Magic Leap One

» see magicleap.com as it lived, on the wayback machine

~ the eulogy ~

magic leap one launched in six US cities. i was not in any of them. the device cost $2,295 in 2018 dollars and weighed more than half a kilo on the head. it came in two parts wired together by a permanent cable. you could only buy it through magic leap's own white-glove delivery service that required a personal calibration appointment. that delivery model alone tells you the device was not for consumers, regardless of what the marketing copy said. by 2018 i was 34, doing computer vision and edge ai for a living, and i could read enough of the launch specs to know that the on-board compute and the field of view were not where the marketing had been promising they would be.

magic leap was founded in 2010 by rony abovitz, the same person who later became famous for being awkward at trade-show keynotes. the company raised over $2 billion across the next eight years from google and alibaba, with follow-on rounds also drawing fidelity and the saudi public investment fund. the pitch was a "lightfield" display that could place virtual objects in the real world with realistic occlusion and focus depth. the demo reels were vivid. the 2015 "whale jumping out of a gymnasium floor" video was viewed tens of millions of times. it was also not what the shipped product did, and journalists who got pre-launch hands-on time were under nda about exactly how different the real device was from the marketing.

august 8 2018 was the public launch. you could buy a magic leap one creator edition in six cities. the device weighed about 316 grams as a headpiece, with a separate puck-shaped lightpack computer connected by cable and clipped to a belt. magic leap had been forecasting hundreds of thousands of unit sales in the first year. according to later reporting, the actual first-year shipment count was around 6,000 units. the company spent the next year trying to pivot toward enterprise customers, and in april 2020 abovitz stepped down. peggy johnson, formerly of microsoft, took over and accelerated the enterprise pivot.

~ the dossier ~

~ what it was ~

the device was real hardware doing real optical engineering. magic leap's waveguide display approach was different from a hololens 1's holographic optical element approach, and the multi-focal-plane scheme tried to address vergence-accommodation conflict by rendering virtual objects at two different focal distances. that was a legitimate piece of work and it produced images that genuinely sat in front of you rather than floating on glass. the problem was the field of view. it was roughly 50 degrees diagonal, which left virtual objects clipped at the edges and broke the immersion that the demo reels had implied.

the developer story was the other half of the collapse. magic leap had been quiet for years about what the actual platform was. when the SDK shipped, developers found a system that required a custom OS called lumin along with a 3D content pipeline that did not map cleanly to unity or unreal at the start. the hardware audience did not exist in commercially meaningful numbers. the third-party app catalog never grew past a couple of hundred titles, most of them tech demos.

~ the wind-down ~

peggy johnson's enterprise pivot kept the company alive but the consumer story was over by 2020. magic leap 2 launched in september 2022 as an explicitly enterprise device, sold to companies for medical-imaging and industrial-maintenance use cases. magic leap one continued to be supported as a legacy device through 2024, including the cloud account system that the headset required to function. on december 31 2024 magic leap shut down cloud services and core functionality for the magic leap one. existing units kept on shelves became unusable. the second-hand resale market collapsed overnight.

~ what it left ~

the magic leap one is the standout hype-vs-reality consumer-AR product. between 2010 and 2018 magic leap had the biggest pre-launch raise of any AR startup along with the longest pre-release secrecy of any comparable hardware program. the magazine coverage in those years was more reverent than any other startup not run by elon musk. when the device finally shipped, it shipped to roughly 6,000 buyers in six cities at a price point that excluded almost everybody who had been excited by the demos for seven years. that gap between the size of the hype and the size of the audience is what every later AR launch (including apple vision pro) has been measured against. magic leap two is still being sold to enterprise customers as of 2026, but the consumer ar headset bet, in its 2010-2018 form, is over.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »