Dynamics has sold or leased about 250 robots to date and business is accelerating. As part of the early adopter program, the first 120 robots took about seven months to sell. Another 120 followed in three and a half months, essentially doubling the rate of sales. Next year, the company plans to give Spot a recharging station and robot arm.
Compared to big manufacturing robotic companies, 250 robots is not a lot. But Playter points out it’s a big achievement “for a novel robot like Spot.” Other robotic startups would love to get that sort of market validation. “We’re penetrating, we’re establishing a market, and people are starting to see value. We’re adapting Spot to be a solution for some of the industries we’re targeting,” Playter said.
Spot’s success means the company is beating its own internal targets. “We are meeting — actually exceeding — some of our sales goals for Spot,” Playter said. “We had ambitious goals this year, but we met our Q1 goal. We’re meeting our Q2 goal. We have ambitious Q3 and Q4 goals. I think we’re probably going to meet or exceed them this year. To become profitable, these products do have to become successful. They have to scale. But right now, I think we’re beating plan.”
The company now has a roadmap to profitability. “I think we’ll be profitable in about two and a half years,” Playter said. “2023-2024 is when I’m projecting that we are cash positive.”
To hit that milestone, Boston Dynamics is simultaneously developing robots for logistics (think production, packaging, inventory, transportation, and warehousing). “And that, frankly, is gaining a lot of attention from the logistics industry,” Playter said. “I am confident that’s going to succeed. That sort of business focus is sort of a new thing for us. But I think that part is going really well. But can we do that and still have the Atlases of the world doing backflips and pushing the boundaries and having all of that coexist in one company? So far, so good. That’s the challenge in front of us, I think, though, to allow these differing skills to coexist in one company.”
Handle slated for 2022
In our interview with Raibert, he categorized the company’s next three robots by time: today (Spot), tomorrow (Handle), and the future (Atlas). Raibert called Handle the “tomorrow” robot because at the time he expected it to ship in 2021. The company has pushed that date back by at least another year.
“We’re building a version of Handle which we will launch in 2022,” Playter said. “We have customers lined up that we’re doing pilots with, essentially prototypes of those machines now. But we’re designing the version for scaled manufacturing, and we expect those first systems to be available in about 2022.”
The timeline appears to have slipped, but Playter doesn’t see it that way.
“Handle hasn’t really been delayed,” Playter said. “We’ve got a new design of Handle. We decided we need to change the design before we commercialize it. So that’s what’s going on. I wouldn’t say it’s delayed. I would say that we sort of rethought exactly what we wanted to do there. And so now we have an iteration on that design, which we are beginning to prepare for manufacturing. And it takes time. To really design something for manufacturing and the reliability you need, it takes a couple of years.”
Boston Dynamics isn’t yet ready to share what the redesigned Handle looks like. The company will make that public “sometime in 2021.” Broadly speaking though, the design change will make the robot “faster and more efficient in a logistics setting,” Playter promised.
Boston Dynamics is really gunning for logistics next. “The opportunities in logistics are large, and we’re going to have the first mobile case-picking robot that can pick up and put down boxes, whether it’s in the back of a truck or in your warehouse or at the end of a conveyor,” Playter said. “Basically, any of the box-picking tasks that are sort of ubiquitous in a warehouse, I think Handle will be able to do.”
Pick for now
But that’s in 2022 at the earliest. Until then, Boston Dynamics is selling Pick, a depalletizing vision system and computer that costs $75,000. Pick is not a robot — it needs to be attached to existing commercial robots. Companies sell integrated robot setups that use Boston Dynamics’ Pick for between $200,000 and $400,000.
Handle is the mobile version of Pick. So, will the former kill the latter? “I think it will have its place in the world, but I do believe that in the long run having a robot that’s mobile that can do what Pick does will end up superseding Pick,” Playter said. “The thing is, there’s need for that application now. Ultimately, the market for a fixed position, depalletizing or palletizing robot is limited because there’s only so many places that you get enough flow of product coming through that you want a robot to do that. But if you have a mobile robot, it can now do that job in several places.”
A mobile robot is naturally superior to a stationary one, but there’s a cost trade-off. “Yeah, but I don’t think cost will ultimately be that much higher,” Playter said. “There is more complexity in our thing, for sure. But I think we’ll be able to manage that cost.”
Pick may not be a robot, but it’s Boston Dynamic’s first product in the logistics space, and it’s getting traction.
“The purpose of Pick is to have an advanced machine learning-based vision system that lets you look at a pallet so that a robot can go pick up boxes that it hasn’t seen before, or what’s called mixed-SKU pallets — pallets of different kinds of product,” Playter said. “And it takes an advanced vision system to be able to deal with that. But we’re beating out more established competition in some early sales. And the neat thing is the vision systems with Pick are going to be the same ones that we use with Handle.”
When Handle launches, it will do the same task in mobile robot form that Pick does bolted to the ground. In this way, Boston Dynamics is establishing credibility with Pick before the full mobile robot is ready.
Atlas is still the future robot
The robot that arguably gets the most buzz is Atlas, which Boston Dynamics uses to test and build ideas. Once proven, they get pulled out and put into existing robots or turned into entirely separate robots. Indeed, the techniques that let Handle move and balance its upper body were first developed as part of Atlas, even though Handle has wheels and Atlas is a humanoid robot with two legs. The team also built its own valves and high-performance control systems for all the hydraulics in Atlas. That knowledge was then used to build a new pneumatic gripper for Handle, which Playter said exceeds the performance commercially available today.
