Crusoe, a vertically integrated artificial intelligence infrastructure provider, announced on September 15, 2026 that it has entered a multi-year cloud partnership with Perplexity. The agreement will see Perplexity use Crusoe Cloud to train frontier models and serve them in production through managed inference services. It also gives Crusoe employees access to Perplexity Enterprise Pro and Max, while financial terms and allocated power capacity were not disclosed.
Scope of the Cloud Agreement
Under the arrangement, Perplexity will run its full model lifecycle on a single platform. Training workloads will be handled by dedicated clusters powered by NVIDIA GB300 NVL72 systems and NVIDIA InfiniBand Solutions. Production serving will move through Crusoe Managed Inference, a service designed to optimize latency and throughput for large-scale AI applications.
Technology and Performance
Crusoe positions its managed inference offering as a response to the real-time demands of AI products used by millions. The partnership is intended to support high-throughput and low-latency delivery, which Perplexity identifies as central to user experience. According to the companies, the infrastructure is engineered to maintain cost efficiency while supporting popular and frontier models in production.
Enterprise Productivity Access
Beyond infrastructure, Crusoe will adopt Perplexity Enterprise Pro and Max for its workforce of approximately 1,800 employees. The tools are expected to support web and internal knowledge search, multi-step research, data analysis, and access to advanced artificial intelligence models. This element connects the commercial relationship to enterprise productivity improvements across Crusoe's daily operations.
Perplexity Platform Scale
Perplexity operates as an AI agent platform that is focused on accurate artificial intelligence and was founded in 2022. The company says its flagship product now handles more than 1.5 billion questions globally every month. That level of usage makes real-time inference performance, availability, and system efficiency essential to the cloud relationship with Crusoe.
Leadership Commentary
Crusoe co-founder and CEO Chase Lochmiller said that fast-moving AI companies need infrastructure that scales across the entire model lifecycle. Perplexity co-founder and CEO Aravind Srinivas emphasized that at Perplexity's scale every millisecond of latency is felt by users. NVIDIA's Dion Harris noted that the unified computing approach enables training, fine-tuning, and serving with the speed required for agentic AI.
Commercial Terms and Market Context
Although the companies did not disclose the value of the agreement or the power capacity assigned to Perplexity, the deal adds to Crusoe's expanding roster of AI chip rental clients. Crusoe recently announced a contract worth close to 13 billion dollars with Jane Street over five years. This latest agreement indicates continued demand for high performance AI infrastructure services across both enterprise and consumer-focused platforms.
Broader Infrastructure Momentum
Crusoe's infrastructure capacity exceeded 4.9 gigawatts as of June, with advanced-stage projects surpassing 40 gigawatts. Its major commitments include a 1.2-gigawatt Oracle project and a 900-megawatt Microsoft campus in Abilene, Texas. The company also recently opened a 400,000-square-foot manufacturing facility in Tulsa, Oklahoma, bringing its total U.S. manufacturing footprint to roughly one million square feet.
By combining dedicated NVIDIA systems, managed inference services, and enterprise software adoption, the partnership aligns operational scale with product performance. It illustrates how AI infrastructure providers are moving beyond basic compute rental to integrated training and deployment offerings. The agreement marks another step in Crusoe's rapid expansion and Perplexity's effort to strengthen services for both consumer and enterprise users.