Scality Launches On-Premises AI Inference Platform for Enterprises
Scality, a provider of enterprise data storage and AI infrastructure software, has introduced AI Inference Factory, an integrated software platform for running AI inference on enterprise-owned infrastructure rather than relying exclusively on cloud-based model services.
The platform combines validated open-weight models, inference-serving software, a workload control plane, and Scality AI Data Infrastructure (ADI). It supports models from providers including Mistral, Google, DeepSeek, and Qwen.
A central component is ADI's shared storage architecture, which extends key-value (KV) caching beyond GPU memory. This allows inference servers to retrieve previously processed context instead of repeatedly computing it.
In company testing, Scality reported KV cache retrieval speeds up to 72 times faster than recomputation for a 439,000-token context. These results have not been independently verified.
The platform also separates prompt processing from token generation, allowing organizations to scale those GPU workloads independently.
For enterprise IT teams, the offering addresses concerns about unpredictable inference costs, data residency and operational control as AI applications move into production. AI Inference Factory is available now through software licensing or as a fully managed service.
Posted by Pure AI Editors on 10/12/2026