Electronica 2026: chipmakers build AI platforms for robots and data centres

A shift from chips to platforms

Chipmakers at electronica 2026 in Munich spent less time talking about individual components and more time presenting complete platforms, according to EE Times. Editors reporting from the show describe a clear emphasis on system- and platform-level thinking, especially around physical AI and edge AI.

Physical AI is the term the industry now uses for intelligence that acts in the real world: drones, robots and humanoid machines rather than software that only answers questions. The major semiconductor companies at the fair all presented hardware and software bundles meant to help developers assemble such systems faster.

That bundle typically combines microcontrollers or processors, development tools, and software that uses agentic AI. Agentic AI refers to AI systems that can carry out multi-step tasks on a developer's behalf, such as generating code or configuring hardware, rather than simply answering a single prompt.

Neural processing spreads to every microcontroller

One of the clearest trends on the show floor concerned microcontrollers, the small chips that run everyday embedded devices. EE Times' Sally Ward-Foxton observed that almost every microcontroller now includes a dedicated NPU, a neural processing unit designed to run AI models locally.

This matters because it moves AI computing out of the cloud and into devices at the edge, where data is produced. A drone, a sensor or a household appliance can process information on the spot instead of sending it to a distant server.

The same report notes that companies such as Nordic Semiconductor want to be seen as solution providers rather than silicon suppliers. That means offering reference designs, software and support around the chip, a pattern visible across the industry.

Smaller start-ups are joining in as well. Alongside the established microcontroller makers, a growing number of young companies are building AI agents for code generation and other design tasks, changing how embedded software gets written.

Europe's strength in microcontroller design was also a talking point at the show, along with the many robotic hand designs on display, a sign of how quickly humanoid robotics hardware is diversifying.

AI data centres dominate the power halls

The power electronics side of the show revolved almost entirely around the AI data centre, according to Aaliya Shaukat, editor-in-chief of PowerElectronicsNews.com, speaking to EE Times. Training and running large AI models demands enormous amounts of electricity, and that demand is reshaping how power is delivered inside facilities.

Several technologies stood out. Exhibitors showed solid-state circuit breakers and transformers, which use semiconductors instead of mechanical parts to switch and protect electrical circuits. Vertical power delivery, which brings voltage conversion physically closer to the processor, also drew attention.

Consolidation is following the money. EE Times highlights Analog Devices' acquisition of Empower Semiconductor as one of the recent deals in this space.

Meanwhile, supply of wide-bandgap semiconductors, materials such as silicon carbide and gallium nitride that handle high voltages efficiently, is being squeezed by data centre demand. Wolfspeed is moving into GaN, shorthand for gallium nitride, and Infineon is scaling GaN production on 300-mm wafers, the larger wafer size that lowers cost per chip.

Executives set out their roadmaps

The show also featured keynote conversations with senior chip industry figures. EE Times lists interviews with NXP chief technology officer Lars Reger on physical AI, robots, drones and trust, and with Maria Heriz of STMicroelectronics on the company's plan to move all of its microcontrollers onto one STM32 platform.

Those interviews underline the platform message. A single unified microcontroller family with shared tools lowers the barrier for engineers, which is exactly what the platform push is about.

Another theme at the fair was the all-electric society, the idea that electrification of transport, industry and buildings will rely on advances in electronics. EE Times covered this as one of the broader narratives of the event.

What comes next for the industry

Looking ahead, Shaukat expects AI data centres to remain the top story in power electronics, with physical AI and humanoid robotics following as the next wave. That sequencing suggests the current investment cycle in data centre hardware will be followed by demand from machines that operate in factories, homes and public spaces.

For readers who design or buy embedded systems, the practical effect is a richer but more complex supplier landscape. Chipmakers are bundling software, AI tools and subsystems with their silicon, so choosing a vendor increasingly means choosing an ecosystem.

For the wider industry, the show confirmed two forces pulling semiconductor companies in the same direction: the appetite for power-hungry AI infrastructure, and the arrival of intelligence in everyday physical devices. EE Times' coverage suggests chipmakers intend to serve both, and to sell more than chips while doing it.