Qualcomm's earbud silicon skips the phone and talks to the cloud itself

A buyer who wants a pair of buds to stop sending anything anywhere goes to the phone, because the phone is how the buds reach the internet. Qualcomm's new part changes that route. Snapdragon Sound Elite Gen 2 includes integrated micro-power Wi-Fi 6E, which Qualcomm calls a world first for a headphone chip, and which Tom's Guide reports lets the device connect to cloud audio directly ^[3].

The camera is a hardware decision an OEM makes once ^[10]. The radio is a second network path, and whoever writes the privacy page and staffs the support queue has to describe it. Qualcomm's pitch is that the direct-to-cloud link supports AI services locally and on the cloud, with an on-chip NPU and an ecosystem of apps and agents built for the part ^[4][6].

Two of the launch numbers can be put together. Qualcomm claims up to twice the AI capability of the previous generation ^[4], in a chip it says fits a 30% smaller footprint with 40% lower power consumption ^[1]. Divide 2 by 0.6 and you get roughly 3.3 times the AI work per unit of power ^[11]. The Tom's Guide account does not say what workload produced either figure, and the two may not describe the same block of silicon.

On whether anyone wants the resulting device, the strongest published case and the doubt come from the same piece. "My tech nerd take is of course that this is a solid silicon upgrade for upcoming devices," wrote Jason, a managing editor of computing at Tom's Guide ^[8]. He then asked whether this is a case of bad timing, pointing to the privacy problems of AI smart glasses and always-on recorders, to how much of people's lives get used to train models, and to the rise in covert recording ^[9]. The camera AirPods Pro that set expectations for the form factor are still a rumor by his account ^[7].

For a team deciding whether to put the camera in, two sentences have to be writable before the spec sheet matters. The first names the job the camera does in the first ten minutes that the phone in the same pocket cannot. The second says how the person across the table learns the camera is on. If the first sentence will not come, the camera is a demo. If the second will not come, the question Tom's Guide asked lands on a support agent.

The rest of the part does not depend on that choice. aptX lossless, ANC that adapts to the room, an AI microphone algorithm for voice pickup, and local models that let a siren through the noise wall ^[5] need no consent conversation.