Ambiq Shows How Ultra-Low-Power Edge AI Can Run in Wearables

Smarter wearables need lower power

Wearable devices are getting smarter. Smart rings, fitness bands, smartwatches, hearables and health monitors now collect more sensor data and run more local processing.

That creates a major design challenge. Engineers need to add artificial intelligence (AI) features without draining the battery.

At Microelectronics US, ipXchange spoke with Dr Adam Page, Head of Artificial Intelligence at Ambiq. He explained how Ambiq supports AI in battery-powered devices through low-power silicon, software development kits and model deployment tools.

What makes Ambiq different?

Ambiq’s core technology is Subthreshold Power Optimised Technology (SPOT). This allows its silicon to run at much lower voltages. As a result, devices can reduce power use in always-on applications.

This matters for products such as smart rings, smart glasses, health monitors and other small wearable devices. These products need local processing, but they also need long battery life.

In the past, Ambiq devices often acted as low-power sensor hubs. They worked close to microphones, inertial measurement units (IMUs), photoplethysmography (PPG) sensors or electrocardiogram (ECG) front ends.

Now, Ambiq is moving into a larger role. Its devices can act as the main processor in more advanced wearable designs.

Running AI models on-device

The video demos use Ambiq’s Apollo510 platform. This platform uses an Arm Cortex-M55 central processing unit (CPU) and supports edge AI workloads on-device.

Ambiq shows several AI tasks running locally. These include health signal processing, speech enhancement and AI-based sensor data compression.

One demo focuses on remote patient monitoring. The system processes ECG and PPG signals directly on the device. It uses AI models for denoising, segmentation and rhythm classification.

This is important because wearable health signals are often noisy. Motion, skin contact, sensor position and electrical noise can all affect signal quality. Local AI processing can clean and interpret that data before the device sends anything to the cloud.

Improving audio and reducing data

Another demo shows speech enhancement. This is useful for hearables, smart glasses and voice-controlled products. The system reduces background noise while keeping speech clear.

This shows that edge AI is not only about vision. Audio, health data and sensor processing all matter in wearable AI.

Ambiq also demonstrates AI-based compression for wearable sensor data. Even when devices run AI locally, companies may still want ECG, PPG or IMU data for cloud analysis or product improvement.

Sending that data can use a lot of power. Ambiq’s compression approach reduces the amount of data that the device needs to store or transmit. In the video, Ambiq discusses compression levels of up to 16x or 32x while keeping the key signal information.

Why the software matters

A major point from the video is that low-power silicon alone is not enough. Engineers also need software that deploys AI models efficiently.

Ambiq discusses its runtimes, development kits and profiling tools. These help engineers optimise AI workloads across the processor, memory and acceleration features.

For design engineers, the takeaway is simple. Wearable edge AI depends on the full system, not just raw compute. Ambiq’s demos show how low-power silicon, efficient software and model optimisation can help bring more capable AI features into small battery-powered devices.