Northwestern Prints Artificial Neurons That Successfully Talk to Living Brain Cells

Northwestern University engineers published a landmark neuromorphic computing paper this week. Flexible, printed artificial neurons generate electrical signals realistic enough to successfully trigger responses in living mouse brain tissue. The paper appeared April 15 in Nature Nanotechnology. Coverage peaked April 17-18.

The breakthrough

What the device is:

• Flexible polymer substrate
• Aerosol-jet-printed with two nanomaterial inks:
  • Molybdenum disulfide (MoS2) nanoscale flakes as the semiconductor
  • Graphene as the electrical conductor
• No silicon, no CMOS foundry, no custom lithography. Prints at atmospheric pressure.

What it does:

Produces three distinct biological-like firing patterns:

1. Single spikes (individual action potentials)
2. Continuous firing (sustained rate coding)
3. Bursting patterns (grouped spikes separated by quiescence)

All three match how real cortical neurons signal.

The critical test:

The research team placed the printed neurons on slices of mouse brain tissue. The artificial neurons triggered real responses in living neurons. That bidirectional coupling is the milestone. Previous printed or electrode-based devices could measure biological activity or stimulate it crudely. These devices speak in the same dialect as real neurons.

Why this matters for AI

Neuromorphic computing is the alternative to GPU scaling.

Every large AI model runs on silicon GPUs burning kilowatts of power per query. Human brains do comparable reasoning on ~20 watts. The gap is not algorithmic alone. It’s architectural: biological neurons compute with sparse spikes, not dense floating-point matmuls.

Neuromorphic hardware has been a research target for decades (Intel Loihi, IBM TrueNorth, SpiNNaker). All prior approaches built neuron-like circuits in silicon. Northwestern’s printed approach is different:

• Biological compatibility. Can interface directly with living tissue.
• Cost. Aerosol-jet printing is orders of magnitude cheaper than silicon fab.
• Flexibility. Can be implanted, wrapped on curved surfaces, or laminated on skin.

Applications pipeline

Near-term (5-10 years):

• Neuroprosthetics: implants restoring hearing, vision, or movement after neural damage. Print neurons sit at the interface between a biological stump and a prosthetic actuator.
• Brain-machine interfaces: higher-bandwidth and lower-rejection-rate than Utah-array-style electrode implants because the material is tissue-compatible.
• Closed-loop neuroprosthetics for epilepsy, Parkinson’s, chronic pain: the implant senses biological signals AND stimulates in response, using biologically-native patterns.

Medium-term (10-20 years):

• Ultra-low-power AI accelerators that run next to or integrated into sensors, using biological-plausible spiking neural networks rather than conventional deep nets.
• Hybrid bio-silicon systems where living neuron cultures and printed neuron networks interact for computation.

Context in the AI research landscape

April 2026 has been robust for foundational AI research:

• Physical Intelligence π0.7: compositional generalization in robotics
• Qwen3.6-35B-A3B: open-source sparse MoE efficiency gain
• Northwestern printed neurons: hardware substrate for post-GPU AI
• Google Gemma 4: open-weight edge deployment

The collective signal: AI research in April 2026 is simultaneously pushing on model architecture, robotics generalization, and compute substrate. The bottleneck for scaling AGI is moving from “how do we train a bigger model” toward “how do we deploy efficient AI at bio scales.”

Open questions

• Biocompatibility at scale. MoS2 is novel in neural implants. Long-term immune response in mammals beyond mice is unknown.
• Throughput. How many neurons per cm² can the printing process achieve at scale?
• Commercial timeline. The Northwestern team has not disclosed licensing or spin-off company plans as of publication.
• Head-to-head vs silicon neuromorphic. Hasn’t yet been benchmarked against Intel Loihi 2 or IBM NorthPole on standard spiking-network tasks.

Sources

• Northwestern Now: Printed neurons communicate with living brain cells
• Northwestern Engineering: Printed Neurons Communicate with Living Brain Cells
• ScienceDaily: Artificial neurons successfully communicate with living brain cells
• TechXplore: Printed neurons communicate with living brain cells
• Neuroscience News: Printable Artificial Neurons That Talk to Living Brain Cells
• EurekAlert!: Printed neurons communicate with living brain cells