Content options
Neuromorphic computing is a way of designing chips that imitate how the brain works. Instead of keeping memory and processing apart, these chips place them together and spend energy only when a signal actually arrives. Research on tiny magnetic vortices, reported by Science Daily, points to one possible way of building such chips from magnetic materials, but this work is at the laboratory stage and you cannot buy a chip based on it.
Key takeaways
- A neuromorphic chip mimics neurons and synapses, so it can handle some tasks, such as recognising patterns, with far less power than a conventional processor.
- Magnetic vortices are tiny swirling patterns in magnetic material. Researchers are exploring whether they can behave like artificial neurons.
- Today's CPUs and GPUs are still better for most jobs, including spreadsheets, games and training large AI models.
- Chips from companies such as Intel and IBM exist for research, but vortex-based devices are experimental.
What is a neuromorphic chip?
The word comes from Greek: neuro for nerve and morph for shape. A neuromorphic chip is shaped, in a loose sense, like a nervous system. It contains many simple units that act like neurons, joined by connections that act like synapses.
Your brain has roughly 86 billion neurons and runs on about 20 watts, less than a typical bedroom light bulb. It does this around the clock, including while you sleep (our guide to sleep hygiene covers what that rest does for you). A data-centre server doing comparable pattern-recognition work can draw many times more power. That gap is the reason engineers keep returning to the brain for ideas.
We should be clear about one thing. Neuromorphic chips do not think, feel or understand. The phrase "chips that think like brains" is a handy shorthand for a hardware design copied from how neurons pass signals.
How do brain-inspired computers work?
Three ideas separate a neuromorphic design from the processor in your phone or laptop.
1. Memory and processing sit together
In a standard computer, data moves back and forth between a memory unit and a processor. This constant shuttling wastes time and energy, and engineers call it the von Neumann bottleneck. In the brain, a synapse stores a connection strength and takes part in the calculation at the same place. Neuromorphic designs try to do the same.
2. Signals are short spikes, not constant streams
A neuron stays quiet until its incoming signals add up past a threshold. Then it fires a brief spike. Chips built on this idea are often called spiking neural networks. If nothing changes in the input, almost nothing happens inside the chip.
3. Work happens only when there is an event
A conventional chip runs on a clock that ticks billions of times a second whether or not there is anything useful to compute. A neuromorphic chip is event-driven. A camera that sends data only when a pixel changes is a natural partner for it, because a still scene produces almost no work.
How do neuromorphic chips compare with today's processors?
| Feature | CPU | GPU | Neuromorphic chip | Human brain |
|---|---|---|---|---|
| Main strength | General tasks, one step after another | Many parallel calculations, AI training | Low-power pattern recognition and sensing | Learning, adapting, perceiving |
| Memory and processing | Separate | Mostly separate | Closely combined | Combined |
| How it runs | Clock-driven | Clock-driven | Event-driven spikes | Event-driven spikes |
| Maturity | Everywhere | Everywhere | Research and early pilots | Not applicable |
For most people, the sensible position is this: neuromorphic chips will not replace CPUs. They are more likely to sit beside them as specialised helpers for always-on sensing jobs where battery life matters.
Where do magnetic vortices come in?
Most neuromorphic chips today are built from ordinary silicon transistors, which have to be arranged cleverly to imitate a neuron. Some researchers think other physical systems might imitate neurons more naturally. Magnetic materials are one candidate, and the Science Daily report describes tiny magnetic vortices as a way to make computers work more like the brain. We have only the headline and summary to go on, so the details of that particular study, such as the materials, sizes and measured results, are best checked in the original report.
What is a magnetic vortex?
Picture water circling a drain. In a very small disc of magnetic material, the tiny magnetic arrows inside can curl around the centre in the same way. At the middle there is a small core where the magnetisation points up or down. That whole pattern is a magnetic vortex.
Such a vortex can be nudged by an electric current or a magnetic field. It can then move, wobble or settle into a new state, and it can influence nearby vortices.
Why might that resemble a neuron?
A neuron integrates incoming signals, fires when enough has built up, and affects its neighbours. A vortex that responds to inputs, shifts state and couples to nearby vortices has some of those features. Magnetic devices also hold their state without power, which is useful for storing synapse-like values.
Here is the general logic researchers follow with devices like this:
- Make a tiny magnetic disc in which a vortex forms.
- Send in an electrical or magnetic input that acts like an incoming signal.
- Watch how the vortex moves or changes in response.
