Neuromorphic computing: Overview
What it does
The job the technology does for people, in a short paragraph
A neuromorphic chip computes the way a brain does, with many simple neurons that signal by brief electrical spikes. It aims to do tasks such as recognising patterns or controlling robots with far less energy than ordinary chips. Carver Mead set out the idea in 1989 [1]. It is still mainly a research tool; the largest system, Hala Point at Sandia, has 1.15 billion neurons [3].
as of 2026-11-11
How it works
The working principle in plain words, step by step
An ordinary computer keeps memory and processing apart and runs on a clock, shuttling data back and forth on every tick. A neuromorphic chip mixes the two. It is split into many small cores, each holding a group of artificial neurons and the synapses, the weighted links, that feed them. A neuron adds up incoming signals until it crosses a threshold, then fires a spike, a short pulse sent to other neurons, much as Hodgkin and Huxley described for nerve cells [7]. Between spikes, little happens and little power is used. IBM's TrueNorth puts 1 million neurons and 256 million synapses in 4,096 cores [4]. Chips are linked into larger systems, and some, such as Loihi, can change their synapse weights as they run, so the network learns.
as of 2026-11-11
Key figures
The handful of numbers that size it up, each with its unit, year and source
| Figure | Value | Unit | Year | Source |
|---|---|---|---|---|
| Artificial neurons in Hala Point, the largest system, at Sandia | 1.15 | billion neurons | 2024 | [3][9] |
| Loihi 2 processors in Hala Point | 1152 | chips | 2024 | [3][9] |
| Programmable spiking neurons on IBM's TrueNorth chip | 1 | million neurons | 2014 | [4][10] |
| Power used by the TrueNorth chip | 63 | mW | 2014 | [4][10] |
| Processor cores in Manchester's SpiNNaker machine | 1 | million cores | 2018 | [5] |
as of 2026-11-11. Sandia received Hala Point in early 2024 [3][9]; TrueNorth was described in Science in August 2014 [4][10]; the SpiNNaker figure is from the University of Manchester's switch-on announcement of November 2018 [5].
Where it is used
The main places and products the technology is found in
- Real-time simulation of brain regions for neuroscience
- Robots that interpret visual information as it arrives
- Low-power pattern recognition and deep learning benchmarks
- Scientific number crunching for national laboratories
- Research into energy-saving AI hardware
as of 2026-11-11
Scale today
How much of it there is now: units in use, made each year and who uses it
- Installed
- Largest system: Hala Point at Sandia, 1,152 Loihi 2 chips with 1.15 billion neurons, received in early 2024 [3][9]
- Users
- More than 200 members of the Intel Neuromorphic Research Community [9]
- Unit
- neurons; research partners
- Year
- 2024
- Note
- No sales or shipment figures were found; neuromorphic hardware is still used mainly in research.
as of 2026-11-11
Sources
- Carver Mead Lemelson-MIT Program
- Advancing neuromorphic computing at the Neural Exploration and Research Lab Sandia National Laboratories, 2026
- A Million Spiking-Neuron Integrated Circuit with a Scalable Communication Network and Interface Merolla et al., Science 345 (abstract at Yale University)
- 'Human brain' supercomputer with 1 million processors switched on for first time University of Manchester, 2018
- Hodgkin and Huxley's modelling of the action potential The Physiological Society
- Intel Builds World's Largest Neuromorphic System to Enable More Sustainable AI Intel (newsroom), 2024
- IBM Chip Processes Data Similar to the Way Your Brain Does MIT Technology Review, 2014