Vertical AI pushes infrastructure beyond one-size-fits-all
SiliconANGLE Mark Albertson
AI infrastructure is getting split by industry, not built one way for everyone. That means healthcare, finance and manufacturing now want their own stacks, not generic cloud.
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
Enterprise AI is starting to look less like a single platform play and more like a set of industry-specific builds. That was the message from a Supermicro Open Storage Summit discussion featuring Supermicro’s Vince Chen, Kioxia’s Anders Graham, Western Digital’s Brad Warbiany and DDN’s Moiz Kohari: the common pieces of infrastructure still matter, but the way they get assembled depends heavily on the vertical.
Chen said the hard part is not just compute, storage and networking. It’s the software stack, too, and the way all of it gets tuned for a given customer. Supermicro is leaning into that by working with partners on vertically integrated, pre-validated options, so buyers can choose from set configurations instead of building everything from scratch.
The vertical pressure is clearest in sectors where data movement changes the economics. Kohari pointed to financial customers such as BlackRock and State Street, where moving data quickly can reduce capital lockup. He said DDN can help with calculations that, in turn, can free up capital for reinvestment or hedging. He said the same logic shows up in life sciences.
Storage vendors are making their own adjustments. Graham said Kioxia’s BiCS8 technology is now appearing in its CM9 and CD9P drives, which are replacing CM7 and CD8P drives in the Supermicro portfolio during the year. He described those newer drives as bringing large gains in power efficiency and performance, and said tiered storage architectures are already common at the ingestion stage because different use cases carry different cost points.
Warbiany drew a sharp line between AI demos and AI systems that have to run every day. The jump from terabytes to petabytes and exabytes changes the architecture fast, he said, and that is where integrated setups from companies like DDN and Supermicro become useful. Chen added that many enterprises may start in the cloud because it is easier, but data confidentiality and control of know-how can push them back toward private data centers when the stakes get higher.
My take — AI-written commentary, not fact-checked reporting
Vertical AI is the useful kind of AI hype: annoying to market, essential to deploy. The industry has spent years pretending one stack can suit every buyer, and now the bill is coming due. Generic cloud-first thinking still has its place, but the serious money and the serious risk are clearly moving toward purpose-built systems.
Read more about this at: SiliconANGLE