Vultr targets open composable stacks and AMD partnership to lead cloud AI infrastructure

Vultr's pitch to the cloud world right now is refreshingly unglamorous: cheaper, faster, everywhere. Kevin Cochrane, the company's chief marketing officer, laid out the case at AMD's Advancing AI event, telling theCUBE's Dave Vellante that Vultr runs across 33 global regions and claims up to 33% better performance at an average of 82% lower cost than the big hyperscaler offerings. That's a specific enough claim to invite scrutiny, but it also explains why Vultr keeps leaning so hard on its AMD relationship.

That relationship started on the CPU side with EPYC and has since grown to include AMD's Instinct GPU line. Cochrane frames AMD's predictable generational roadmap as the backbone of Vultr's whole price-performance argument. And he's not shy about resurrecting an unfashionable idea: the CPU, he says, never actually went away. It's just found new relevance as workloads shift toward things that don't need a GPU cluster the size of a small country.

The more interesting argument is about where compute needs to live. Cochrane says the training era spawned a wave of neocloud providers because customers just wanted GPU access wherever they could get it, geography be damned. Inference, he argues, flips that logic. Low latency matters, and so does proximity to actual users and employees, which is why Vultr insists on spreading investment across all 33 regions instead of building one mega-datacenter and calling it a day.

Sovereignty is the other pillar of Vultr's pitch, and Cochrane doesn't treat it as a passing political trend. He compares national AI infrastructure to telecom or water treatment — something every country eventually needs as basic infrastructure, regardless of who's in charge. Vultr's compliance work across its regions, he says, is turning into a real competitive edge as more governments start asking hard questions about where their data and models actually sit.

On the product side, Vultr announced a joint solution with VAST Data, SUSE and AMD aimed at robotics, financial services and healthcare, built as a prepackaged open composable stack customers can spin up from Vultr's marketplace in about half a minute. Cochrane is blunt about the philosophy behind it: no interest in competing with partners, just enabling the ecosystem. Looking forward, he points to enterprise contracts with five-year planning horizons and teases more capacity and financing news to come, with Vultr positioning itself as the specialist for the agentic era — chasing the most agents per watt at the lowest price per dollar.