Currently, the Monod implementation runs on a single platform. The faster the CPU, the better — it is extremely CPU intensive, in part because of the matching which takes place in the Swarm layer (especially as it is extremely inefficient right now! FIXME: that). There are plans to extend Monod to run in a distributed/clustered fashion across multiple platforms. And it is conceivable that Monod could accommodate non-standard computational platforms.
However, this is not the case right now.