Qualcomm Launches an AR-Focused Chipset, and What the Future of XR Processing Could Look Like

Qualcomm announced the launch of its Snapdragon AR2 chipset, an Augmented Reality (AR) specialized chip design promising improved efficiency and Original Equipment Manufacturer (OEM) design flexibility thanks to distributed processing across multiple chipset nodes. The chip package has three components that can be placed in separate areas of a Head-Mounted Display (HMD). This is combined with a host device for heavy-duty processing, like graphics, to create a best-case scenario for spreading compute across available devices/locations.

Compared with Qualcomm’s own Extended Reality (XR) chipsets, having specialized, separated chips targeting important AR processing elements like Artificial Intelligence (AI), computer vision, and sensor fusion should prove more efficient for AR while physically being more compact—an important element for a challenging form factor like smart glasses. As the XR market introduces more AR devices to a wide consumer audience, design, ease of use, and capability are all equally important. Having more physical flexibility for hardware OEMs to design without sacrificing capability should be a win-win. Qualcomm already powers almost 100% of the XR market today, and so a preemptive AR chip before the expected consumer AR influx is well timed.

The AR2 chipset itself is not necessarily groundbreaking in terms of pure compute or efficiency; however, it showcases a few key design ideologies that we can expect to see more of going forward.

• Focus on AI: AR’s reliance on machine vision, combined with AI’s potential to increase both efficiency and raw compute capability when trained, means there will be greater emphasis on dedicated AI cores and improving those cores going forward. Virtual Reality (VR) could benefit as well, but the VR form factor is intrinsically more flexible than AR with greater room for components, heat management, and so on.
• Targeted compute capabilities: AR2 does not handle any graphics processing as it is demanding and as there is a host (tethered) device to handle that processing. Instead, AR2 handles latency-sensitive and AR mission-critical compute on devices, like image processing and AI engines. Connectivity also fits here as AR2 prioritizes Wi-Fi 7.
• Distributed compute: Going a step further from separating a single chipset into multiple chipsets for form factor flexibility, the next generation of “distributed processing”—as Qualcomm is calling it—is distributing compute location across both local devices and the network.

These ideologies combine for some welcome improvements—namely, a 50% reduction in power usage compared with XR2. For devices that don’t need all-onboard graphical compute or that prioritize design over pure compute, AR2 will be a welcome option. There is also an interesting angle with on-device operating systems—Android is the go-to, but an AR2 device connected to an Android smartphone would not need Android on the HMD and instead could use something more efficient like Linux.

The competitive landscape for XR processing has simultaneously been interesting and simple. Qualcomm owns the HMD market today; the company lists over 60 XR devices supported with the Snapdragon XR lineup (both XR1 and XR2).  

The broader AR market is already fragmented, and there is some danger of fragmenting what has so far been a cohesive Qualcomm effort with the XR chipset lineup. However, as AR demand increases across many verticals and use cases, efficient and flexible silicon will be a boon. ABI Research estimates over 40 million smart glasses will ship in 2027, increasingly within the consumer ecosystem. Qualcomm’s increased AR efforts are a welcome addition in a more and more complex AR market that favors silicon specialization to maximize compute efficiency and performance. Outside of hardware, partnerships will be a vital element to success, so Qualcomm’s official partnerships with Adobe, Microsoft, and Niantic will further open critical content creation, enterprise, and end-user segments.

Launching an AR-focused chipset can be seen as forward-looking today. The XR market has waffled between AR and VR for years. Today, however, attention is on VR, perhaps thanks to the constant news cycle around Meta and its VR efforts of late. Also, Qualcomm’s dominance in the total XR market means that there is no real pressure to change things up with a new product branch. However, the next battleground is AR—with players like Apple, Google, and Meta launching smart glasses over the next two years, with significant competition from other names as well. Qualcomm wants to maintain its market share (Meta is already a confirmed partner), and this involves not only securing new customers but also scaling existing ones—both in number of units and number of stock keeping units. 

However, Apple is an interesting foil in this landscape and perhaps the most notable one. The company’s XR headset is expected in 2023, likely competing with Meta’s Quest Pro and similar enterprise-minded VR-first devices. A future, pure AR device is expected afterward. Both of these devices, assuming there is some stand-alone processing onboard, will not run on Qualcomm technology. What they will run is, of course, not yet known, but Apple’s efforts with in-house silicon—and more importantly their success with it—makes it likely the company will keep XR processing in house as well.

The rest of the competitive marketplace is less clear and impactful. Other than Apple, there aren’t any notable devices on the horizon that will push a new silicon competitor. Instead, distributed compute is perhaps a greater competitive opportunity from some of the silicon incumbents. Intel and NVIDIA, for example, will not likely push an XR targeted chipset but instead look to stream content to an HMD remotely from an Intel or NVIDIA machine—data center or otherwise. As more compute is pushed off-device, the onboard chipset needs to change as well—even more focus will be put on latency-sensitive compute and connectivity. That could bring other chipset vendors into the fold—such as MediaTek, a company that has not prioritized high compute chipsets but now has an addressable low-power XR market to target.