Spiking Neural Processor T1
The Spiking Neural Processor T1, a tiny artificial intelligence (AI) chip inspired by the human brain, has arrived and will increase the battery life of smart devices.
O first “neuromorphic chip” of the world was presented at CES 2025 – the largest technology fair in the world, which took place last week in Las Vegas.
As described by , this chip Mimics human brain architecture and introduces artificial intelligence capabilities on smart devices with limited power consumption, such as Wi-Fi connected lamps, doorbells or smoke alarms.
What is a “neuromorphic chip”?
The Spiking Neural Processor T1 is a neuromorphic processor – meaning its architecture is organized to mimic the brain’s pattern recognition mechanisms.
For example, when we feel something – be it a smell or a sound – we have different types of neurons that fire to identify sensations.
Likewise, on the chip, different groups of artificial neurons record spikes.
As Live Science explains, the underlying principle is spike neural network (SNN) – where a neural network is a set of machine learning algorithms and the spikes it produces are similar to the signals produced by brain cells.
SNN algorithms also tend to be about 100 times smaller in file size than conventional deep neural networks used in large language models.
Batteries can last six times longer
In an interview with Live Science, on the sidelines of CES 2025, Sumeet KumarCEO of processor company Innatera Nanosystemsresponsible for this innovation, explained that there are three fundamental layers in the T1 chip:
- The first is the computing engine based on SNN, which has a power dissipation of less than 1 milliwatt and a latency, or delay, that is typically less than 1 millisecond for most applications.
- The second layer includes conventional deep neural networks.
- The third layer includes a standard processor that deals with the functioning of the system.
The T1, or similar chips, would increase battery life up to six times in some smart devices and scenarios.
For example, a prototype smart doorbell built with the T1 processor that could detect a person’s presence using radar technology lasted 18 to 20 hours, versus one or two hours for a conventional Wi-Fi-based product that sends image data. and video for servers.
The chip is already being prepared for mass production this year, with samples being sent to device manufacturers. The first products with the T1 neuromorphic chip are expected to hit shelves in 2026.
