Fireworks, the platform for specialized intelligence, today announced an expanded collaboration with NVIDIA. Fireworks will expand its capacity in Europe with NVIDIA Vera Rubin and Grace Blackwell Platforms. The planned expansion will also leverage the NVIDIA DSX AI factory platform and NVIDIA Spectrum-X Ethernet networking for leading performance and throughput per watt. At Fireworks Forge, its inaugural user conference, the company also introduced new capabilities spanning enterprise inference workloads: an Ultrafast inference preview for interactive applications that need immediate responses, multi-day prompt caching and background mode for long-running agents, and higher rate limits to support growing demand.
“With NVIDIA, we’re expanding our European capacity with the Vera Rubin and Grace Blackwell Platforms to help more enterprises run specialized intelligence at scale,” said Lin Qiao, co-founder and CEO of Fireworks. “Improving inference across the full spectrum of enterprise workloads will change the very nature of work, from interactive applications that need immediate responses to agents that keep working in the background. Every enterprise needs the speed, efficiency, and capacity to support both.”
This NVIDIA accelerated computing footprint will support deployments across Europe beginning in 2027. The added capacity will help Fireworks meet growing regional demand for open and specialized models as enterprises move more production AI onto the platform. 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 intelligence of open models has reached an inflection point, opening the door for enterprises to build specialized AI that transforms how they work,” said Jensen Huang, founder and CEO of NVIDIA. “I’m delighted to see demand for Fireworks skyrocketing. Together, we’re expanding Fireworks’ computing capacity with NVIDIA Vera Rubin and Grace Blackwell to fuel its spectacular growth and bring specialized intelligence to more enterprises.”
As enterprises deploy agents and more complex AI workflows, a single task can involve repeated model calls and large amounts of shared context. The speed of each response, the ability to reuse that context, and the capacity to accommodate demand all affect application performance and economics.
Enterprise AI workloads span a broad spectrum, from interactive applications where users are waiting for an immediate response to agents that work through complex tasks over extended periods. Fireworks is introducing capabilities tailored to each:
- Fireworks helps customers accelerate latency-sensitive workloads with Fireworks Ultrafast, now in public preview on the NVIDIA Blackwell Platform. Leveraging NVIDIA’s full-stack accelerated computing platform with over 1000 co-designed domain-specific libraries, AI models, and tools such as CUDA-X, Nemotron, and Nsight, Ultrafast combines faster model computation with efficient communication across GPUs to support responsive, real-time AI applications, achieving inference speeds exceeding 1,500 tokens per second.
- Multi-day prompt caching and background mode for long-running agents. Multi-day prompt caching reduces the cost of reusing large amounts of shared context across repeated model calls. Background mode dramatically lowers inference costs for work that can run without an immediate response. Together, these capabilities make it more economical for agents to keep working in the background over extended periods.
- Higher rate limits and sub-minute cold starts to support growing demand. Increased limits give customers more room to scale both interactive applications and background agent workloads.
About Fireworks
Fireworks is the specialized intelligence platform built to give companies control of their AI economics and ownership of the intelligence that sets them apart. Teams use Fireworks to run, ensemble, and train open-weight models, with day-zero access to new releases. Founded by the team behind PyTorch, Fireworks delivers trillions of tokens per day to companies around the world. Learn more at fireworks.ai.