Reality finally set in for Rony Abovitz in May. The augmented reality headset that had been under development for nine years inside his company, Magic Leap, had been a colossal flop. In a tearful address over video conference, Abovitz told employees that he would resign.

Abovitz, whose infectious optimism helped Magic Leap secure total investments of about $3.5 billion, didn’t stay down long. Just as his replacement, Peggy Johnson, was taking over, he tweeted that he was “working in stealth mode on something new:-).” The cryptic message was accompanied by a change to Abovitz’s Twitter bio referencing something called Project Phoenix. Abovitz, one could assume, is the mythical bird. That makes Magic Leap the ashes.

Magic Leap once burned bright. Many high-profile investors made the pilgrimage to a swampy, downtrodden suburb of Miami, where they became convinced Abovitz was building a kind of Apple for computers strapped to people’s faces. Private demonstrations of the technology, which made it appear as though digital objects viewed through the headset existed in the physical world, helped procure capital from China’s Alibaba Group Holding Ltd., AT&T Inc., Google and the chipmaker Qualcomm Inc. Magic Leap’s plan was to squeeze the technology down into a consumer device, construct a factory to manufacture it at scale, design an operating system, video games and films and spark the creation of a vast new content industry.

Doing it all, in Abovitz’s eyes, was the only way to outrun the corporate giants who also wanted to own the future. “It’s like if we were a coffee company, and we would have acquired a mountain and the soil and created the coffee bean in the particular climate and then made the roaster and controlled all the parameters,” Abovitz told Bloomberg at Magic Leap’s headquarters in Plantation, Florida, in 2018. “If you look at the best computing products, at the history of them, people that had hardware and software integration and understand the entire consumer experience, they built the best overall products.” He elaborated in a separate interview: “This is like Apple in 1978.”

At its peak—a peak that predated any public evidence of an actual product—technologists raved about Magic Leap’s potential. Shortly before he was appointed CEO of Google, Sundar Pichai joined Magic Leap’s board in 2014 and declared the product would “revolutionize the way people communicate, purchase, learn, share and play.” Pichai quietly withdrew from the board in 2018 and installed a Google subordinate in his place.

After Magic Leap’s $2,300 headset bombed, the startup narrowed its focus to professional applications, tried unsuccessfully to sell the company and fired more than half of its staff. Investors wrote down their stakes by an average of about 94% over a 12-month period ending in June, a steeper decline than WeWork, according to data collected by Zanbato, a research firm that tracks institutional investors.

The new CEO, Johnson, is trying to revive the business through partnerships. Magic Leap is engaged in discussions with Amazon.com about packaging the headsets with Amazon’s cloud services, according to three people familiar with the talks. The conversations are at an early stage and may not result in a deal. A spokeswoman for Magic Leap declined to comment, and Amazon didn’t respond to request for comment.

Abovitz responded to an interview request with a message consisting entirely of link to a research report, which estimates long-term growth in the augmented reality market. His spokesman later clarified that there would be no interview and referred subsequent questions to Magic Leap, which declined to comment. People familiar with Abovitz’s next project said it centers on building entertainment content for smartphones and augmented reality devices, including Magic Leap.

The co-founder’s departure came as little surprise to those who worked with him. Interviews with over two dozen people familiar with Magic Leap’s operations, including current and former employees, investors and business partners, suggest Abovitz’s world-building aspirations had become increasingly disconnected from the company’s reality. When employees found they would be unable to deliver on Abovitz’s vision, Magic Leap went from being one of the most intriguing tech startups outside of Silicon Valley to a parable about believing one’s own hype. The company’s failings reflect a broader struggle within the industry to commercialize a technology with so much promise. The question for Magic Leap is what parts of Abovitz’s dream, if any, can be salvaged.

Magic Leap was always Rony Abovitz. Cherubic and curly haired, Abovitz evokes a mix of Apple’s Steve Wozniak and Gene Wilder from Charlie and the Chocolate Factory. He spent his early childhood in the eastern suburbs of Cleveland before his extended family moved en masse to Florida in his adolescence. Abovitz studied biomedical engineering, drew cartoons and threw javelin at the University of Miami.

After college, Abovitz co-founded a medical robotics company named Mako Surgical Corp. in 2004 and helped take the company public in 2008. Two years later, he stepped down as chief technical officer and into the amorphous role of chief visionary officer. By Abovitz’s own account, he was drifting, daydreaming in the evenings and looking for something else to do. As a thought experiment, he began building a fictional universe called Hour Blue and soon saw it as a business opportunity. At first, Abovitz envisioned some sort of media venture, a film, perhaps, or a game. Then he became fixated on building one aspect of his fantasy world—augmented reality headsets—for people back on Earth.

Intrigue spread through the tech industry in the years after Magic Leap’s founding in 2011. Abovitz liked to keep it weird: A TEDx Talk in 2012 titled “the synthesis of imagination” consisted entirely of Abovitz wordlessly dancing around in a spacesuit alongside several people in furry monster outfits. Another video showed a life-size whale splashing through a school gymnasium. Investors flew from the Bay Area to Florida and signed strict non-disclosure agreements so they could strap on unwieldy prototypes. Then they wrote big checks and raved coyly about how they now understood the future of computing.

Abovitz drew people in with a nerdy charm and naked ambition, said people who worked with him. He alternates between bursts of inspiration and intense dissections of medical research or sci-fi movie plots in hard-to-follow monologues that many people find captivating.Magic Leap recruited at technical conferences by handing out cards that read, “Wizards Wanted,” and the staff was peppered with physicists, game developers, supply chain experts and at least one legendary science fiction novelist, Neal Stephenson. “It was a grand vision,” said Khizer Khaderi, a California-based ophthalmologist and entrepreneur who Abovitz recruited in 2014 as an adviser.

As a leader, though, Abovitz was divisive. He often pitted Magic Leap teams against one another, sending them to develop competing versions of similar ideas and then let the projects fall into stasis instead of choosing a winner, according to the accounts of over a half-dozen former colleagues, who requested anonymity to avoid alienating Abovitz or the historically litigious Magic Leap, which has filed several suits alleging trade secret theft. Abovitz had the final say on all matters but had little patience for details, they said. The company settled into a kind of paralysis, unable to either deliver on or dial back its ambitions.

Spencer Lindsay, an early employee, described his time at Magic Leap as complete chaos but expressed appreciation for Abovitz’s vision. “He was so kind to geeks and a believer in magic,” he said. Lindsay was fired in 2017 due to what he said were conflicting instructions about what his job was supposed to be. In the end, Lindsay felt Abovitz wasn’t equipped to run a company the size of Magic Leap. “Rony really believes in this and tried his hardest,” Lindsay said. “When it came to making concessions to reality, he didn't have that ability.”

In September 2018, Abovitz invited two Bloomberg reporters to Florida to see the Magic Leap factory. The companycommands a massive building in a drained stretch of the Everglades in Plantation, a town that gets its name from a failed attempt to plant rice fields in the early 20th century. The first stop was Abovitz’s office, a glass-walled room on the second floor. His interior design style could be defined as hobby-shop chic. There were models of spaceships on the coffee table, sketches of ray guns on white boards and shelves stacked high with action figures, sci-fi books and other knickknacks, including a half-dozen R2D2s from Star Wars, a Willy Wonka lunchbox and a photograph of an Apple I computer signed by Steve Jobs.

Abovitz, as he does, quickly slipped into fantasy when describing his management style. “A lot of tech startups, it’s like the Spider-Man movie, and everyone talks about Spider-Man. Everyone in my company knows that my goal is to distribute Magic Leap, to be more like an Avengers team.” When asked whether he had succeeded, Abovitz shrugged. “I actually think turning yourself into Spider-Man is kind of cool,” he said.

