Nearly two years ago, Google announced a new robotics division that had secretly snapped up more than half a dozen companies.
The state of those efforts is now in flux, and the group is in a difficult position as it tries to meet a goal of creating consumer robot technology by 2020.
Called "Replicant" internally, the initiative is a collection of roughly 10 acquired companies that insiders admit have little in common, with teams scattered across different countries and working on unrelated projects.
Replicant is an independent part of Alphabet, the new holding company that contains Google and all its newer businesses, but it hasn't had a permanent leader since founder Andy Rubin left in 2014 and it's grappling with questions about how to achieve Rubin's ambitious vision of creating the first wave of "consumer products that interact with the physical world."
Rubin told Business Insider that "the strategy was defined" when he left, but it seems like the group has lost much of its unification since then.
"The technology pieces we have are incredible," says one member of Google's robots team. "We just have to commit to a particular direction to go in and focus."
Alphabet has been in discussions with potential CEO candidates and is evaluating its options, the person said, but finding a person with the right combination of robotics expertise, business savvy and bold vision is no easy feat.
As Google parent company Alphabet seeks to build a collection of game-changing businesses in everything from hardware to healthcare, the robotics division is a reminder of the challenges facing the organization's new bet on its moonshot projects.
Business Insider spoke to several current and former Googlers, as well as robotics industry insiders, to provide a window into one of Alphabet's most ambitious and risky bets.
Google declined to comment on this piece.
How Replicant got started
Replicant's existence within Google and its current predicament are both tied to its founder, Andy Rubin.
Rubin joined Google in 2005, when Google bought his smartphone company Android. For the next 10 years, he slowly built Android into an enormous business that had sold more than 1 billion handsets to people all over the world.
Known for being extremely demanding, as well as a technical genius with a sharp business acumen, Rubin stopped leading Android in March 2013, and he and CEO Larry Page decided on a new mission: Rubin would create Google's new robotics division.
Page is interested in robots — we once heard a story about how at one of his extravagant birthday parties, Page enlisted a robotic arm that chucked party favors to guests — and had a broad vision of creating general purpose bots that could cook, take care of the elderly, or build other machines, but the actual specifics of Replicant's efforts were all entrusted to Rubin.
Rubin has said that Page gave him a free hand to run the robotics effort as he wanted, and the company spent estimated $50 to $90 million on 8 wide-ranging acquisitions before the end of 2013. Rubin led nine acquisitions total.
Among others, Google bought Schaft, which makes humanoid robots, Boston Dynamics, whose animal-esque bots have incredible indoor and outdoor independent locomotion abilities, and Redwood Robotics, whose robotic arms are more suited for factory floors.
If the assortment sounds disjointed, that's because, in some ways, it was.
One employee tells Business Insider that the acquisition strategy relied on the fact that building a robot requires excellent software systems combined with state-of-the-art hardware. So, Google started snapping up companies that had top entrepreneurial talent and unique accomplishments in each one of those areas.
Google has been tightlipped about what exactly it hoped to achieve with the robots, and there have been reports that the company was looking at everything from factory automation to robot assistants.
According to an email Rubin sent the team, which was viewed by Business Insider, the overarching goal was to create and release consumer-oriented robotic technology before 2020.
"We have built a world-class organization that is well on its way to launching its first products based on robotic technologies," Rubin wrote when he left in 2014.
Before 2020, he wrote, "this team will take more than 20 years of robotic research and launch a suite of 1.0 products that will be the foundation for future consumer products that interact with the physical world."
When the guru leaves...
But when Rubin left Google to start his own hardware incubator, Playground Global, in October 2014, it came as a huge blow to the team.
Replicant cofounder James Kuffner had been managing the research and development side of things, but was thrust in charge of the entire division when Rubin left. Although Kuffner is a long-time expert in cloud-based and humanoid robotics, he didn't have the product development experience that his former boss had.
After Rubin left, many of the people who had joined Google through the robotics M&A felt confused and disappointed, according to several sources. "The division was harmed," one person said.
Marc Raibert, CEO of Boston Dynamics, by far the most high-profile company Google bought, felt especially taken aback, we're told. (Raibert declined to comment on this piece.) Like many others, Raibert believed in Rubin's leadership and vision.
Rubin defends his decision to leave, telling Business Insider via email that "I left after the team was assembled and the strategy was defined."
While he insists his departure wasn't a shock to those who knew him well, he concedes that the nine companies he acquired were "probably the most surprised."
"I personally pitched each and every one of them on the strategy and convinced them to join this crazy idea," he says. "So they saw me as their sponsor, and perhaps seeing your sponsor move on was difficult — I don't know."
Sources say that after Rubin left, there was no one who knew how to tie all the disparate acquisitions together. Rubin had provided a destination, but the team no longer had a roadmap or a guide to get there.
Rubin takes issue with the point, noting that he felt confident in the three people who had "volunteered to take on everything" when he left.
