CoreWeave (CRWV) saw its shares surge nearly 6% in premarket trading on Wednesday after announcing a multi-year agreement to support inference operations for Perplexity, an emerging AI-driven search engine backed by Jeff Bezos and Nvidia.
As part of the deal, CoreWeave will become a key backend cloud partner for Perplexity AI. The company will run its next-generation inference tasks on dedicated NVIDIA GB200 NVL72 clusters operated by the cloud provider.
The platform will serve as a foundation for Perplexity’s Sonar and Search API products as they expand, as noted by the companies.
“AI applications running in production require more than just access to raw infrastructure – they require best-in-class performance and reliability as well as a cloud platform designed end-to-end for AI that simplifies compute operations,” Max Hjelm, senior vice president of revenue at CoreWeave, noted.
AI inference is the real-time execution phase of AI models, when trained models are used to make predictions or generate outputs based on new input data. This process can vary from answering questions, making recommendations, classifying data, to powering real-time features like search results, image recognition, or language translation.
For Perplexity’s product ecosystem, inference speed, latency stability, and scalability directly affect the user experience.
“We’re proud to partner with Perplexity as they scale their inference workloads on CoreWeave’s AI cloud,” he stated.
Dmitry Shevelenko, chief business officer at Perplexity, highlighted the provider’s technical capabilities and collaborative approach as key factors in the decision.
“We were impressed by the combination of CoreWeave’s technical aptitude and partner-first mindset that help AI-native companies accelerate their growth and scaling goals,” said Shevelenko, recognizing the role of CoreWeave in enabling Perplexity to improve infrastructure efficiency and model quality for delivering powerful AI search and automation services across sectors.
The search firm has already begun deploying workloads using the cloud provider’s Kubernetes service. It is also using W&B Models for training and fine-tuning as part of a broader multi-cloud strategy.
Specialized GPU cloud operators have become increasingly vital partners for AI companies facing growing computational demands. CoreWeave has posted leading results in MLPerf benchmarks and holds platinum rankings in SemiAnalysis ClusterMAX evaluations for performance and reliability.
The arrangement also sees the cloud company adopt Perplexity Enterprise Max internally, giving employees access to web search, research tools, and advanced AI models through a single interface.