Amazon is doing what Apple did with the iPhone: taking ownership of the silicon inside its most important consumer gadgets. On 2 July 2026, the head of Amazon's devices business, Panos Panay, told CNBC the company will push its own custom AI chips, branded AZ3 and AZ3 Pro, across its lineup of Echo smart speakers and Fire TV streaming sticks, with more devices on the way.
The first devices to ship with the new chips are Echo speakers and Fire TV streaming sticks. AZ3 powers the Echo Dot Max. AZ3 Pro powers the Echo Studio, Echo Show 8, and Echo Show 11. Both chips include a new AI Accelerator designed to run language models and vision transformers locally, on the device, rather than round-tripping every request to the cloud.
That distinction is the strategy. Amazon says the new silicon delivers more than 50 percent better wake-word detection and supports a sensor-fusion platform it calls Omnisense, which combines camera, audio, ultrasound, Wi-Fi radar, accelerometer, and Wi-Fi channel-state data (Amazon announcement). For Alexa Plus, the company's upgraded, generative-AI-flavored voice assistant, the pitch is that the chip lets the device understand context, anticipate intent, and respond fast enough to feel ambient rather than transactional.
Panay told CNBC Amazon is moving "away from a world of apps and screens" toward voice and ambient AI. The horizontal read, that Amazon has decided apps are losing to agents, is easy to overstate. The vertical read is more grounded: Amazon wants to control the silicon so it can control the local-versus-cloud boundary, the bill of materials, and the pace at which it can ship new device categories.
The Apple parallel is not perfect. In the CNBC interview, Panay acknowledged Amazon will continue using Qualcomm chips for other purposes, so this is vertical integration layered on top of an existing supply chain, not a clean Qualcomm displacement (Futurum Group analysis). The point is to own the parts of the stack that define how the device hears and what it can do without phoning home.
There is a quieter motivation underneath the strategy. As memory and neural-processing-unit-driven component costs have risen across the consumer-electronics industry, Amazon did not quantify how much it saves by making its own silicon, but it now has direct leverage on the biggest line items in the Echo bill of materials (CNET coverage). That mirrors the margin protection Apple has historically used on the iPhone, absorbing component-cost shocks with in-house silicon rather than passing them on.
The roadmap also points past the kitchen counter. Panay confirmed a "whole roadmap of on-the-go devices" beyond the home, contextualizing Amazon's 2025 acquisition of Bee, a wearable AI brand whose device clips to clothing and listens continuously (Crypto Briefing summary). The Humane AI Pin and the Rabbit R1 struggled commercially as ambient-AI concepts, in part because the silicon to run language models locally was not yet cheap or efficient enough. If Amazon is serious about on-the-go Alexa Plus, it is partly because it now believes the silicon is finally ready to make ambient AI a category rather than a concept demo.
The privacy question hangs over the strategy. Echo Show 11 still auto-processes voice for analysis with no user opt-out, and Amazon has not loosened that policy as the chips get more capable (CNET coverage). If the strategy is to push more AI onto the device, the device also gets more microphones and more ability to interpret the room without asking. That is a caveat, not a contradiction, but it is the part of the strategy Amazon has not yet answered.
What to watch next: whether the on-the-go roadmap produces a shipping product in the next year, whether Amazon opens up Omnisense to third-party developers, and whether the AZ3 line spreads to Ring cameras and Fire tablets. If ambient AI is going to win, the device under it has to be good enough to justify the trade. Amazon is now betting it can build that device itself.