As demand for AI accelerates, access to high-performance computing has become one of the most important constraints on growth. This has helped create a new category of cloud provider: the neocloud.
A typical neocloud arrangement involves three parties. The data center operator, the neocloud provider, and the neocloud customer. Each relationship is governed by its own commercial agreement and SLA
Tsafrir Oranski, Parametrix
Neoclouds specialize in delivering GPU-based computing infrastructure for AI workloads. Unlike traditional cloud providers, which offer a broad portfolio of infrastructure and software services, neoclouds are generally focused on GPU servers, racks and clusters, high-performance networking, storage, and the management tools required to operate large-scale compute environments.
Their growth is helping solve a critical problem for AI companies: access to scarce and expensive computing capacity. But the neocloud model also creates a contractual exposure from the potential mismatch between the Service-Level Agreements (SLAs) that protect the neocloud and the SLAs they provide to their customers.
