AI search and answer firm Perplexity signed a "multi-year" agreement with Crusoe covering the development and operation of its frontier AI models.
Perplexity will use clusters powered by Nvidia's top-end GB300 systems to train its models and then deploy them through Crusoe's inference service, according to a Tuesday announcement. Inference is essentially the process AI models use to respond and generate answers for an end user.
The agreement will now put Perplexity's model training and operations on the same cloud platform.
The financial terms and compute capacity of the deal were not disclosed.
"Running AI at Perplexity’s scale means every millisecond of latency is felt by users," Perplexity co-founder and CEO Aravind Srinivas said. "It’s high-throughput, low-latency inference backed by next-generation hardware."
Perplexity claims that its AI answer engine handles more than 1.5 billion questions each month.
In return, Crusoe will also provide Perplexity Enterprise Pro and Max to its reported 1,800 employees for research and data analysis.
The agreement follows reporting that Crusoe raised more than $3 billion in a funding round valuing it at roughly $30 billion, nearly triple its $10 billion valuation after a 2025 funding round.
Earlier this month, Crusoe also reportedly signed a five-year compute contract worth around $13 billion with quant trading firm Jane Street.
Crusoe has been around since 2018, with roots as a bitcoin mining company. It has since shifted toward supplying data centers, computing systems, and cloud services for companies looking to develop and run AI models.