Northwestern University engineers printed artificial neurons that activated living mouse brain cells. The work was published April 15 in Nature Nanotechnology.

The devices use nanoscale flakes of molybdenum disulfide and graphene deposited onto flexible polymer films. Molybdenum disulfide acted as the semiconductor. Graphene served as the conductor. Aerosol jet printing placed the inks only where needed.

What did the Northwestern team actually build?

The artificial neurons produced spikes matching the timing and duration of real neural signals. They fired single spikes, continuous trains, and bursting patterns.

Mark C. Hersam led the study from the McCormick School of Engineering, where he holds the Walter P. Murphy Professorship and chairs materials science. Hersam also directs the university's Materials Research Science and Engineering Center and holds joint appointments at the Feinberg School of Medicine and the Weinberg College of Arts and Sciences. Vinod K. Sangwan, a research associate professor at McCormick, co-led the work.

The printed hardware relies on a memristor network. Polymer residue inside the ink is partially decomposed to form narrow conductive filaments. Current constriction through those filaments produces the firing events.

How does the printed device produce neuron-like spikes?

Earlier researchers stripped the polymer binder from printed inks because it hurt electrical performance. Hersam's group instead decomposed the binder partially. Current then drove further decomposition in a spatially inhomogeneous pattern.

That process leaves one conductive path through each device. The signals span a temporal range previous artificial neurons could not reach. Organic-material versions spiked too slowly. Metal-oxide versions spiked too fast.

"Other labs have tried to make artificial neurons with organic materials, and they spiked too slowly," Hersam said. "Or they used metal oxides, which are too fast. We are within a temporal range that was not previously demonstrated for artificial neurons. You can see the living neurons respond to our artificial neuron."

Testing took place with Indira M. Raman, the Bill and Gayle Cook Professor of Neurobiology at Weinberg. Her team applied the artificial signals to slices of mouse cerebellum. The applied spikes matched key biological properties, including timing and duration. Living neurons responded reliably to the printed device's output.

Why does the brain outperform silicon by such a wide margin?

Modern computing stacks billions of identical transistors onto rigid two-dimensional chips. Every device behaves the same way. The system stays fixed after fabrication.

Biological neurons differ. Many specialized types exist. They form soft three-dimensional networks that rewire constantly. Hersam argues the contrast demands new materials and new fabrication methods.

What does this mean for AI power budgets?

Hersam frames the work as a response to the power wall behind AI training. He estimates the brain operates five orders of magnitude more efficiently than a digital computer.

Data centers now consume gigawatts of power. They demand dedicated nuclear capacity. Cooling strains regional water supplies.

"To meet the energy demands of AI, tech companies are building gigawatt data centers powered by dedicated nuclear power plants," Hersam said. "It is evident that this massive power consumption will limit further scaling of computing since it's hard to imagine a next-generation data center requiring 100 nuclear power plants."

The printed process reduces material waste. Additive manufacturing deposits ink only where required. Costs stay low. Environmental burden falls.

The paper, "Multi-order complexity spiking neurons enabled by printed MoS2 memristive nanosheet networks," was published in Nature Nanotechnology, 2026, DOI 10.1038/s41565-026-02149-6. It received support from the National Science Foundation. It points toward brain-machine interfaces and neuroprosthetics for hearing, vision, and movement. Commercial timelines remain undefined, but the temporal fidelity shown in mouse tissue sets a benchmark that earlier printed neurons could not match.