Can brain-inspired computers match the human brain’s power consumption?

The race to build machines that think like us has hit an energy hurdle: the human brain runs on roughly 20 watts, about the draw of a standard LED lamp. Brain-inspired computers promise brain‑like efficiency, but can they truly stay within that tiny power envelope?

brain-inspired computers power

Neuromorphic chips, such as Intel’s Loihi and IBM’s TrueNorth, are engineered to fire only when needed, mimicking neuronal spikes. Early prototypes consume ↑ 15% less energy than conventional GPUs when running sparse workloads, according to Reuters.

“The energy envelope of neuromorphic chips is still a moving target,” says Dr. Elena Ruiz, senior researcher at IBM.

Scaling the technology remains a challenge. While a single chip may sip a few watts, integrating thousands could push total draw toward the brain’s 20‑watt sweet spot. Analysts at Bloomberg warn that without breakthroughs in low‑leakage transistors, the promise may stay theoretical.

Words by Dr. Aris Thorne (Artificial Intelligence Researcher).