The chipmaker announced two new offerings for AI wearables: a mixed reality platform and a turnkey kit for smart glasses. CEO Cristiano Amon says more than 40 designs are in the pipeline.
Qualcomm, the company best known for the modem in your phone, is now publicly betting that the next major computing platform is something you wear, not something you carry. On Tuesday, CEO Cristiano Amon told CNBC the chipmaker has more than 40 AI-wearable designs in development, spanning jewelry, earbuds with cameras, pins, and watches. The pipeline figure, not a shipping count, is the clearest signal yet that Qualcomm wants to be the silicon and software layer for whatever comes after the smartphone.
Qualcomm used the day to announce the two pieces of that bet. The first is Snapdragon Reality Elite, a new chip platform for mixed-reality glasses that succeeds the company's XR2+ Gen 2 line. The second is the Scalable Turnkey AI-Ready Toolkit, or START, a bundle aimed at hardware makers that want to ship AI glasses without building the stack from scratch, as TechCrunch reported.
Snapdragon Reality Elite is built for two distinct form factors. Standalone headsets use video see-through, or VST: cameras on the device capture the world, and the headset overlays digital images on top of that feed. Tethered glasses use optical see-through, or OST, blending digital graphics into the actual see-through lenses. The distinction matters because it determines what the user sees, how heavy the device gets, and how long the battery lasts. The platform supports 4.4K resolution per eye at 90 frames per second, up from 4.3K per eye on its predecessor.
The concrete capability Amon is anchoring on is local AI. Qualcomm says the platform can run a 3-billion-parameter language model on-device, meaning the AI runs directly on the headset rather than in a remote data center, at 45 tokens per second, the standard measure of how fast a model produces text. Local inference matters on a face-worn device because it cuts the round-trip delay of reaching a server and keeps personal sensor data from leaving the device. Qualcomm is also claiming up to 60 percent higher graphics performance, 30 percent higher CPU performance, and 160 percent higher neural-processor performance than the prior generation. Those figures come from Qualcomm's own benchmarks, not independent testing, and they should be read as a marketing claim until third parties put them on a test bench.
START is the strategically more interesting half of the announcement, because it is what turns a chip into an ecosystem. The package bundles an AR chip, a software platform, companion apps, and a white-label program, meaning hardware makers can sell products built to Qualcomm's design under their own brand. Three reference designs are available: an audio-plus-camera frame modeled on Meta's Ray-Ban smart glasses, a monocular display template with a single lens, and a binocular display template with two. Inspecs and O'Neill, owned by eyewear group TitanFlex, are the first named white-label partners, with Qualcomm saying the toolkit will expand to other form factors later.
The first shipping signal is XREAL's Project Aura, which Qualcomm showed off at Google I/O. A second device, from Play for Dream, is on the roadmap. Both names are still announcements, not store shelves, but they show the platform is moving from slide deck to a real partner pipeline.
Amon's argument, in the CNBC interview, is that consumers will reach for an AI agent through a device they wear all day and that can see the world around them. Apps, he said, are going to change, and "agents are going to be the new app." He pegged smart-glasses shipments at "multiple tens of millions" a year already, and said they could reach "hundreds of millions" in a couple of years, putting the category within range of the 1.26 billion smartphones shipped in 2025, a figure from Counterpoint Research cited by CNBC. Amon also named Apple and Samsung as incumbents whose position could be reshaped, and pointed to OpenAI's 2025 acquisition of Jony Ive's hardware startup io as evidence that AI companies are moving into consumer hardware.
The bet has obvious friction points. The percentage gains are vendor benchmarks, and the 40-plus device count is a pipeline number, not installed base. On-device AI of the kind Qualcomm is pitching has historically struggled to clear the latency, heat, and battery bars that have kept headsets from going mainstream. Whether the new platform, the reference designs, and the white-label partners can clear those bars is the question that turns an announcement into a category.