Neuromorphic processor platforms
A family of research and commercial chips implements large-scale spiking computation: digital many-core designs (Intel Loihi, IBM TrueNorth, the ARM-based SpiNNaker, BrainChip Akida) and mixed-signal/analog systems (BrainScaleS, DYNAP, and similar). They provide many programmable spiking neurons, event routing, and in some cases on-chip plasticity, and are the usual testbeds for neuromorphic BCI decoding — running or emulating decoders at low power on the bench.
Crucially, these are board- or system-scale platforms for prototyping and edge inference, not implantable devices: they are far too large, power-hungry, and unqualified for chronic biological use. Their role in BCI is to validate algorithms and estimate energy, informing the eventual design of a purpose-built, implant-grade neuromorphic front-end that does not yet exist in humans.
'Neuromorphic' spans a spectrum from analog subthreshold to fully digital SNN accelerators; only the low-power, event-driven end is relevant to an implant, and even that must shrink by orders of magnitude to be implantable.