The primary purpose of the tour was to show off Magic Leap’s manufacturing facility, located a floor below Abovitz’s office in the part of the building he referred to as “little Shenzhen.” Any piece of consumer gadgetry is made up of many components from an array of manufacturers. The main processor in the Magic Leap 1 headset, for instance, is made by Nvidia Corp., and the chip that manages the device’s cameras comes from Intel Corp.

Abovitz insisted that Magic Leap make one component itself: the lenses used to enable the headset’s augmented reality features. Each lens, known as a diffractive waveguide, is etched with tiny grooves that redirect light across the surface before being steered into a wearer’s eye. Because the lenses are transparent, users can see both the physical world and the digital images the headset creates. Existing technology had limitations, and Abovitz thought Magic Leap could make some breakthroughs in secret.

He also really wanted to build a factory. His previous company, Mako, had its own production facility, and Abovitz saw it as critical to helping him understand how best to push the technology forward. (He’d also discovered that showing surgeons around a factory floor bustling with robots was an effective way to win their business.) In consumer electronics, though, components are typically the domain of specialty manufacturers, many of which are in Asia. Some of Magic Leap’s executives and investors questioned the wisdom of an expensive undertaking to make parts that could be purchased elsewhere, according to two people with knowledge of those discussions.

Abovitz went ahead anyway. He mined the staff of Tessera, an optics business in North Carolina, to build a manufacturing team. The first key recruit was Paul Greco, a soft-spoken engineer with a taste for loud Hawaiian shirts. Greco had spent 16 years at Motorola, where he helped run a smartphone production facility in Plantation. At Greco’s suggestion, Magic Leap moved into the old Motorola building. He then set out to gut and renovate the ground floor to establish an operation every bit as utilitarian as the rest of the office was whimsical.

Greco led a tour of the production lines, where workers in clean-room bunny suits operated machinery. At the time, he was preoccupied with adding a second assembly line that would be nearly autonomous, he said. “Rony loves it, because essentially we’re going to have the robots, like little R2D2s, that are literally going to move the materials around,” Greco said.

Magic Leap designed a supply chain fit for a multinational corporation. The lenses and other materials were built in Florida and shipped to Guadalajara, Mexico, where a partner assembled the headsets and sent them back to the U.S.

Abovitz ran Magic Leap as though it was already a serious rival to companies like Apple, which wasn’t always the cheapest or most effective approach. “I’m not sure why there wasn’t anybody who said you can’t build this out, and you can’t turn all this stuff on at this stage,” said an early Magic Leap investor.

Like many investors, this one had initially been won over by a dazzling demo, only to watch things go wrong, not just technically but culturally. For those on the inside, Abovitz’s vision had become a sort of blindfold to common business sense. “Everyone was so bought in, they’d drunk the Kool-Aid together. Nobody stopped to say, ‘This product sucks,’” the investor said. “The first time I put on the real one, it was just like, ‘Oh shit. You guys did not deliver on your promise.’”

As Abovitz was publicly declaring he would crush the Big Tech companies, he maintained a backchannel with nearly all of them, just in case. The CEOs of Apple, Facebook Inc. and Google each journeyed to Florida for discussions about a potential acquisition in 2016, according to the technology news site the Information. The talks with Apple progressed far enough that Abovitz flew to Cupertino, California, to meet with the company’s senior leadership, said a person familiar with the trip. Abovitz named the sale talks Project Batman.

Silicon Valley has long been seduced by the idea of augmented reality and has struggled nearly as long to make the technology useful. When Google’s attempt, Glass, flopped, it steered the product toward medical professionals. Microsoft Corp.’s HoloLens was an impressive gaming machine but unaffordable to most. Now it’s mostly targeted at businesses. Apple’s work on a product that combines augmented and virtual reality, code-named N301, has been ongoing since at least 2015, with more than 1,000 engineers dedicated to the project. Progress has been halting.

