In December, we predicted we’d see rapidly scaling demand for edge and item-level AI. And the start of 2026 has not disappointed. A key theme of this year’s CES has been intelligence moving closer to where data is generated – to the edge and, eventually, to individual items.
Announcements were focused around on-device AI capability, systems enabling real-time AI decision-making at the edge, and item-level intelligence in everyday objects – from small consumer gadgets to wearables and even connected toys, like smart Lego bricks.
Edge AI moves beyond inference
Historically, edge AI was largely about inference, running pre-trained models locally to reduce latency, bandwidth, and power consumption.
Now, on-device learning is beginning to expand, allowing more devices to adapt and respond in real time to their environment, especially in applications where size, cost, and power efficiency matter as much as performance.
This shift makes intelligence more dynamic, creating opportunities for predictive analytics, context-aware automation and even personalisation – ideal for brands wanting to get closer to their customers and deliver a curated experience.
Specialised chips power this transformation
For edge and item-level intelligence to scale, hardware must be efficient, versatile, and capable of fitting into new form factors.
Low-cost, flexible, and compact semiconductor solutions are increasingly enabling intelligence to be embedded into objects themselves, whether in packaging, wearables, or industrial devices.
Real-time, item-level data allows products to respond to conditions, personalise experiences, and support automated decision-making in ways that batch data or cloud-only processing cannot.
What’s next
This shift signals a structural change for the semiconductor industry. Efficiency, scalability, and integration are becoming central, and intelligence everywhere is moving from aspiration to expectation.
CES 2026 shows that the next phase of digital transformation will be defined not just by what devices can do in the cloud, but by what they can sense, learn, and act on locally, at both the edge and item level.
