FuriosaAI's 180W Renegade (RNGD) inference accelerator, a 5nm chip built around a tensor contraction processing core, is now live inside Equinix's carrier neutral Lisbon datacentre, opening a new colocated inference option for European enterprises.
FuriosaAI frames its Equinix Lisbon deployment as a way to deliver "the RNGD accelerator's power efficiency" through Equinix's global colocation footprint, per CEO June Paik (Furiosa press release). The concrete read: a 5nm Tensor Contraction Processor at 180W per accelerator, up to eight per NXT RNGD Server, is now physically resident inside a carrier-neutral hall (RNGD spec sheet, developer overview).
European operators can colocate inference workloads on Korean-designed silicon without signing a managed-cloud contract, a thin but real opening for buyers wary of single-vendor stacks and datacentre operators hunting for density at constrained power budgets.
The May Broadcom rack-scale pact (HPCWire) explains how Furiosa intends to scale the platform. Lisbon is the first visible European proof point, not the roll-out.