CloudSphere Announces New AI Data Center Platform Built for Large-Scale Model Training
SEATTLE — Cloud infrastructure company CloudSphere Technologies announced Monday a new data center platform designed to support large-scale artificial intelligence training and inference workloads.
The platform, called CloudSphere AI Fabric, combines high-density GPU servers, high-speed networking, distributed storage and software tools for managing large AI clusters.
The company said the system was developed for enterprises and research organizations that need to train large language models, multimodal systems and other compute-intensive AI applications.
“AI infrastructure is becoming one of the most demanding workloads in modern data centers,” said Laura Chen, chief technology officer at CloudSphere. “Customers need computing, networking and storage to operate as a coordinated system rather than as separate components.”
CloudSphere said a single AI Fabric cluster can support thousands of accelerator chips connected through a low-latency network.
The company also introduced software that automatically allocates computing resources based on workload priority, model size and available hardware capacity.
According to CloudSphere, the management system can detect failed nodes and move workloads to healthy servers without requiring developers to manually restart entire training jobs.
The platform also includes tools for tracking GPU utilization, energy consumption, storage throughput and network performance.
AI developers can use the monitoring dashboard to identify performance bottlenecks and estimate the infrastructure cost of individual training or inference workloads.
CloudSphere said energy efficiency was a major focus of the new design.
The company has introduced liquid cooling in selected server configurations and uses automated power management to reduce energy consumption when computing resources are idle.
The announcement comes as technology companies continue investing heavily in infrastructure for generative AI.
Training advanced models can require thousands of specialized processors operating continuously for weeks or months, creating significant demand for electricity, cooling and high-speed networking.
CloudSphere said it is also expanding several data center campuses to support expected growth in AI workloads.
The company plans to add approximately 400 megawatts of additional computing capacity across facilities in North America and Europe over the next three years.
Some of the new facilities will use long-term renewable energy contracts, while others will include battery storage systems designed to reduce pressure on local electricity grids during peak demand.
Industry analysts said power availability has become one of the biggest constraints on new AI data center construction.
In some regions, developers are facing delays because local grids cannot provide enough electricity for large computing facilities.
CloudSphere said its infrastructure software can distribute certain workloads between different data centers based on computing availability, electricity cost and network capacity.
The company is also testing scheduling systems that can delay non-urgent AI training jobs until periods when electricity demand is lower.
The first CloudSphere AI Fabric systems are already being deployed for selected enterprise customers.
The company said wider commercial availability will begin early next year.
CloudSphere also plans to introduce smaller configurations aimed at companies that want dedicated AI infrastructure but do not require hyperscale computing capacity.
The company expects demand for specialized AI infrastructure to continue growing as more organizations move artificial intelligence projects from experimental environments into production systems.
