Magic Leap Is Cutting 193 Jobs and Getting Out of the Headset Business

Magic Leap filed a WARN notice in July. It cuts 193 jobs at the company's Plantation, Florida headquarters, effective October 1. The affected list runs through software, hardware, UX and design, product management, manufacturing engineering, quality, technical program management, and a set of senior engineering and executive roles.

That is not a trim. That is the product organization.

What Magic Leap keeps is the optics. The company is repositioning as a waveguide supplier and device integration partner for other firms building AR glasses. It wants to sell the hardest part of the display stack to people who will put their own name on the finished product.

Read that as a failure if you want. The more useful read is that Magic Leap already told everyone this was coming, and almost nobody treated it as news.

The 2024 deal was the announcement

Magic Leap partnered with Google in 2024, then extended it to a three-year agreement. The joint reference glasses weigh under 50 grams in a monocular design. Google supplies the Raxium microLED light engine. Magic Leap supplies the waveguides and the optics.

Look at what each side contributed. Google brought the platform, the light engine, and the ecosystem. Magic Leap brought a piece of glass. That division of labor was public for two years before the WARN notice, and it described the company's actual product with more precision than any product page did.

The layoffs did not change the strategy. They removed the people who were still executing the old one.

Waveguides are the bottleneck, and the bottleneck is where the margin is

Every set of AR glasses needs a way to get an image from a tiny projector into a lens and then into an eye. Diffractive waveguides do that job. They are also brutally hard to manufacture at yield, which is why so many AR products are demos that cost more than a used car.

Magic Leap uses Jet and Flash Imprint Lithography to make surface-relief gratings. The pitch is fewer process steps and faster cycles than competing methods. Yield is the entire game in optics, and process steps are where yield goes to die.

Kevin Curtis, the company's SVP of advanced optics, put the case plainly. "Waveguide performance is what separates AR that consumers want to wear from an expensive demo." Scott Carden, SVP of display engineering and manufacturing, framed the pivot around scale: "We're solving the toughest challenges of scaling waveguide production, from industry-leading waveguide performance to manufacturing and full-device integration."

Both quotes are supplier language. Performance, manufacturing, integration. Nobody at a company that still intends to ship its own headset talks like that in public.

What four billion dollars bought

Magic Leap has raised more than $4 billion since Rony Abovitz founded it in 2010. Saudi Arabia's Public Investment Fund is the majority shareholder and has put in over $1 billion on its own. Alibaba, Qualcomm, and Google are on the cap table. Abovitz left the CEO job in 2020, when the company first turned away from consumers.

Two headsets shipped in fifteen years. Magic Leap 2 arrived on September 30, 2022 at $3,299 for the base tier, rising to $5,000 for the enterprise configuration. It undercut Microsoft's HoloLens 2 by $200, which tells you the market it was actually competing in, and it never reached the one it was funded for.

Selling components is not the business anyone pitched in 2016. It is, however, a business.

The pattern is getting hard to miss

This is the third version of the same move in a month. EssilorLuxottica bought Lynx for the software and shelved the headset. Lenovo folded its XR business into AI wearables. Now Magic Leap keeps the optics team and lets the product team go.

In all three cases the valuable asset turned out to be a capability, not a device. Companies are separating the two and keeping the half that other people need to buy. That is what a component market looks like while it forms.

What it means if you are buying

Enterprise customers running Magic Leap 2 should ask for support commitments in writing, and soon. The WARN list includes the hardware, quality, and program management functions that keep a shipped device alive. The company has not said the headset is finished. It has laid off a large share of the people who would keep it going.

For everyone else, the practical effect arrives later and wearing someone else's badge. If a pair of AR glasses ships in 2027 with a display that finally looks worth the money, there is a reasonable chance the waveguide inside came from Plantation, Florida. The name on the temple will belong to a company that never had to raise four billion dollars to learn how to make it.