Breaking

Crusoe Signs Perplexity Multi-Year Cloud Deal for AI Training and Inference

September 15, 2026 at 13:36 EDT

Infra & Chips
AI Agents
Sales & Business

Crusoe and Perplexity announced a multi-year cloud partnership on September 15, 2026, putting Perplexity's model training, fine-tuning and production inference on Crusoe Cloud. The agreement gives Perplexity access to dedicated NVIDIA GB300 NVL72-powered clusters and NVIDIA InfiniBand Solutions, while Crusoe will deploy Perplexity Enterprise Pro and Max to 1,800 employees.

Sep. 15, 2026 | Crusoe + Perplexity

One cloud lane for Perplexity's model lifecycle

The partnership moves training, fine-tuning and production inference onto Crusoe Cloud, using NVIDIA GB300 NVL72 clusters and NVIDIA InfiniBand Solutions.

Blackwell Ultra GPUs per GB300 NVL72 rack

Grace CPUs in the same rack-scale system

1,800

Crusoe employees getting Perplexity Enterprise Pro and Max

1.5B+

Questions Perplexity says it answers each month

Managed Inference benchmark

Crusoe cites larger gains against vLLM on a Llama-3.3-70B four-node deployment.

9.9x

Time to first token

5.0x

Throughput

Full lifecycle path

Train frontier models on GB300 NVL72

Fine-tune within the same platform

Serve through Managed Inference

Positive reading

Access to high-end NVIDIA capacity and one platform for training, fine-tuning and production serving.

Open questions

Contract value, assigned capacity and deployment locations were not disclosed.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.