AMD, Supermicro and MinIO target the enterprise data pipeline bottleneck
SiliconANGLE Mark Albertson ● Covered by 2 sources
AMD, Supermicro and MinIO say enterprise AI’s real choke point is data movement, not raw compute. The fix they’re pushing is open lakehouse storage, up to exabyte scale.
Based on reporting by SiliconANGLE, Mark Albertson — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
The AI boom has not erased a very old problem inside big companies: getting data where it needs to go. AMD, Supermicro and MinIO say enterprises are still stuck with fragmented, mostly unstructured information that is hard for AI systems to query, govern and move. The result is a bottleneck in the data pipeline, not a simple shortage of chips.
Varun Selvaraj of AMD said many customers are discovering they do not have a big compute problem so much as a data pipeline problem. In his telling, the shift is moving away from raw processing power and toward the hard work of moving data cleanly across systems. That matters because AI infrastructure has to keep feeding CPU and GPU systems with massive amounts of information.
Greg DeMichillie of MinIO made the storage case bluntly: GPUs are too expensive to sit around waiting. He said the storage layer has to keep them productively fed, and that modern lakehouse design should eliminate fragmentation by keeping data in one storage environment. MinIO says it can do that at exabyte scale, and its AIStor platform supports Apache Iceberg tables and the Iceberg REST Catalog.
Supermicro is trying to package that approach with AMD EPYC systems, balancing CPU resources, PCIe lanes and memory across the data layer. Albert Tan said the company’s idea of open is flexibility and choice, with pre-validated systems that are ready to go in the data center. It is the usual enterprise promise, only with less poetry and more cables.
Iceberg is the bigger story underneath the sales pitch. DeMichillie called its rise a major shift, pointing out that open table formats were once a strange idea in a world where every database had its own proprietary format. Now, he says, wide adoption is letting vendors push Iceberg deeper into the storage tier, which gives enterprises a way to support both traditional analytics and AI workloads without locking themselves into one vendor’s house rules.
My take — AI-written commentary, not fact-checked reporting
This is one of those rare enterprise AI debates where the boring answer is probably the right one: storage, standards and plumbing matter more than another round of GPU worship. Open formats like Iceberg are winning because they make the mess slightly less expensive to live with. The industry keeps selling magic; the real breakthrough is making the data stop behaving like soup.
Read more about this at: SiliconANGLE