On paper, onsemi’s agreement to acquire Synaptics is a portfolio expansion but strategically, it is something more interesting: an attempt to build the architecture for machines that can sense, interpret, connect, and act without waiting for the cloud to do all the thinking.
The first iteration of IoT was mostly about visibility. Put sensors on assets, connect devices, collect data, then send it somewhere else for analysis. That model helped companies see more of their operations, but it left a gap between information and action. In a factory, vehicle, robot, grid asset, medical device, or warehouse system, the most valuable intelligence is often the ability to act locally. It has to read the environment, process signals, make a decision, stay connected, and do all of it within the limits of power, heat, latency, cost, and reliability.
This is the context for onsemi’s agreement to acquire Synaptics in an all-stock transaction valued at approximately $7 billion (onsemi, 2026). The deal brings together onsemi’s strength in power management and sensing with Synaptics’ Edge AI compute, wireless connectivity, interface technologies, and software assets.
Physical AI operates under different pressures from data center AI. In the cloud, intelligence can lean on scale, with compute, data, and infrastructure concentrated in one place. At the edge, it has to work inside physical limits. Robotic systems cannot send every decision away and wait for a response. Vehicles need intelligence close enough to act when milliseconds matter. Industrial sensors have to become smarter without draining power or adding fragility. Edge intelligence therefore has to be fast, efficient, local, and tightly integrated with the machine it is meant to guide.
Already, onsemi has deep positions in power management and image sensing, two of the essential layers for machines that need to see and operate efficiently. Synaptics adds Edge AI compute via its Astra processors and neural processing units, alongside Wi-Fi, Bluetooth, GPS, human-machine interface technologies, and software assets designed to support edge development.
Altogether, the deal is less about buying another chip company and more about closing the loop. Physical AI needs power, sensing, connected compute, and control to function as a single system. A sensor without local interpretation is limited. Computing without efficient power management is difficult to deploy widely. Connectivity without control creates data traffic rather than autonomy.
Customers want systems that can detect conditions, interpret them locally, and respond quickly enough to make a difference. That could mean smarter factory automation, more responsive robotics, safer vehicles, better energy infrastructure, or equipment that can recognize failure patterns before they become downtime.
In that world, semiconductor suppliers cannot behave like catalog vendors. They need to solve integration problems. Industrial and automotive customers do not want to assemble five disconnected technologies and carry all the design risk themselves. They want platforms that integrate the right sensing, compute, connectivity, software, and power architecture into a form that can withstand real-world operating conditions.
The companies said the deal would expand onsemi’s total addressable market to $243 billion by 2030, with opportunities across automotive, industrial automation, robotics, AI infrastructure, and connected intelligent systems. That figure gives the transaction scale, but the harder test is coherence. A larger portfolio will not matter if customers still experience it as separate products. The value will come from turning the combined assets into architectures that engineers can actually design into real machines.
The interesting tension is that the market may still read this through yesterday’s AI lens. Investors have been trained to look for data centre leverage, accelerator demand, and cloud infrastructure exposure. Physical AI is less clean as a story. It is distributed. It is application-specific. It depends on long customer cycles, hardware integration, software support, safety requirements, and industrial credibility. That makes it harder to model, but potentially more durable if onsemi can execute.
But there are some risks. Hardware portfolios are difficult to combine while software ecosystems take time to mature. Industrial customers move carefully because their products often have long life cycles and high reliability demands. Physical AI will not reward an Genelegant strategy alone. It will reward systems that work in demanding environments, over time, with fewer integration headaches.
If the deal closes in mid 2027, subject to shareholder and regulatory approvals, onsemi will be better placed for a version of IoT that is no longer satisfied with simply reporting what happened. The next generation of connected machines will be expected to recognize what is happening, decide what matters, and respond before the moment has passed.
That is the shift from IoT to Physical AI and the market onsemi is trying to reach before it becomes crowded.