The chip news worth watching at IFA is not the one promising a virtual person in every appliance. It is the one designed to make a product team stop adding boards.
Qualcomm has introduced two lower-cost Dragonwing processors, the commercial-grade Q-2390 and industrial-grade IQ-2390. Both combine a quad-core CPU, Adreno 704 graphics, a real-time RISC-V microcontroller and local AI hardware. The IQ-2390 adds the industrial details—time-sensitive Ethernet, ECC memory protection and an operating range Qualcomm specifies as -30°C to 115°C.
None of that makes a new phone faster. It could, however, change what belongs inside a smart display, retail terminal, building controller or robot. Local inference, vision, connectivity and deterministic control usually bring separate compromises. Qualcomm’s pitch is that a single, long-lived platform makes those trade-offs less punishing.
A useful measure of “AI at the edge”
The numbers are deliberately modest. Qualcomm lists 1.1 dense TOPS of AI performance for the IQ-2390—not a frontier-model machine, and not meant to be one. This is enough for tightly defined work such as camera analysis, an interface that adapts to sensor input, or a controller that must keep working if its cloud connection is slow or unavailable.
| Platform | Intended setting | The hardware distinction | What it makes plausible |
|---|---|---|---|
| Dragonwing Q-2390 | Consumer, commercial and enterprise IoT | LTE Cat 4, GPS, camera/display support and flexible I/O in one platform | Smart appliances, retail terminals, home robots and fitness equipment |
| Dragonwing IQ-2390 | Industrial edge systems | Two Gigabit Ethernet ports with TSN, a real-time MCU, ECC protection and ruggedized operating range | Machine vision, gateways, HMIs, PLC-adjacent systems and building controls |
The shared architecture includes one Arm Cortex-A78 core and three Cortex-A55 cores, with support for Android, Ubuntu, Yocto Linux and Zephyr. That menu matters. An appliance maker may want Android for a rich screen, an industrial supplier may want Zephyr for a predictable control task, and neither should have to re-engineer the board to choose.
The real purchase is time
Industrial devices do not live on the annual upgrade treadmill. Qualcomm says the IQ2 family is supported through 2036, subject to change. That is more revealing than a shiny demo: a factory controller or building-management panel needs an available processor, software maintenance and a believable replacement path for years.
The Q-2390’s appeal is more mundane but no less important. Integrating application compute, a camera path, AI acceleration, connectivity and security can reduce board area and the number of components a manufacturer has to qualify. If the final product is a point-of-sale terminal or a smart appliance, fewer integrations can be the difference between a useful local feature and a cloud-dependent one that gets cut before launch.
That does not automatically make local processing private, resilient or good. Those properties come from the product’s software, update policy and data handling. But moving a narrow perception or control task onto the device gives designers an option they often lack when every action is a round trip to a service.
Early access is not availability
There is a sizable boundary around this launch. Qualcomm says customers can see system-on-modules at IFA 2026 and join an early-access programme; evaluation kits are expected in early 2027. There is no finished consumer gadget to recommend, no retail price and no promise that every claimed application will materialise.
Independent coverage from CNX Software and TechTimes likewise frames the announcement as an embedded-platform launch rather than a new end-user device. That is the right reading. Its success will show up later in the boring places: a kiosk that responds locally, a controller with a longer useful life, or a camera system that does not need to stream every frame away.
For GearPulse readers, this is the part of the AI hardware cycle that eventually shapes the gadgets in front of us. It is similar to the dynamic behind NVIDIA’s effort to turn idle PCs into a local AI pool: capability is only interesting once the infrastructure becomes practical enough to disappear.
Bottom line
Qualcomm’s Q-2390 and IQ-2390 are not glamorous chips, and that is their strongest quality. They aim to package just enough local intelligence, real-time control and lifecycle planning for product makers who cannot afford a data-centre mindset. Watch for the first shipping systems, not the IFA booth. If those products use the integration to become simpler and more dependable, this quiet launch will matter far more than its 1.1 TOPS figure suggests.