# Alluxio

> Alluxio builds a distributed caching layer between AI and analytics compute and persistent storage. Founded from UC Berkeley research, it sells enterprise software that keeps frequently used data close to GPUs and query engines while leaving the durable source of truth in place.

- **Founded:** 2015
- **Headquarters:** San Mateo, United States
- **Founders:** Haoyuan (H.Y.) Li (Founder, chairman and CEO), Amelia Wong (Co-founder and VP of Marketing)
- **Products:** Alluxio Enterprise AI, Alluxio Enterprise Data, Alluxio Community Edition
- **Notable:** Fireworks AI reports 2–3 minute model loads for 800 GB-plus models at more than 100 concurrent replicas, down from 20-plus minutes or longer., Fireworks AI reports 800 GB/s to more than 1 TB/s aggregate throughput and a 50% reduction in multi-cloud egress costs in its deployment., Alluxio reported 99.57% GPU utilization for ResNet50 in MLPerf Storage v2.0.

## Products & services

- **Alluxio Enterprise AI** — Distributed caching for AI training, model distribution, inference, and checkpoint workflows, with enterprise support.
- **Alluxio Enterprise Data** — Unified data access and caching for analytics and SQL workloads across storage systems.
- **Alluxio Community Edition** — Free, Apache 2.0 licensed open-source distributed caching software oriented toward analytics workloads.

## Achievements

- Fireworks AI reports 2–3 minute model loads for 800 GB-plus models at more than 100 concurrent replicas, down from 20-plus minutes or longer.
- Fireworks AI reports 800 GB/s to more than 1 TB/s aggregate throughput and a 50% reduction in multi-cloud egress costs in its deployment.
- Alluxio reported 99.57% GPU utilization for ResNet50 in MLPerf Storage v2.0.
- Alluxio said its 2021 fiscal-year sales grew 3.5 times over the prior year and its project exceeded 1,100 GitHub contributors.

## Latest updates

- **2026-06** — Announced an AI infrastructure solution with Oracle Cloud Infrastructure focused on keeping GPUs supplied with data.
- **2025-08** — Released Enterprise AI 3.7 with sub-millisecond time-to-first-byte capability, faster cache preloading, and S3 role-based access control.
- **2025-05** — Released Enterprise AI 3.6 with model distribution, checkpoint writing, and multi-tenancy improvements.

## Links

- Website: https://www.alluxio.io/
- LinkedIn: https://www.linkedin.com/company/alluxio-inc-/
- Twitter/X: https://x.com/Alluxio
- GitHub: https://github.com/Alluxio
- YouTube: https://www.youtube.com/channel/UCpibQsajhwqYPLYhke4RigA
- Facebook: https://www.facebook.com/alluxio/

---

Profile page: https://yespress.io/alluxio
Published by YesPress — https://yespress.io
Last updated: 2026-09-29