Mushroom-Based Chips Show Living Mycelium Can Do Real Computing

    A DARPA-funded study in Scientific Reports describes morphologically tunable mycelium chips for physical reservoir computing, bio-based analog hardware made for under a dollar a chip.

    Mushroom-Based Chips Show Living Mycelium Can Do Real Computing A DARPA-funded study in Scientific Reports describes morphologically tunable mycelium chips for physical reservoir computing, bio-based analog hardware made for under a dollar a chip. Aaron Rafferty July 01, 2026 Key Takeaways A DARPA-funded study describes working computer chips grown from living fungal mycelium that use the material's structure to compute. The method, called physical reservoir computing, lets the researchers tune what a chip does by shaping how the mycelium grows. The team reports more than three million chips per harvest at under a dollar each, though the approach remains an early proof of concept. A team backed by the Defense Advanced Research Projects Agency says it has built working computer chips out of living mushroom material, and that the grown structure of the fungus can do real computation. The work, published June 10 in Scientific Reports and first posted as a bioRxiv preprint , describes what the authors call morphologically tunable mycelium chips for physical reservoir computing. Reservoir computing uses the messy dynamics of a physical material, in this case the branching network a fungus grows, to process signals the way a small neural network would. By shaping how the mycelium grows, the researchers say they can tune what the chip computes, a method they describe as programming with morphology rather than code. Cheap, grown, and analog The pitch is as much about manufacturing as computing. The team reports that a single harvest can yield more than three million chips, that the chips stay stable after three months, and that each one costs less than a dollar. The group, which includes Ecovative founder Eben Bayer, lays out a design, grow, and compute pipeline that models

    Loading full article…