Dive Brief:
- Cost-conscious enterprises are leaning on object storage to feed data-intensive AI and ML workloads, according to a Tuesday MinIO report. The data software company surveyed 656 IT leaders.
- Organizations are planning to grow object storage deployments as they operationalize large language model technologies, the report found. Respondents said object storage would hold three-quarters of their data within the next two years, up from 70% at the time of the survey.
- The cloud storage market is expected to more than double in size to $128 billion by 2028 as enterprises accelerate adoption and use of generative AI and traditional ML capabilities, according to Omdia research published in June. Object storage comprises more than two-thirds of all cloud-based data repositories, the research firm said.
Dive Insight:
Generative AI technologies put a spotlight on enterprise data strategy, as executives sought tangible gains from the technology and IT leaders grappled with practical cost and security issues. Object storage, the cloud’s most ubiquitous and least pricey data repository, remained the go-to solution for enterprise architects.
More than half of respondents cited AI as the top reason for choosing object storage over block storage and file storage, the two other main options. Performance requirements and scalability were also key factors in the decision.
While vendors raced to embed generative AI tools in existing enterprise platforms, IT leaders also focused on traditional ML and analytics capabilities, MinIO found. The most common workloads for object storage included:
- Advanced analytics such as Spark, Presto/Trino, SQL Server and Snowflake, cited by 54% of respondents
- AI model training and inference, cited by 51% of respondents
- Data lakehouse storage, cited by 44% of respondents
“Traditional AI is very much alive and well,” MinIO said in the report. “Enterprises haven’t jumped on the generative AI bandwagon so enthusiastically that they’ve left traditional AI in the dust.”
To mitigate risk and control costs, hybrid data ecosystems are the norm. One-third of respondents said their organization had an even mix of cloud and on-prem infrastructure, while only 19% were entirely in cloud and just 3% were mostly or fully on-prem.
“Nearly all of our enterprise customers wanted to control their data,” Ugur Tigli, CTO at MinIO, said in the report. “They didn’t feel comfortable injecting their proprietary information into LLMs using cloud services.”
