CoreWeave Prepares for Inaugural Conference as AI Backlog Tops $100B

Specialized cloud infrastructure provider CoreWeave Inc. is preparing to gather enterprise clients, partners, and technology analysts in San Francisco for its inaugural user conference, Fully Connected, scheduled for Sept. 30 through Oct. 1. The event comes as the firm expands its compute infrastructure tailored for artificial intelligence model training and operational deployment.

Founded in 2017 under a different corporate name, CoreWeave operated as a cryptocurrency mining business before pivoting in 2019 to rent out surplus, discounted graphics processing units. Since making that transition, the company has built dedicated cloud architecture designed to support large-scale machine learning training and real-time AI inference workloads.

Over the past five months, CoreWeave has introduced several specialized enterprise offerings. In May, the company launched autonomous self-improvement capabilities for AI agents using operational data, asserting the tooling can reduce training costs by over 40% and accelerate training speeds by approximately 1.4 times without compromising output quality. In September, the platform introduced a Physical AI Field Engineering service designed to connect industrial domain expertise with applied machine learning deployment.

The service rollouts follow substantial commercial expansion for the infrastructure provider. According to reporting by SiliconANGLE (https://siliconangle.com/2026/09/18/cloud-platform-coreweave-ai-thecube-fullyconnected/), CoreWeave disclosed in August that its revenue backlog reached approximately $104 billion as of June 30. That backlog figure excluded more than $25 billion in net new customer commitments secured during the early third quarter.

CoreWeave's hardware architecture relies heavily on tight integration with major semiconductor partners. In early June, the enterprise announced the industry's first bring-up and validation of Nvidia Corp.’s Vera Rubin NVL72 platform on CoreWeave Cloud, an integration requiring custom liquid cooling systems, rack control, high-performance networking, and multi-tenant isolation. A high-ranking Nvidia executive is scheduled to speak at the summit.

CoreWeave Chief Marketing Officer Jean English stated that infrastructure reliability and validation speed remain the primary drivers of client onboarding, noting that enterprises often migrate to CoreWeave after encountering latency or stability bottlenecks on competing platforms. John Furrier, an analyst at theCUBE Research, noted that enterprise software buyers are increasingly adopting an 'asset light' model to access advanced AI infrastructure without undertaking multi-billion-dollar direct capital expenditures in data center construction.