In a fresh sign of how fiercely cloud providers are scrambling to meet the soaring demand for artificial intelligence compute, cloud vendor Vultr is committing more than $1 billion to build a major AMD-powered GPU cluster in Springfield, Ohio. The project, slated to come online in early 2026, marks the company’s largest infrastructure investment to date and demonstrates that AI is no longer the exclusive territory of the hyperscalers.
Vultr’s plan centers on a 50-megawatt installation equipped with 24,000 AMD Instinct MI355X accelerators. While the scale is far smaller than the gigawatt-class complexes being assembled by Microsoft, Google and Meta., Vultr’s value proposition is different. The company pitches itself as the cost-efficient alternative, offering cloud GPU compute at rates roughly half of what the hyperscalers charge. This pricing gap, the company claims, gives independent cloud providers room to compete, especially as demand for training and inference workloads accelerates far faster than supply.
What’s most striking about the Ohio buildout is that Vultr doesn’t yet have an anchor tenant lined up. The company is financing and constructing the massive AMD cluster on spec, a confident move in a market where customers often reserve capacity years in advance. The company says conversations with potential clients are underway and that demand signals are strong enough to justify moving forward.
Working with AMD
Vultr was an early adopter of AMD’s Instinct line, and this new deployment broadens that relationship considerably. The company plans to integrate AMD’s next-generation MI450 GPUs and the chipmaker’s Helios rack-scale architecture as they become available, reinforcing a roadmap that spans multiple product cycles. AMD, meanwhile, is pushing hard to capture share from NVIDIA, touting its accelerators as both high-performance and more cost-efficient. For AMD, a multi-year commitment from a fast-growing cloud provider offers an important validation point.
Yet Vultr isn’t going exclusive. NVIDIA GPUs remain part of the company’s broader strategy, a pragmatic nod to the reality that many customers insist on NVIDIA hardware for training their largest models. Still, the Ohio campus is notable for the sheer concentration of AMD silicon. It’s one of the largest such deployments announced by an independent cloud vendor.
Debate About GPU Depreciation Cycles
The state of Ohio, eager to position itself as a tech infrastructure hub, is backing the project with support from JobsOhio, the Dayton Development Coalition, the Greater Springfield Partnership and the governor’s office. For the region, the facility represents both economic growth and a foothold in one of the world’s most capital-intensive technology sectors.
Vultr’s expansion also highlights the financial engineering now required to build AI data centers. The company raised $333 million in equity last year, followed by $329 million in credit financing this June. Notably, $74 million of that financing was asset-backed, with GPUs serving as collateral. It’s a vivid illustration of how AI chips have become investment-grade assets, even as some analysts warn that depreciation cycles may shorten as hardware improves. In contrast, other experts counter that six years remains a conservative estimate for useful GPU life, and that industry fears of a bubble overlook how severely compute capacity still lags demand.
In any case, Vultr’s strategy seems to be: AI infrastructure is massively underbuilt, and independent clouds can grow by offering accessible performance at competitive prices. If that thesis holds, the Springfield campus will mark a turning point, both for Vultr and for the growing roster of cloud providers carving out space in an AI market once thought to be the exclusive domain of trillion-dollar giants.
