In 2012, a small team of Google Inc. engineers and business staffers met with several of the world’s largest car makers to discuss partnerships to build self-driving cars.

In one meeting, both sides were enthusiastic about the futuristic technology, yet it soon became clear that they would not be working together. The Internet search company and the automakers disagreed on almost every point, from car capabilities and time needed to get it to market to extent of collaboration.

It was as if the two were “talking a different language,” recalls one person who was present.

As Google expands beyond Web search and seeks a foothold in the automotive market, the company’s eagerness has begun to reek of arrogance to some in Detroit, who see danger as well as promise in Silicon Valley.

For now Google is moving forward on its own, building prototypes of fully autonomous vehicles that reject car makers’ plans to gradually enhance existing cars with self-driving features. But Google’s hopes of making autonomous cars a reality may eventually require working with Detroit, even the California company acknowledges. The alternative is to spend potentially billions of dollars to try to break into a century-old industry in which it has no experience.

“The auto companies are watching Google closely and trying to understand what its intentions and ambitions are,” said one person familiar with the auto industry, who asked to remain anonymous because of sensitive business relationships.

“Automakers are not sure if Google is their friend or their enemy, but they have a sneaking suspicion that whatever Google’s going to do is going to cause upheaval in the industry.”

No Steering Wheel

Analysts estimate Google has invested tens of millions of dollars in an effort that’s ultimately a side project. But car companies, all too familiar with the devastating financial and brand damage of recalls, would see any hiccups with the self-driving car as a threat to their main business.

Nowhere is the disconnect more evident than in Google’s latest prototype. Two people sit abreast in the tiny pod-shaped car, which has a flexible windshield for safety and is topped by a spinning cone that helps navigation. The electric vehicles, unveiled in May, are limited to a maximum speed of 25 miles per hour and do away with several decades-long constants in motoring: the steering wheel, brake pedal and accelerator pedal.

Google co-founder Sergey Brin has described self-driving cars as an on-demand service that consumers summon when needed. That would represent a seismic shift from a longstanding model based on individual ownership, an annual $375 billion U.S. market according to J.D. Power.

Moreover, a study by consulting firm KPMG last year found that American consumers would trust brands like Google and Apple more for self-driving cars than they would automakers.

General Motors’ global product development chief Mark Reuss recently said Google could become a “very serious competitive threat.”

Evolution vs. Revolution

Chris Urmson, director of Google’s self-driving car group, would not discuss any negotiations with automakers but argues that self-driving cars will benefit car companies and consumers by expanding the number of car users.

“I’m confident that when there is technology that makes sense, and when there is a business model that makes sense, that there will be interest and partnerships” with car makers, Urmson told Reuters in an interview.

Self-driving cars can free people to do more of the things that earn Google money, such as Web search. But Urmson said Google is still figuring out how to make a profit from the technology.

“I would imagine that this is probably different than just making more time for people to click on web sites,” he said.

Car makers such as GM, Mercedes and Volvo have been developing their own autonomous vehicle technology for years.

But most favor an incremental approach to self-driving cars, in which features such as lane centering and parking assistance are gradually integrated into vehicles. Car makers are also hesitant to invest in new features until they are certain there is enough demand to pay for them.

That approach and car makers’ long development process are at odds with Google’s ambition to create a fully autonomous car in one swoop. The Internet company seemed to have little patience for Detroit, according to people involved in the 2012 talks with automakers.

“There was a certain amount of arrogance on the Google side, in the sense of ‘We know what we’re doing, you just help us,'” said a second person, representing a major car maker, who was involved in discussions with Google.

“We’d say, ‘Well you don’t really know that much. And we’re not going to put our name on a project like that because if something goes wrong, we have a lot more to lose.'”

Another potential sticking point is maps developed by Google and essential for its robo-cars to operate, says Sven Strohband, a robotics expert who worked at Volkswagen until 2006 and was not involved in the discussions. That data, compiled by Google, can be extraordinarily detailed, down to the height of curbs or location of signs.

“The question is who owns the data,” he said. “You need to have frequent map updates and your car can only go where you have really accurate map data.”

Without a driver to blame when accidents happen, the vehicles could bring greater liability for car makers.

Google’s assurances to one car maker that it would take responsibility for accidents due to its technology, and that the data collected by the cars makes it easy to pinpoint fault, was dismissed, according to the first person involved in the 2012 discussions.

“I just couldn’t believe my ears and was like ‘Wow you live in a bubble,'” the person said. “Car makers never get to decide who is at fault. It’s the lawyers, the judge and the jury.”

Starting Small

Whether Google opts to license its technology or seeks to build cars to its specifications, Google will need Detroit for the last mile, say industry experts and insiders.

Google has made headway in less sensitive areas such as entertainment and navigation. In January, Google teamed up with GM, Audi, Honda and Hyundai to form the Open Automotive Alliance to incorporate its Android operating system, the software for mobile phones and tablets, into cars.

And it has taken steps to understand regulations better, hiring Ron Medford, the National Highway Traffic Safety Administration’s former Deputy Director, in November 2012.

“My view on this is both parties probably need each other,” said Strohband, now Chief Technology Officer at venture capital firm Khosla Ventures.

A source at one automaker said the company talks to Google on a weekly basis about auto matters, though they have not partnered on self-driving cars.

Some in the industry predict fully automated cars will be available as soon as 2020, though research firm IHS Automotive does not expect the cars to be widely available until 2035. For now, Google is starting small with 100 to 200 prototype cars. It wouldn’t identify manufacturing partners, though industry reports pinpoint Michigan-based Roush Enterprises, which assembles small volumes of custom vehicles such as race cars. Roush declined comment.

To build anything more than a couple thousand cars would likely require an automaker partner. Industry insiders point to critical systems such as steering and suspension, the intricacies of working with hundreds of suppliers and high-volume production at consistent levels of reliability as skills that cannot be learned overnight.

While Tesla Motors offers an example of an outsider breaking into the business, the electric car maker has benefited from a hefty government loan and from having access to the shuttered GM-Toyota NUMMI car manufacturing plant in Fremont, California.

The cost to launch a new car model, including costs of developing and tooling, is generally $1 billion to $1.5 billion. For a company starting from scratch, such as Google, that cost would likely be higher, say auto industry experts.

Some industry observers have suggested that Google should pair up with Tesla, which is also developing self-driving technology and which shares Google’s Silicon Valley mindset. With roughly $60 billion in cash, Google could also acquire a smaller auto company, some speculate, though they note that such a move would involve more ongoing costs, liabilities and cultural challenges then Google may be willing to accept.

“Google is the 800-pound gorilla in the room and nobody wants to miss the boat,” said Edwin Olson, assistant professor of computer science at the University of Michigan, who works with Ford on an automated vehicle project. “But at the same time I don’t think automakers want Google to be dictating terms if the time comes and Google is the only game in town.”

(Additional reporting by Deepa Seetharaman in San Francisco, editing by Edwin Chan and Peter Henderson)