Wasabi is taking on public cloud hyperscalers with fast-access object storage for AI use.
Traditionally, Wasabi supplies single-tier, S3-compliant object storage in its cloud at lower-than-Amazon prices and no egress fees. Thus was based on disk drive media. A change was signalled when it raised $32.5 million in October and said it was setting up AI data storage cloud facilities with export to external GPU server targets. Now it has launched its Fire high-performance storage class that supports AI workloads.
Co-founder and CEO David Friend said: “Object storage is the backbone of AI, but customers shouldn’t have to choose between speed and cost. With Wasabi Fire, we’re delivering NVMe performance at disruptive prices, allowing organizations to cost-effectively store the critical data needed to train AI.”
Wasabi Fire uses NVMe SSD media, saying that the service is “purpose-built for compute-intensive AI and ML training, real-time inference, high-frequency data logging, and media pipelines.” Fire delivers single-digit millisecond response time, and is five times faster than Wasabi’s regular S3 storage.
The price is $19.99/terabyte/month with no egress or API request charges, and this is claimed to a fraction of AWS fast S3 object storage cost. Amazon S3 Express One Zone, with the same single-digit millisecond latency, costs $0.11/GB/month. That’s $112.64/terabyte/month, 5.6 times more than Wasabi Fire, before taking egress charges into account.
Wasabi has also opened a new storage region in San Jose, Calif., as part of its relationship with IBM Cloud; it’s co-located with the IBM infrastructure. By having a Silicon Valley location, Wasabi can offer Fire-class storage to AI startups there.
Alan Peacock, General Manager of IBM Cloud, said: “We’re excited for Wasabi to expand into Silicon Valley with the IBM Cloud San Jose data center. Wasabi Fire on the IBM Cloud is designed to give clients the benefits of IBM’s secured enterprise-grade infrastructure.”
Wasabi Fire is available for early access in this region. Register to apply for access here.
