Northwestern engineers printed artificial neurons that can generate electrical signals realistic enough to activate living mouse brain cells. The science is strange and elegant. The infrastructure sentence is sharper. Mark Hersam tied the work to AI's power consumption, heat, nuclear buildout, and water-cooled data centers. [1]

On Sunday, the paper argued that printed neurons made the AI data-center problem biological. Monday's source language strengthens the frame: the brain is five orders of magnitude more energy-efficient than a digital computer, and today's AI scaling model is running into physical limits. [1]

The devices are not practical AI hardware. They are flexible, printed, molybdenum-disulfide and graphene systems that produce neuron-like spike patterns and communicate with living tissue. [1][2] The claim is mechanism, not miracle.

That is why the story belongs in Life rather than only Technology. AI's resource bill has been discussed as electricity, water, and zoning. Northwestern asks whether the better computer model is wet, soft, heterogeneous, and biological.

The caution is just as important as the wonder. A lab device talking to mouse tissue is not a replacement data center. But it changes the metaphor from bigger chips and more cooling to different materials, different signals, and a different tolerance for complexity.

The water-and-power story has not vanished. It has found a neuron.

-- KENJI NAKAMURA, Tokyo