"I am known to be somewhat impatient and I had a nervous energy," he says. "The last thing you want me doing is hovering over a bunch of brilliant roboticists and asking them 'Are you done yet?'"
Kuffner remained in the business leadership role until earlier this year, when Johnathan Rosenberg replaced him.
But Rosenberg, a long-time Google vet who helped lead Motorola and wrote a book about management with Eric Schmidt, is another placeholder while the company searches for someone who is a better fit.
To succeed Rubin, the division needs someone with an intense robotics and research background combined with robust product experience. That was Rubin’s strength: He was the kind of person who, for fun, had built a giant robotic arm to make him coffee whenever he sent it a text message , but had also shipped more than a billion Android smartphones.
Even though the company has been in discussions with candidates and evaluating options, Rubin simply isn’t an easy leader to replace.
One former high-ranking Google executive says the robot group is a "mess that hasn't been cleaned up yet." The robot group is a collection of individual companies "who didn't know or care about each other, who were all in research in different areas," the person says. "I would never want that job."
Incredibly hard problems
Alphabet's collection of robotic companies — which a current employee describes as the "crown jewels of the industry" — still have the opportunity to work on interesting problems and challenges.
Replicant is trying a lot of different things in parallel and learning along the way, with the capabilities of each of the robotics groups advancing.
And while Replicant is Google's largest dedicated robotics technology investment, machine learning and computer vision efforts led by senior researcher Jeff Dean and artificial intelligence developments through Google's DeepMind are applicable to robotics as well.
But, still, some Replicant employees feel frustrated that the teams are moving forward in relative isolation. Although many of the robotics employees are based in the Bay Area, Google also has factions in Massachusetts, where Boston Dynamics is based, and Toyko, because of Schaft.
Several of these teams have lost employees in the last year, according to a LinkedIn search, and some employees crave more collaboration.
A person who worked at one of the robotics companies that Google acquired told Business Insider that Amazon was also eyeing the business before the search giant officially sealed the deal. That person would ultimately end up wishing that Google hadn't won the bidding war.
"Google seems more aimed at a moonshot while Amazon is building a ladder to the moon," the person said, frustrated by the idea that if Amazon had acquired the company, the technology would likely already be loose in one of the ecommerce giant’s fulfillment centers, instead of plodding along in R&D.
Not a "robot army," by any means
Those who work in the robots group believe it is tackling one of the world's most difficult technological problems.
If computers are the most complex things that have been created by humans, a robot is that much more complex. You need to connect very advanced software and artificial intelligence with an incredibly precise mechanical device with motors, sensors, and the ability to recognize the world around it.
During the heyday of coverage of Google's robotics acquisition, publications casually suggested that Google was building a "robot army." And, when you watch the videos of Boston Dynamic's crazy robots in action, it’s easy to feel vaguely terrified by the idea of Google building a herd of robot dogs.
The fact that Boston Dynamics is a military contractor didn't help allay those fears. Google has since said it will honor existing Boston Dynamics contracts with the Department of Defense, but will not pursue new military business.
But the broader thrust of the group is not about building a Terminator. There have been a lot of interesting experiments around the idea of robots helping humans, whether through manufacturing, in the home, or with logistics. Google will eventually share some of the results of experiments, even if it isn't close to releasing a product, according to the robot group insider.
Factions of Replicant also work with universities, like MIT or Carnegie Mellon, where Kuffner is an adjunct professor. For example, the Meka Robotics team that Google acquired has collaborated with Stanford University to find new ways to explore the Red Sea, by providing hardware to develop underwater robot arms.
"We’re optimistic about the technology, and that it’s going to help people," the robot group insider said.
The ripple effect
For now, the biggest legacy of Google's effort may be the impact on the close-knit robotics industry.
Some say that Google's acquisition spree breathed life into the community. One insider says that Google's investment helped stimulate the industry in the same way that Google's self-driving car project has catalyzed similar efforts at other companies.
For a while, the private sector didn't want to touch robotics because it was seen as too expensive without any short-term uses. Most progress came from either the government, through DARPA, or in university settings. Existing robots worked in factories or fulfillment centers. But Google's investment and its acquisition spree accelerated more private and consumer-focused technological development (and piqued the interest of niche robotics companies looking for a viable exit strategy).
"It felt like a frenzy," David Cann, CEO of telepresence company DoubleRobotics, tells Business Insider. "That was a very exciting time for robotics. It really validated the space."
But Google's embrace of robots has come at a cost.
A former Boston Dynamics employee tells Business Insider they have heard from people inside Google who feel frustrated by how their once open culture is now shrouded in secrecy.
When the Schaft robot that Google bought in 2013 had to drop out of DARPA's robotics challenge last year, it gave some the impression that the company was impeding overall robotics progress. The goal of the competition was to accelerate the development of robots to help during disaster situations.
Perhaps even more important are the expectations now resting on Alphabet's high-profile robot effort.
If Alphabet doesn't figure out how to make money in this space and the robot company winds down, it will send a bad message to the market, the former Boston Dynamics employee said.
"I hope it works out well," the person said.