Qumulo Intros Cloud AI Accelerator with Cisco to Create GPU Liquidity

Why It Matters

By removing data‑gravity bottlenecks, Qumulo’s accelerator can dramatically lower idle GPU costs and accelerate AI time‑to‑value, a critical advantage as enterprises scale AI workloads across hybrid clouds.

Key Takeaways

• Enterprise GPU utilization averages only 5%, leaving 95% idle
• Cloud AI Accelerator streams data directly to GPUs, avoiding copy stages
• NeuralCache predicts caching, cutting GPU data load times up to 64%
• Works across AWS, Azure, Google Cloud, OCI, and Cisco UCS
• Enables flexible AI workloads, reducing idle GPU costs dramatically

Pulse Analysis

The AI boom has exposed a stark inefficiency: most enterprise GPUs sit idle because data must be staged on flash storage before compute can begin. Industry vendors have responded by tightly coupling storage to GPU clusters, a fix that improves peak performance but leaves the majority of the hardware underutilized. Qumulo’s Cloud AI Accelerator tackles the root cause—data gravity—by delivering data on‑the‑fly from any location to the GPU, turning storage islands into a fluid data fabric.

At the heart of the solution is Qumulo’s Cloud Data Fabric, a distributed file and object repository that spans on‑premise, edge, and public‑cloud environments. NeuralCache, an AI‑driven predictive cache, anticipates data access patterns and pre‑loads blocks into CPU DRAM, slashing load times by up to 64 percent. Integrated with Cisco’s networking and security stack, the accelerator provides secure, low‑latency pathways to major AI services such as Microsoft AI Foundry, AWS Bedrock, and Google Vertex AI, all without replicating data.

For enterprises, the financial impact is compelling. Reducing the heavy‑load phase eliminates weeks of idle GPU time, translating into substantial cost savings and faster model training cycles. As AI workloads become more distributed, the ability to tap any available GPU capacity—whether in a data center, edge node, or public cloud—offers a strategic edge. Qumulo’s approach positions it as a key enabler of agile, cost‑effective AI infrastructure, challenging traditional storage‑centric GPU deployments and prompting competitors to rethink data‑fabric strategies.