AI Moves On-Device: Who Is Winning in PCs, IoT and Robotics, and What It Means for APAC Retail

As AI shifts from the cloud onto PCs, edge devices and robots across APAC, the brands that win sell-through will be those whose products are properly demonstrated on the retail floor by trained specialists.

The most consequential shift in consumer technology right now is not a new model release. It is where the computing happens. On-device AI across APAC is moving the workload off distant data centres and onto the hardware in front of the customer, from Copilot+ laptops in a Sydney JB Hi-Fi to robot vacuums on a Harvey Norman endcap. For the brands selling that hardware, the change is commercial, not just technical, because a feature that runs locally has to be shown to be believed.

This piece maps where on-device AI stands across three fronts, PCs, edge and IoT, and robotics, gives an honest read on who leads each, and lands on what it all means for anyone responsible for retail sell-through in the region.

From the cloud to the device: why AI is moving on-prem

For a decade, AI meant the cloud. That is changing fast. Latency, privacy, battery life and cost are pushing inference workloads onto the endpoint. IDC expects that by 2030, 50% of enterprise AI inference workloads will be processed locally on endpoints or edge nodes rather than in the cloud. That is a structural redistribution of where intelligence lives.

The money is following. Worldwide spending on edge computing reached $265 billion in 2025 and is expected to nearly double to roughly $450 billion by 2029, driven by the rise of edge AI. Meanwhile over two-thirds of all enterprises globally will deploy edge AI by 2029, up from just 10% in 2025, according to Gartner.

What matters for retail brands is the consumer-facing edge of this trend. On-device AI is arriving in three product categories your customers actually buy at a counter: PCs, connected home and IoT devices, and robots. Each one is a considered purchase, and each one carries a story that a spec sheet cannot tell.

AI PCs: the NPU race and who is shipping the silicon

The AI PC is the clearest example of on-device AI reaching mass retail. Gartner forecasts worldwide AI PC shipments will total 77.8 million units in 2025, representing 31% of the total global PC market by the end of the year, rising from 15.6% in 2024 to 54.7% in 2026. The category is becoming the default, not the premium exception.

The bar is now defined. Microsoft classes a Copilot+ PC as a Windows 11 device with an NPU capable of more than 40 TOPS, at least 16GB of RAM, and qualifying silicon from AMD Ryzen AI, Intel Core Ultra or Qualcomm Snapdragon X. That spec is where the silicon fight plays out, and the leadership picture is genuinely contested depending on how you slice it.

The read is this: Intel still ships the most AI PC silicon, Apple leads the broad AI-capable category on its own terms, and Qualcomm has proven it can take real premium retail share where its battery and NPU story is demonstrated well. The APAC context sharpens the stakes. IDC reports the Asia/Pacific PC market grew 11.6% in 2025 to 106.6 million units but projects a 13.7% decline to 92.0 million in 2026. In a shrinking unit market, the AI PC premium is where margin lives, and Canalys estimates AI-capable PCs carry a 10% to 15% price premium over similar non-NPU machines.

Edge AI and IoT: intelligence at the point of use

Below the PC sits a far larger and more fragmented category: everything from robot vacuums to AR glasses to connected appliances. This is where on-device AI becomes invisible to the spec sheet and visible only in behaviour, how well a Dreame or Ecovacs robot vacuum maps a room, how an Anker Eufy camera recognises a face without sending video to the cloud.

The market is substantial. The global edge AI hardware market is projected to grow from USD 26.14 billion in 2025 to USD 58.90 billion by 2030, at a CAGR of 17.6%. On the connectivity side, total global IoT connections are forecast to reach 40.8 billion by 2030, and APAC sits at the centre of this.

North East Asia accounts for around 70% of all global cellular IoT connections by the end of 2025, according to the Ericsson Mobility Report.

Two things stand out for retail brands. First, the retail and services sector is set to account for nearly 28% of total global edge computing spending in 2025, meaning the retail environment itself is becoming an edge AI venue. Second, on-device AI is going mainstream in the smartphone, with GenAI-capable smartphones forecast to reach 45% of global shipments in 2026, up from 36% in 2025. When the phone in a customer's pocket already does on-device AI, the bar for explaining the premium device on the shelf rises with it.

Robotics and physical AI: the next platform fight

The newest front is physical AI, robots that perceive and act in the real world. This is where the platform war is being fought hardest right now, and the two names to watch are NVIDIA and Qualcomm.

NVIDIA has the incumbent advantage. Its Jetson platform and robotics stack have attracted over 2 million developers and 150+ partners since 2014, with Jetson Orin alone enabling more than 7,000 customers to deploy edge AI, and early Jetson Thor adopters including Boston Dynamics, Figure and Amazon Robotics. Qualcomm is pressing in: at CES in January 2026 it introduced its Dragonwing IQ10 Series robotics processor, a full stack built with partners including Figure, Kuka and Advantech.

The growth projections are large and APAC-weighted. Goldman Sachs raised its humanoid robot shipment forecast to roughly 6.5 million units by 2035, valuing that market at about $138 billion, with shipments rising from around 75,000 in 2026 to 890,000 by 2030. On the industrial side, China installed about 354,000 industrial robots in 2025, a 20% rise accounting for 59% of all global installations. Consumer robotics, from vacuums to lawn mowers to companion devices, will ride the same silicon and software curve into retail.

Who is in the lead, and what the near-term scoreboard looks like

An honest scoreboard, by segment:

No single company owns on-device AI end to end. That fragmentation is exactly why the retail floor matters: customers cannot be expected to understand which silicon does what, so the brand's representation at point of sale carries the message.

What this means for APAC retail: products that have to be demonstrated

Here is the commercial point. An AI PC, an edge-AI robot vacuum and a GenAI smartphone share one trait: their most valuable capabilities are invisible on a spec sheet. A customer in The Good Guys cannot see 45 TOPS. They can see a laptop transcribe a meeting in real time, or a robot vacuum identify and avoid a cable on the floor. The value only becomes real when it is shown.

That is a problem for brands, because generic retail staff cannot tell a Snapdragon X story, explain why an NPU matters for battery life, or run a live on-device AI demo under the pressure of a Saturday shop floor. These are considered, demonstrated purchases carrying a double-digit price premium, and the premium evaporates if nobody on the floor can justify it. When Snapdragon X launched into premium Windows retail, the share it took above $800 was won in exactly those demonstrated moments, not on the box.

This is the gap VMS Group closes across APAC. Trained product specialists, launch teams and structured product enablement in JB Hi-Fi, Harvey Norman, The Good Guys and equivalent chains across the region turn an invisible on-device capability into a sale. As more of the portfolio becomes AI-first, the quality of your field force becomes the deciding variable in sell-through.

If you are planning a launch or want your retail floor ready for the next wave of on-device AI products, start a conversation with our team. It is a practical next step toward making sure your technology actually sells where customers decide.

Sources