CoreWeave Lands Perplexity AI Inference Deal

Key Takeaways

• Perplexity signed a multi-year agreement to run AI inference workloads on CoreWeave’s GPU cloud infrastructure.
• The deployment relies on NVIDIA GB200 NVL72 clusters designed for large-scale AI inference performance.
• CoreWeave continues expanding through major AI infrastructure deals with companies like Meta, OpenAI and now Perplexity.

CoreWeave has cemented its position as the go-to GPU cloud for AI-native companies, adding Perplexity to a client roster that already includes OpenAI and Meta.

The company announced on March 4, 2026, that it entered a multi-year strategic partnership with Perplexity to support the AI company's inference workloads on CoreWeave Cloud. The agreement also includes piloting new services across both organizations.

Under the deal, Perplexity will power its next-generation AI inference workloads using dedicated NVIDIA GB200 NVL72-powered clusters. The infrastructure aims to keep pace with Perplexity's rapid growth and the requirements of its Sonar and Search API ecosystem.

CoreWeave will also deploy Perplexity Enterprise Max across its organization, giving employees the ability to search the web and internal knowledge, run multi-step research and work with advanced AI models.

Table of Contents

• What the CoreWeave-Perplexity Partnership Includes
• CoreWeave’s Strategy: Power, GPUs and Hyperscale Contracts
• Scaling AI Requires Specialized GPU Infrastructure
• CoreWeave Background

What the CoreWeave-Perplexity Partnership Includes

Perplexity has begun deploying inference workloads on CoreWeave’s platform as part of the partnership’s initial rollout. The deployment combines specialized GPU infrastructure, orchestration services and model-management tools designed to support high-volume AI inference.

Infrastructure Component	Role in the Deployment
NVIDIA GB200 NVL72 clusters	Dedicated GPU infrastructure for inference workloads
CoreWeave Kubernetes Service	Managed orchestration for AI workload deployment
W&B Models integration	Model training, fine-tuning and management tools
Perplexity Enterprise Max	Web and internal knowledge search with AI models

CoreWeave’s Strategy: Power, GPUs and Hyperscale Contracts

CoreWeave executed an aggressive growth strategy throughout 2025 and early 2026, anchored by its $9 billion all-stock acquisition of Core Scientific in June 2025. The deal delivered approximately 1.3 gigawatts of gross power capacity across 10 US data centers plus over 1 GW of expansion capacity.

On the partnership front, the company secured a $14.2 billion agreement with Meta in September 2025 to supply cloud and AI compute infrastructure through 2031. Plus, CoreWeave's relationship with OpenAI expanded through three successive agreements totaling $22.4 billion, including an additional $6.5 billion cloud contract in September.

The company completed a successful IPO in March 2025, raising $1.5 billion in the largest US tech IPO since 2021. It subsequently acquired MLOps platform Weights & Biases for roughly $1.7 billion.

Despite closing 2025 with $5.13 billion in revenue and a $66.8 billion backlog, February earnings revealed a $1.17 billion net loss for 2025.

Scaling AI Requires Specialized GPU Infrastructure

Specialized GPU cloud infrastructure has become essential for enterprises deploying production-grade AI systems that demand low latency and operational efficiency.

Cloud providers like CoreWeave offer managed Kubernetes services alongside on-demand Nvidia GPU access, designed for flexibility and cost efficiency in AI agents and machine learning workloads.

CoreWeave Background

CoreWeave, founded in 2017, provides high-performance GPU compute infrastructure for AI-first organizations, research labs and creative studios. The company offers bare-metal GPU and CPU compute nodes, AI-optimized storage, networking and managed services such as Kubernetes. Its platform serves AI research labs, model builders and enterprises with large-scale training data needs.