Fireworks expands European capacity onto Vera Rubin and Grace Blackwell, aiming to serve enterprise agent inference from 2027
Synopsis
Fireworks announced an expanded collaboration with NVIDIA to add capacity in Europe on the NVIDIA Vera Rubin and Grace Blackwell platforms, planning to use the NVIDIA DSX AI factory platform and Spectrum-X Ethernet, with the added capacity supporting European deployments beginning in 2027. The company also introduced enterprise inference capabilities: an Ultrafast inference preview on the NVIDIA Blackwell Platform reaching inference speeds exceeding 1,500 tokens per second, multi-day prompt caching and background mode for long-running agents, and higher rate limits with sub-minute cold starts. These target the latency and cost that arise when a single agent task involves repeated model calls and large amounts of shared context.
Interpretation
Fireworks will expand its capacity in Europe with the NVIDIA Vera Rubin and Grace Blackwell platforms, planning to draw on the NVIDIA DSX AI factory platform and NVIDIA Spectrum-X Ethernet for performance and throughput per watt. The expansion builds on Fireworks' work with NVIDIA to optimize inference for enterprise workloads, connecting accelerated computing with the software and model expertise companies need to run their own specialized intelligence. The stated schedule is that this NVIDIA accelerated computing footprint will support deployments across Europe beginning in 2027, meeting regional demand for open and specialized models.
For enterprise inference, Fireworks splits capabilities between interactive applications that need immediate responses and agents that keep working over extended periods. A single agent task can involve repeated model calls and large amounts of shared context, so response speed, context reuse and capacity all affect application performance and economics. Fireworks Ultrafast is in public preview on the NVIDIA Blackwell Platform, combining faster model computation with efficient communication across GPUs to reach inference speeds exceeding 1,500 tokens per second; multi-day prompt caching reduces the cost of reusing shared context across repeated calls, background mode lowers inference costs for work without an immediate response, and higher rate limits with sub-minute cold starts support scaling both interactive and background agent workloads.
Perspective
The report covers the expanded collaboration and inference capabilities Fireworks announced at its Forge user conference; the European deployment is planned to begin in 2027, and Ultrafast is in public preview.
The announcement does not give the size of the added European capacity, the model and input conditions behind Ultrafast's speed above 1,500 tokens per second, or the cost reduction from multi-day prompt caching and background mode.