Magic Leap executives consistently bristled at reports of the company’s profligacy over the years, saying rivals were actually spending more but obscuring their projects deep within massive balance sheets. Still, Abovitz never figured out how to distill the magic produced by the enormous prototypes into a viable product, former employees said.

The headset Magic Leap released two years ago was based on a different technique, and its limitations were evident immediately. A narrow field of view meant that digital images had to be small or risk appearing cut off, and the headsets couldn’t be used reliably outdoors.The company found tricks to compensate. In one demo billed as a huge dinosaur wandering around an office, it created the effect by situating the creature at the far end of an open hallway so it would fit in the display.

The Magic Leap 1, with its steampunk-inspired design, did offer glimpses of potential. One app allowed people to drop small digital creatures around a room, where they’d adorably bump into the legs of coffee tables and fall off chairs. (They disappeared completely if you weren’t looking directly at them, then suddenly reappeared when turning your head back.) These gimmicks wore off quickly, though. Despite having spent heavily on content and touting collaborations with CNN, the National Basketball Association and Walt Disney Co.’s Lucasfilm, there wasn’t much to do with the device.

Experts in augmented reality hardware were underwhelmed, too. Palmer Luckey, the co-founder of Facebook’s virtual reality headset Oculus VR, dissected Magic Leap’s device and determined it was “the same technology everyone else has been using for years.”

Even engineers who used to work at Magic Leap questioned whether the company’s signature lenses will ever be fit for a consumer device because they use too much power and absorb too much light to operate outside of controlled environments.

As Magic Leap floundered, it began to re-examine its options. The company’s senior leadership seriously considered multiple acquisition offers late last year, according to people with knowledge of the deliberations. It decided to raise more capital instead.

The cash spigot that had been flowing to startups throughout Magic Leap’s lifetime slowed to a trickle this year when the coronavirus pandemic struck. Investors balked at Magic Leap’s requests for more money, angering Abovitz, who had said privately that he didn’t sell the company last year because investors assured him of their support, according to a person familiar with his thinking at the time. Magic Leap made another attempt to sell itself this spring and then resorted to layoffs. In his departing address to employees in May, Abovitz largely blamed the coronavirus, an assertion that rang false to some people in the audience, according to a person in attendance.

Somehow Apple, which doesn’t even sell a headset today, has emerged as the company most likely to own the mass market when it releases one as soon as 2022. For businesses, Microsoft is the clear leader.

That leaves Magic Leap and its new CEO in an awkward spot. If Abovitz was a visionary with trouble staying grounded, Johnson was there to execute. With two decades of experience at Qualcomm, Johnson was one of Satya Nadella’s first big hires after he was named Microsoft CEO in 2014. Johnson helped Microsoft repair relationships with companies like Salesforce.com Inc. and Samsung Electronics Co., which had frayed in the Steve Ballmer era.

Johnson started her job at Magic Leap last month and had to wait weeks before getting into the office due to the pandemic. She declined an interview request, and Magic Leap didn’t provide anyone else for an interview. She will likely focus on medical and industrial applications for the product, which have proven to be the only realistic near-term markets for augmented reality.

She’s also focused on making Magic Leap more competitive with the HoloLens, a $3,500 headset that Microsoft sells to manufacturing and medical institutions and that critics say is technically superior to what Magic Leap offers. The startup has a handful of corporate customers signed on, and Johnson’s first task is to close other deals in the works, including the one with Amazon, according to people familiar with the plans.

Some of the people who left or were dismissed by Magic Leap this year have landed at Apple and Facebook. Even several current employees acknowledge Magic Leap will never fulfill the goals Abovitz so convincingly pitched. Abovitz offered a different message on the day he resigned: “We have created a new field. A new medium,” he said. “And together we have defined the future of computing.” It’ll be up to someone else to make that future a reality.