Alibaba Cloud unveils CPFS storage, claims 69% reduction in AI training costs
At the 2026 Yunqi Conference on September 22, Alibaba Cloud launched its next-generation Cloud Parallel File Storage (CPFS) for AI training clusters, claiming up to hundreds of TB/s throughput and 100 PiB capacity. The company stated the system reduces model startup time by 50%, increases compute utilization by 30%, and lowers AI storage costs by 69%. Alibaba Cloud also introduced KVCacheStore, a caching engine for inference that reportedly improves cache hit rates by over 20%.
Reference imageEditorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page reads the event directly, while its address stays stable when the title changes.
- Summary awaiting refresh
Summary awaiting refresh
Cross-source coverage
Reporting timeline
Alibaba Cloud Launches New CPFS Storage, Cutting AI Costs by 69%
At the 2026 Yunqi Conference, Alibaba Cloud announced the launch of its next-generation high-performance storage solution, CPFS (Cloud Parallel File Storage), designed for AI training clusters. The new system offers throughput of up to hundreds of TB/s and billions of IOPS, with a single file system capacity expanded fivefold to 100 PiB. It aims to accelerate model training for clusters of up to one million GPUs and support large-scale model training and multimodal data processing. According to the company, in real-world training scenarios, the new CPFS can reduce average model startup time by 50%, increase peak compute utilization by 30%, lower AI storage costs by 69%, and double business processing capacity. The report is attributed to the Shanghai Securities News.
Read sourceAlibaba Cloud Launches New High-Performance Storage CPFS, Cutting AI Storage Costs by 69%
At the 2026 Yunqi Conference, Alibaba Cloud announced the launch of its next-generation, fully self-developed high-performance storage solution, Cloud Parallel File Storage (CPFS). Designed for next-generation AI training clusters, CPFS delivers up to hundreds of TB/s throughput and billions of IOPS. It increases single file system capacity fivefold to 100 PiB, aiming to accelerate training for million-card models and support large-scale model training and multimodal data processing. According to the company, in actual training scenarios, CPFS can reduce average model startup time by 50%, increase peak computing power utilization by 30%, lower AI storage costs by 69%, and double business carrying capacity. The report is sourced from Shanghai Securities News.
Read sourceAlibaba Cloud Launches Next-Gen CPFS Storage for AI Training at 2026 Apsara Conference
At the 2026 Apsara Conference, Alibaba Cloud officially launched its next-generation high-performance storage CPFS (Cloud Parallel File Storage) for AI training clusters. The system offers up to hundreds of TB/s throughput and hundreds of millions of IOPS, with a single file system capacity expanded fivefold to 100 PiB. It is designed to accelerate model training on million-GPU clusters and support large-scale model training and multimodal data processing. In real-world training, the new CPFS can reduce average model startup time by 50%, increase peak compute utilization by 30%, lower AI storage costs by 69%, and double business capacity. The new file system also features a 10x improvement in metadata performance and a 100x increase in the number of files, scaling to a single file system of 100 PiB capable of holding trillions of files. Enterprises can manage training corpora, model parameters, checkpoints, and experimental results in a unified file system without repeatedly splitting or migrating data for different training phases, and can scale capacity and throughput synchronously with GPU clusters.
Read sourceShow 4 older updatesHide older updates
Alibaba Cloud Launches New CPFS Storage, Cutting AI Costs by 69%
At the 2026 Yunqi Conference on September 22, Alibaba Cloud announced the launch of its next-generation high-performance storage system, CPFS (Cloud Parallel File Storage), designed for AI training clusters. The new system offers up to hundreds of TB/s throughput and billions of IOPS, with a single file system scaling to 100 PiB, a five-fold increase. It aims to accelerate training for million-card models and handle large-scale multimodal data. In practical training scenarios, it reduces average model startup time by 50%, increases peak compute utilization by 30%, lowers AI storage costs by 69%, and doubles business capacity. The system eliminates the need to split or migrate data across training phases. Alibaba Cloud also introduced KVCacheStore, a caching engine for inference that sits between GPU memory, host memory, and remote storage, achieving a 20%+ cache hit rate improvement. Alibaba Cloud storage product head Jiang Jiangwei stated the company will continue to innovate across compute, networking, and storage to provide a high-performance, scalable, and cost-optimized storage foundation for large model training and AI applications.
Alibaba Cloud launches new AI storage CPFS, claims 69% cost reduction
At the 2026 Yunqi Conference on September 22, Alibaba Cloud launched a new generation of high-performance storage, CPFS (Cloud Parallel File Storage), designed for next-generation AI training clusters. The company claims the system delivers up to hundreds of TB/s throughput and billions of IOPS, with a single file system scaling to 100 PiB, a five-fold increase. In practical training, Alibaba Cloud states the new CPFS can reduce average model startup time by 50%, increase peak computing power utilization by 30%, lower AI storage costs by 69%, and double business capacity. The system eliminates the need to split or migrate data across training phases. Additionally, Alibaba Cloud introduced KVCacheStore, a new storage layer for large model inference that sits between GPU memory, host memory, and remote shared storage. It uses a 'storage for compute' approach to handle large caches from long contexts and multi-turn dialogues, claiming a 20%+ increase in cache hit rate. Alibaba Cloud storage product leader Jiang Jiangwei stated the company will continue to advance computing, networking, and storage innovation.
Read sourceAlibaba Cloud Unveils New Storage System to Cut AI Costs by 69%
At the 2026 Yunqi Conference, Alibaba Cloud announced the next-generation CPFS (Cloud Parallel File Storage) designed for AI training clusters. The system delivers up to 100 TB/s throughput and billions of IOPS, with single file system capacity expanded fivefold to 100 PiB, supporting million-card model training. It addresses bottlenecks from massive small-file reads and checkpoint writes that cause GPU clusters to idle. The new CPFS uses a separated data and metadata service architecture with horizontal scaling, integrating Alibaba's Pangu distributed storage and a full-stack self-developed pipeline. It reduces model startup time by 50%, boosts peak compute utilization by 30%, and cuts AI storage costs by 69%. For inference, Alibaba Cloud also launched KVCacheStore, a storage layer that improves cache hit rates by over 20%. Alibaba Cloud storage product head Jiang Jiangwei stated the company will continue advancing compute-network-storage synergy to optimize AI infrastructure.
Read sourceAlibaba Cloud Launches New CPFS Storage, Claims 69% AI Cost Reduction at 2026 Cloud Summit
At the 2026 Yunqi Conference on September 22, Alibaba Cloud announced the next-generation Cloud Parallel File Storage (CPFS) designed for AI training clusters. The new system delivers up to hundreds of TB/s throughput and billions of IOPS, with single file system capacity expanded fivefold to 100 PiB. Alibaba Cloud claims the CPFS reduces model startup time by 50%, increases peak compute utilization by 30%, lowers AI storage costs by 69%, and doubles business capacity. The system supports unified management of training data, model parameters, checkpoints, and experimental results without data migration. Additionally, Alibaba Cloud introduced KVCacheStore, a storage layer between GPU memory and shared storage that improves cache hit rates by over 20% for large language model inference. Alibaba Cloud storage product head Jiang Jiangwei stated that the company will continue to coordinate computing, networking, and storage innovation to accelerate data processing for AI.
Read source