- Place several such discs close together so they influence one another, like a small network.
- Check whether the network can recognise simple patterns or learn from repeated inputs.
Steps 4 and 5 are the hard ones, and they are where most of the open questions lie.
What could lower power use mean in rupees?
The numbers below are an illustration, not a measurement of any real vortex chip. They show why efficiency matters when machines run all day.
Assume electricity costs about ₹8 per unit (a kWh). That is a rough round figure, and your actual tariff will differ by state and category.
- A 300 W accelerator running 24 hours uses 0.3 kW × 24 h = 7.2 kWh a day. At ₹8 per unit, that is 7.2 × 8 = ₹57.60 a day, or about ₹1,728 over 30 days.
- A 20 W device running 24 hours uses 0.02 kW × 24 h = 0.48 kWh a day. That costs 0.48 × 8 = ₹3.84 a day, or about ₹115 over 30 days.
The difference is about ₹1,600 per device per month. Multiply that by thousands of devices in a data centre, or by millions of sensors that run on batteries, and efficiency becomes a major design goal. The caveat is that a 20 W chip and a 300 W accelerator are not doing the same job. A fair comparison needs the same task, which is exactly what researchers are still testing.
What exists today, and what is still lab-stage?
Working neuromorphic hardware does exist. Examples include IBM's TrueNorth, described at its 2014 launch as having about one million neurons and 256 million synapses, Intel's Loihi research chips, and the SpiNNaker machine built at the University of Manchester. A few companies sell small neuromorphic processors aimed at low-power sensing. These are mostly used by researchers and in pilot projects, not in everyday phones or laptops.
Magnetic vortex computing sits well behind them. Here is a fair way to sort what is known from what is not.
- Shown in labs: magnetic vortices can be created, controlled and made to interact in tiny structures.
- Being explored: whether groups of vortices can do useful, brain-like computation and learn.
- Not yet shown: manufacturing them at chip scale, in billions, at low cost, and wiring them to ordinary electronics reliably.
Any claim of a commercial product based on vortices in the near term should be treated with caution.
Myths and facts
Myth: neuromorphic chips will make computers conscious
Fact: they copy a style of signalling, not a mind. Nothing in the design gives a chip awareness or understanding.
Myth: these chips will replace the processor in your phone
Fact: they suit narrow, always-on tasks such as detecting a wake word or spotting motion. General computing still needs conventional processors.
Myth: a research headline means a product is coming soon
Fact: many lab results never reach the market. Moving from a working device to a manufacturable chip often takes many years.
Why should an ordinary reader care?
If this research succeeds, the gains would show up in small, quiet ways. Hearing aids and wearables could last longer between charges. A battery-powered sensor in a field or a warehouse could run for months while still spotting unusual patterns. Cameras could filter what they see on the device itself instead of sending everything to a server.
It also connects to the larger AI story. Much of today's AI runs on power-hungry servers in the cloud, which we explain in our guide to cloud computing. Software that acts on its own, covered in our piece on AI agents, will need lots of computing too. More efficient hardware could help keep the cost of all this in check.
Frequently asked questions
What is a neuromorphic chip in simple words?
It is a chip designed to work like a small network of brain cells. Its parts send short signals to each other and store information where they compute. This makes it efficient for tasks like recognising patterns in sound or images.
Are brain-like computer chips available to buy?
Some small neuromorphic processors are sold to developers and researchers, and larger research systems exist at universities and companies. You will not find one in a mainstream phone or laptop today.
What is magnetic vortex computing?
It is an approach that uses tiny swirling magnetic patterns in nanoscale discs to store and process information. Researchers are studying whether these vortices can imitate neurons and synapses. It remains experimental.
Is neuromorphic computing the same as AI?
No. AI is a set of software methods, while neuromorphic computing is a hardware design. Neuromorphic chips can run certain AI workloads efficiently, but most AI today runs on GPUs and CPUs.
Will neuromorphic chips make GPUs obsolete?
Unlikely in the foreseeable future. GPUs are excellent at the heavy maths used to train large AI models. Neuromorphic chips are expected to complement them in low-power, always-on situations rather than replace them.
Bottom line
Neuromorphic computing borrows the brain's habit of combining memory with processing and working only when something happens. Magnetic vortices are an interesting candidate material for building such chips, but the work is early and unproven at scale. If you follow this field, watch for results on manufacturing and on real tasks compared with conventional chips, and read the original research report, not just headlines, before drawing conclusions.