“Atlas we’re not really envisioning commercializing, at least not yet,” Playter said. “Atlas remains an aspirational robot that really forces us to advance the state of the art in both software and hardware design. And so we still just have a few Atlases. We use that to motivate our R&D work really. People identify with the humanoid, obviously, so that makes it interesting. It’s a complex robot, and so it forces you to develop techniques to deal with all of the degrees of freedom that are not traditional techniques.”
The Atlas team has recently done some work to author behavior software on Atlas much more rapidly. What used to take six months to code, the team can now do in a few days, thanks to advanced optimization tools. “And those tools will become available really to all of our machines, but we’re using Atlas as the way of motivating the development,” Playter said. “But Atlas is too expensive and too complicated to commercialize anytime soon.”
Breaking down the various teams
There’s a lot going on inside Boston Dynamics. To put it into perspective, we asked where the company puts its focus. Playter laid out his answer in terms of team sizes.
The Spot team is about 100-110 people. The Handle and Atlas teams are about 70% and 20% that size, respectively.
“So Atlas is small,” Playter said. “It’s a research team, sort of the critical mass needed to create these interesting behaviors and performances. Handle is rapidly growing because we’re preparing to launch a product. It’s following sort of a similar trajectory [to the one] that Spot did.”
Boston Dynamics needs about 20 or 30 people to prototype a robot. Playter thinks the Handle team will grow to about the same size as the Spot team “over the course of the next two years.”
At that point, “our success in achieving profitability depends on being able to scale that product as well,” Playter said. “But there’s a large opportunity there that I think will lead to growth in the sales of that robot as well. I expect to have two successful products by the time we get there. Hopefully, we’ll start building that third one that you asked about.”
The next robot
Boston Dynamics’ near-term focus is to expand sales for Spot and Pick/Handle. But there’s room for more robots in the long term.
“Our ambition is really to take the lessons that we’ve learned over decades of R&D work in terms of how to build lightweight, highly mobile manipulation machines and apply them to a bunch of different industries,” Playter said. “I foresee that once we really get Spot totally established as a successful product that we can probably redirect some of that team to go building the next product that might be something specific for, I don’t know, construction or forestry. There’s lots of potential applications out there that I think we could go build other bespoke robots for, that are still general-purpose mobile manipulation robots but probably need to be different than the ones we have. I would like to go build a series of these things.”
Spot may be modular, but even modularity has a limit. That’s why Handle exists. Playter believes Boston Dynamics ultimately needs to build more robots in different form factors.
“In the construction industry, there’s lots of big heavy things that need to be held in place,” Playter said. “There’s a lot of two-person lifting jobs that might be able to be a one-person lifting job if you had a robot helping to carry some of that load, or something like that. These are confined spaces. These can be big and heavy. It can be drywall. It can be HVAC equipment. It can be ducting. You can’t really use robotics in those environments now because the robots are typically fixed-position robots, and they’re big and heavy. But what if you could build a robot that was light enough to be mobile on a construction site but big enough to pick up, I don’t know, a piece of drywall and hold it up while somebody else screwed the screws or managed the whole process? That’s the thing I envision, but Spot is not going to do that, and neither is Handle. It needs to be a bigger robot that can handle that kind of stuff but could still be mobile.”
Robots in the home
In June, Raibert revealed that the company wants to sell Spot for the home “someday.” We asked the new CEO on the thinking there.
“We really started with business applications because we knew that the robot is still an expensive proposition for a personal consumer, I think,” Playter said. “You can’t build a robot with the capabilities that Spot has at the price point of, I don’t know, an autonomous vacuum cleaner or something like that. So we really started with business because we knew it was going to need to probably exceed the price of the consumer. The value of the robot is clear and makes sense from a return on investment kind of application.”
The big exercise was to figure out how to reliably build a robot at scale, including manufacturing, service, and support. As the company gets better at that, it has to figure out how to keep driving costs down.
“I think that the vision of eventually having a robot for the home is still strong,” Playter said. “I think it’s going to take a little while to get there because I think the cost needs to be for some personal consumer-type thing in the few thousands of dollars. And we’re obviously in the several tens of thousands of dollars now. But I think everything that we will learn in terms of the capabilities of the machines and producing them and supporting them is all providing the foundation for eventually doing something at sort of that consumer level.”
Price and value proposition
As with any consumer product, price is key. We asked Playter what his target price would be for a home robot. “Well, certainly below $10,000. Go look at your robotic vacuum cleaners that are $1,000.” But reaching that price point would be easier said than done. “I think it may be hard to get Spot to $1,000, but I think several thousand dollars might be feasible at large enough scale,” Playter said.
The value proposition, which could be anything from entertaining humans to helping them, remains an open question.
“Honestly, I don’t know yet,” Playter said. “I’m not really focused on that. I think that’s a ways off still. To be honest, I’m focused on the industrial applications of the robot we have now and the other robots that we’re building for logistics. That’s really where I’m focused. Part of the larger picture for us is we aspire to build these transformative robots, and I think it’s a great vision to imagine building a less expensive version that people would want in their home, but honestly that’s a little bit over the horizon for us.”