What is Vultr?
Vultr is the world’s largest privately held independent cloud platform. Launched in 2014 as an alternative to the big three hyperscalers, our mission has been to deliver the most reliable, easy-to-use cloud platform with the best price-to-performance in the industry. We were founded by a group of engineers, 100% bootstrap — the only company any of us know in tech that scaled north of $100m in revenue profitably without outside capital, without a single person in sales or marketing. After passing the $100m threshold we brought in a new CEO, JJ Cardwell, to scale enterprise sales and marketing and to pursue the next wave of cloud infrastructure opportunities, notably GPUs for AI use cases.
Since then our business has expanded while staying true to being built by developers for developers. We’ve advanced our enterprise go-to-market to help mainstream developers inside larger enterprises and public sector agencies retool and re-envision the next decade of AI-native applications. We operate 32 global data-centre regions across all six inhabited continents, reaching around 90% of the world’s population in 2–40 milliseconds, and we maintain a hyperscaler-grade security and compliance posture — SOC, NIST, GDPR, DORA support and a FedRAMP roadmap — while remaining the price-per-performance leader in both core cloud and AI infrastructure.
With hyperscaler dominance — AWS, Google, Microsoft — how do you translate your differences into a strategy to capture market share in upcoming AI infrastructure?
We recently surveyed over 2,000 C-level executives responsible for AI budgets and found a major shift: only 18% plan to leverage an existing hyperscaler relationship for AI infrastructure, while 38% are open to partnering with an alternative provider like Vultr. The drivers are clear. Enterprises want access to the latest GPUs and networking components faster than hyperscalers often adopt them, and they don’t want to be constrained by hyperscalers’ push toward custom silicon. Staying current with NVIDIA or AMD generations matters to organisations on the innovation edge, and we aim to be first to market with those architectures.
Beyond hardware, enterprises value specialist skill sets and customer-centric support: we’re experts in AI infrastructure and can help architect deployments that accelerate time to market. There’s also a growing lack of trust in hyperscalers due to opaque cost structures and data-use concerns; organisations want alternatives to avoid vendor lock-in and to preserve ROI on cloud and AI investments. Our market strategy is therefore centred on combining up-to-date hardware availability with deep specialist support, global reach, compliance, and operational efficiency — which together make Vultr a compelling alternative.
You’ve raised capital and are investing heavily in GPUs — how do you see the split between training and inference in Vultr’s future workload?
The early market emphasised scale-up architectures: centralised clusters for training large LLMs, scheduled across data scientists. That remains important for some, but the mainstream enterprise market is focused on building and deploying applications and experiences. Most enterprises will adopt existing open-source models rather than develop custom models for every application — for example, taking models from Hugging Face, packaging them and deploying at scale. For that reason, the dominant workload will shift toward inference rather than bespoke training.
Inference requires a scale-out, distributed architecture — inference needs to happen close to end users to deliver real-time experiences with minimal latency. We do both: scale-up for training and, crucially, scale-out for inference. Projections already show inference spend growing as a share of AI infrastructure through 2026–2028, and that spend will be geographically distributed. Vultr specialises in scale-out architectures that minimise latency and enable the distributed inference footprint enterprises will need.
How do you decide where to build new data-centre regions — what metrics matter most?
It’s a combination of availability of power and land, local regulatory requirements, and — most importantly — where developers live and work, because latency to users is the primary performance metric. We’ve always prioritised price-performance, which translates into minimal latency and excellent user experience. The company’s roots supporting Activision’s global gaming franchise taught us how critical every millisecond is; that same thinking guides our region selection for AI and cloud workloads.
Within that latency and user-density rubric, we factor in regulatory requirements, access to space and power, and the economics of operating in a given market. The goal is to drive greater performance in the global infrastructure stack while ensuring compliance and cost efficiency — so regions are chosen to bring compute physically closer to users, meet local rules, and offer viable power and real estate economics.
What operational challenges have you faced in expansion?
Operationally we’ve focused on planned capacity, investing in and driving utilisation of that capacity. Our operating model — forged as a bootstrap company where every check comes from the founders’ pockets — demanded extreme efficiency in sourcing GPUs, space and power, and then rapidly outfitting and deploying racks customers can consume. That discipline yields our price-to-performance edge: we procure, secure and turn up racks quickly and with high utilisation.
That efficiency is our “special sauce.” Growing globally means executing that automated, repeatable process at scale: sourcing inventory, securing sites, outfitting space and deploying quickly without the luxury of unlimited capital. Our engineering background forced us to build that machine early and maintain it as we expand.
Given your multi-jurisdictional footprint, how do you handle data sovereignty, privacy and user content licensing?
Our platform was designed from day one with regional isolation and data residency in mind: each data-centre region runs in isolation from a networking perspective, so operational continuity and data residency are intrinsic to the architecture. Bits in a bucket in Paris are owned and processed only by the customer; we don’t access or process customer data. Backup and disaster-recovery services require customers to explicitly choose where backups go — we don’t automatically move or replicate data across regions without the customer’s instruction.
That model means customers control residency and sovereignty from day one, which is why we support GDPR, DORA and high standards like HIPAA in the U.S. We simply don’t touch customer data — we can’t process it without explicit configuration — and that clear separation addresses many concerns organisations have about where backups are stored and how data is handled across jurisdictions.
How ready is the United States infrastructure to handle the large-scale demand expected in the coming years?
The United States is exceptionally good at mobilising capital to create and recreate industries. We’re seeing rapid capital mobilisation to fund startups and the supporting infrastructure — data centres, power, and related assets. The U.S. is a global source of innovation and has repeatedly marshalled capital at scale to build new industries; I’m extremely bullish on its ability to do so again for AI infrastructure.
That said, the work requires coordination across investment, permitting, power and supply chains. But overall, the U.S. track record of scaling industries gives me confidence that it can mobilise the resources needed to meet the coming surge in demand.
Last couple of lighter ones — why the name “Vultr,” and do you find parallels between ultra-distance running and leading at Vultr?
We picked “Vultr” as a nod to the ecological role of vultures: they are crucial to a healthy, vibrant open ecosystem. Remove the vulture and the ecosystem slowly collapses; they keep systems healthy by cleaning up and cycling resources. The metaphor extends to our industry: an open, healthy AI ecosystem needs independent, specialised players who ensure innovation can thrive without collapse into single-vendor dominance.
On the marathon side, absolutely — there’s a direct parallel. Building the future of cloud and AI is a 10-year cycle that requires pacing, consistency and relentless execution. Ultra-distance running teaches you to keep moving forward at a steady pace under changing conditions, which is the mindset we apply at Vultr as we scale and innovate over the next decade.
