Inference efficiency becomes a major startup and developer battleground

Volantis reportedly raised $88 million for photonic-memory technology aimed at improving bandwidth and reducing token costs, while local-model tools and recent developer experiments show growing demand for cheaper, hardware-flexible inference. Performance, energy, and commercial claims require independent validation; orbital compute remains experimental rather than usable capacity.