Bloomberg reported Thursday that Google’s parent company has put Boston Dynamics, a prominent robotics division, up for sale. According to the report, which cited anonymous sources, Alphabet executives think “Boston Dynamics isn’t likely to produce a marketable product in the next few years.”
Possible acquirers would include Toyota Research Institute and Amazon, the report said. Toyota Research Institute has been hiring in Cambridge, MA, while Amazon Robotics recently saw the departure of the founders of Kiva Systems, the warehouse-automation company that Amazon acquired in 2012 as the basis of its robotics division.
Google bought Boston Dynamics in 2013, along with a number of other robotics companies. The firm has been around since 1992; it was founded and led by Marc Raibert, a former professor at Carnegie Mellon and MIT. Boston Dynamics is known for its spectacular and haunting videos of humanoid and mule-like robots walking, running, and trying to keep their balance—including a recent video of a new humanoid robot walking in snow, picking up packages, and being pushed by a guy with a hockey stick.
It’s all good fun until Google needs to make money. And that’s the lesson here—if Boston Dynamics ends up being sold, it may be because Google higher-ups got impatient with the research and development timeline, while someone else was willing to invest in it. Though it sounds like there are culture-fit issues as well.
Last fall, I spoke with Jerry Pratt, a humanoid robot expert at the Institute for Human and Machine Cognition in Florida. Pratt’s team came in second place at the DARPA Robotics Challenge finals in 2015, winning $1 million in prize money. Like several other teams, they used Boston Dynamics’ Atlas robot in the competition.
Pratt said something interesting when I asked him whether he might start a new humanoid-robot company. (He previously co-founded a company called Yobotics in the early 2000s.)
“The big problem is there’s no minimally viable product [MVP] that I can think of,” he said. Unlike doing a “lean startup,” where you can make something small to get started and then iterate, he said, for a humanoid robot “the bar’s just so high—if you’re talking about competing against minimum-wage labor.”
Pratt said comparisons to the PC industry don’t hold up for this kind of robot. “In the early days of computers, a crummy one would save you a lot of money,” he said. “A really crummy humanoid robot’s going to lose you a lot of money. The economics isn’t there at all.”
Which is not to say robots aren’t worth working on, of course. It’s just that certain types “are still a tough sell,” Pratt said, when it comes to “what you want in a company.”
Boston Dynamics, meanwhile, has never really operated with an MVP-like mindset. The company has been relentless about pushing the state of the art in robotics and making progress toward robots that can behave and move around on legs like people and animals can.
So who might be interested in Boston Dynamics, if not Google?
The world of humanoid and legged robots is pretty small. Pratt did his PhD research in the MIT Leg Laboratory, which Raibert founded. (Disclosure: Pratt and I overlapped in the Leg Lab, circa 2000-2001.) His thesis advisor was Gill Pratt (no relation), a professor who went on to lead the DARPA Robotics Challenge and now heads up Toyota Research Institute as CEO.
It’s unclear whether Toyota would buy into Boston Dynamics’ approach, but its leadership certainly appreciates Raibert’s work and would have the money to support it. Toyota Research Institute says it’s working on autonomous vehicles, but may also pursue applications in areas like elder care and other types of mobility. (My money’s on Toyota over Amazon, unless the latter wants walking delivery robots as well as drones to replace the U.S. Postal Service—not the craziest idea, actually.)
Other potential acquirers might include Raytheon, which has done work on exoskeletons and other robots, and Teradyne, which bought Universal Robots last year. Another possibility is that Boston Dynamics will become independent again.
