Quectel Opens Mass-Production XR Mainboards to Customers: Hawk Glasses Boards and Dolphin Compute Units

On October 28, during the 2026 Shenzhen International AI+AR Smart Glasses Industry Exhibition, Quectel's Hawk AR+AI glasses solution won the Lingmou Golden Pupil Award for Innovation in Smart Glasses R&D and Manufacturing. The award is an industry signal. A disclosure made around the same event changes how the hardware gets delivered: Quectel develops its XR reference designs to production standards and now opens the PCBA mainboards to customers, who can buy the board outright instead of starting from a schematic. A day earlier, at the 2026 Global Smart Glasses Industry Summit on October 27, Ma Guowen, general manager of Quectel's XR product line, put the problem in plain terms: what consumes the most time, the most money and the most avoidable mistakes in a pair of smart glasses is the board in the temple, not the industrial design.

What Quectel Sells Now: A Board, Not a Blueprint

Quectel splits customers into two tracks. Brands with in-house ID and mechanical design capability can buy the production-grade PCBA directly and handle enclosure, structure and final assembly themselves. Brands without that capability can take a full ODM or white-label route, in which Quectel completes design, manufacturing and test, and the customer ships the device under its own brand. Software is modular: XR OS customization, voice and XR SLAM algorithms are available as services developed jointly with the customer.

Buying a board moves an entire hardware chain from the customer's side of the table to the supplier's. Quectel's stated scope covers hardware architecture and schematic design, PCB design and simulation, micron-level SMT and soldering, functional test and adhesive dispensing, display and sensor calibration, system-level test, and global certification. Each joint in that chain constrains the next. Placement and dispensing accuracy set the datum for the display module; that datum governs how quickly sensor calibration converges; the calibration result feeds straight into system test. No single step is exotic. Turning them into a line that converges repeatably is where the schedule actually goes, and it is the part no customer can buy in on a per-project basis.

Hawk and Dolphin: On-Board and Tethered

Quectel splits the boards into two families. Hawk is the glasses mainboard that sits in the temple, for all-in-one designs. Dolphin is an AR compute-unit board for tethered designs, where the compute sits in a unit outside the frame. Both families are already in mass production. The published lineup:

• Hawk, Snapdragon AR1 Gen1 waveguide board: 4nm, quad-core Cortex-A55 at 2GHz, paired with the WQ7036A BLE SoC (dual-core RISC-V), running Android 16 alongside RTOS.
• Hawk, ARS45 waveguide board: 12nm, quad-core Cortex-A53 at 1.8GHz, 6 TOPS, running Linux with RTOS and aimed at cost- and power-sensitive designs.
• Hawk, BirdBath board: SigmaStar SSC309QL, dual-core Cortex-A32; 1290x1200 per eye at 120Hz; supports Sony 0.68-inch Micro OLED at a 50-degree field of view; camera up to 12MP.
• Dolphin, Dragonwing Q-6690 compute unit: octa-core Kryo, 4nm, 2.9GHz; 6 TOPS NPU; 8GB+128GB or 12GB+256GB; display up to 4K at 144Hz; dual cameras up to 32MP per channel; 5G for China and North America, Wi-Fi 7, Bluetooth 5.4.
• Dolphin, Dragonwing QCS8550 compute unit: 48 INT8 TOPS, same memory configurations.

Read together, the tiering is deliberate. The AR1 Gen1 board carries the fullest ecosystem: Android 16 plus RTOS means application work can reuse mobile development resources instead of starting over. The ARS45 board trades 4nm for 12nm and a lower clock to land at 6 TOPS on Linux with RTOS, which suits products where cost and power budget decide the bill of materials. The BirdBath board inverts the ratio, pairing an SSC309QL with display specifications well above its compute tier — 1290x1200 per eye at 120Hz, Sony 0.68-inch Micro OLED, 50-degree field of view — which points at media consumption rather than spatial computing. Dolphin runs on a different axis: the Q-6690 with 6 TOPS, 4K144 output and 32MP-per-channel cameras for tethered designs that need a screen and capture, and the QCS8550 at 48 INT8 TOPS for heavier workloads.

Where Six to Eight Months Comes From

Ma's comparison is concrete: designing a comparable pair of XR glasses entirely in-house takes a year at best, and more often eighteen months to two years. On Quectel's model, a customer can reach mass production in six to eight months, halving the development cycle or better.

The figure needs unpacking. What gets compressed is the hardware platform's path from zero to production-ready — part selection, schematic, PCB and simulation, SMT, calibration, test, certification, the chain described above. What does not get compressed is just as specific. The roadmap covers compute and display. Optical design, whether a waveguide or a BirdBath module, along with mechanical parts, thermal design and full-device reliability validation, stays with the brand. The two form factors are not interchangeable either: a Hawk board does not fit a Dolphin enclosure, and the reverse does not hold. Six to eight months therefore rests on two preconditions — the board is already finalized, and the customer's ID and mechanical capability are in place. Quectel has not disclosed PCBA pricing, minimum order quantities, lead times or a customer list.

Put those preconditions back into the product-definition phase and a sequencing problem surfaces. Building glasses in-house used to mean hardware bring-up and definition validation ran in parallel, each waiting on the other. Once the board is a purchasable part, the hardware-side wait disappears and definition validation moves to the front: appearance, fit, use case and optical choice have to be settled earlier, or the months saved get consumed by rework later.

HSY Perspective

Opening the PCBA shifts XR glasses competition from whether you can build it to whether you can define it correctly, and the dividing line is optics and structure, not the board. Quectel has pulled schematic, PCB simulation, SMT, calibration and certification onto the supplier's side, so what a customer buys is a settled compute-and-display platform. The schedule Ma cited on October 27 measures the hardware platform's path from zero to one, not the device's.

The division of labor that follows matters more than the months saved. Once the board is a catalog part, brands stop differing on who got their board running first and start differing on product definition, use-case selection, and the trade-offs they make in optics and structure. Quectel's own line — that for most XR brands the real edge lies in product definition and scenario execution rather than rebuilding a hardware platform — describes that split accurately. But the three Hawk boards map onto two optical forms, waveguide and BirdBath, and Dolphin is a tethered compute unit; the roadmap covers compute and display throughout. Optical modules, mechanics, thermal design and device certification stay with the brand. A team with no depth in those areas buys half a program, not a shortcut.

The judgment holds while a customer's definition fits standard optics. Where the product depends on a non-standard optical path or an unusual wearing format, the two form factors stop being interchangeable, reuse of a standard board falls, and six to eight months stops being a predictable schedule. Two things are worth watching: whether the open board lineup keeps expanding, and whether XR OS customization, voice and XR SLAM move from optional services to default bundling. The second is what would mark the shift from selling boards to selling a platform.

Source: Quectel news release, "Quectel XR Wins Another Award as Mass-Production XR Mainboards Open to Customers," October 29, 2026, quectel.com.cn