Vultr

What Vultr actually is

Vultr is the flagship of The Constant Company, LLC — one of the few independent clouds left that can genuinely call itself global. The service launched in February 2014, and the company's own numbers are hard to argue with: over 80 million instances deployed, more than 1.5 million customers across 150+ countries, and a footprint of 34 cloud data center regions spanning North and South America, Europe, Asia, Australia, and Africa. That breadth is the product. Where most providers in /category/vps-providers are regional plays with a story about expansion, Vultr will sell you a small VM in Santiago, Chennai, Honolulu, Johannesburg, or Melbourne today, from the same console, at the same price structure.

The platform is broad but legible. Cloud Compute covers the everyday VMs: Regular Performance (previous-generation Intel, regular SSD) from $2.50/month, and a newer VX1 line pitched at core infrastructure with dedicated CPU, up to 50 Gbps networking, bootable block storage, and provisioning in under 15 seconds. Optimized Cloud Compute adds NVMe-backed instances with no noisy neighbors, split into CPU-, memory-, and storage-optimized shapes plus High Frequency. Cloud GPU sells virtualized NVIDIA (A100, H100, L40S, GH200) and AMD (MI300X, MI325X) accelerators by the hour; Bare Metal provides single-tenant physical servers, including GPU machines and self-service clusters. Around the compute sit managed PostgreSQL/MySQL/Valkey/Kafka, S3-compatible object storage, block storage, file storage, load balancers, a CDN, cloud firewalls, Direct Connect, BGP bring-your-own-IP, and Vultr Kubernetes Engine. The API and Terraform provider are first-class: this is an API-first platform with a good console, not a hosting panel with an API bolted on.

Behind it is real infrastructure — the datacenter partners read like a who's who of colocation (Digital Realty, Equinix, QTS, CoreSite, Interxion, GTT) with NTT, Telia, and Lumen-style transit — and a standardized platform image model that means a VM in Tokyo behaves like a VM in Amsterdam. Our listing's old record claimed Vultr launched in 2026; that is plainly wrong and we correct it below. The company is US-based (The Constant Company, LLC), privately held, with Series D-stage financing reported in 2025.

Pricing: what you really pay

Vultr publishes both monthly and hourly rates for everything in the compute catalog, which is the right default. Cloud Compute Regular starts at $2.50/month ($0.004/hour) for 1 vCPU, 0.5 GB RAM, 10 GB storage, and 0.5 TB transfer — an IPv6-only plan — with the cheapest IPv4-enabled plan at $3.50/month. The familiar $5/month plan gives 1 vCPU, 1 GB RAM, 25 GB, and 1 TB transfer; $10/month gives 1 vCPU, 2 GB, 55 GB, 2 TB; $20/month buys 2 vCPU, 4 GB, 80 GB, 3 TB. High Frequency starts at $6/month (1 vCPU, 1 GB, 32 GB NVMe, 1 TB) and $12/month (1 vCPU, 2 GB, 64 GB, 2 TB). Optimized Cloud Compute runs from roughly $28/month for 1 dedicated vCPU/2 GB up to $3,840/month for 96 vCPUs with 256 GB RAM and 12 TB transfer; memory- and storage-optimized shapes price in the same bands. Most plans are capped at 672 hours per month; VX1 instances bill actual hours with no cap.

Costs that sit outside the headline price matter here. Windows licensing is extra and varies by plan. Automatic backups add 20% to the instance's base price — Vultr's own documentation is explicit about this — and snapshots are billed by storage (third-party references put the rate at $0.05/GB/month). Egress beyond the included transfer allowance is metered. The terms of service are also unambiguous about refunds: there are none, billing is auto-renewing, and Vultr reserves an early termination charge for services stopped mid-cycle. Budget accordingly.

On value, we have our own dated numbers, and they are the most useful thing on this page. In our 2026-04-01 test runs, a Vultr High Frequency 2GB instance in Amsterdam measured a sysbench CPU score of 4,100, 38,000 fio IOPS, and 1,800 MB/s sequential reads for about €12/month. On the same day, Hetzner's CX22 in Nuremberg (2 vCPU, 4 GB, 40 GB NVMe) returned 4,820 / 42,000 IOPS / 2,100 MB/s at €4.35/month, and UpCloud's Simple 2-2GB in Helsinki hit 4,600 / 62,000 IOPS / 2,800 MB/s at €14/month. Vultr's disk numbers were the weakest of that group and its price sits between the two. The complete table is on /benchmarks. The conclusion we draw is not that Vultr is slow — 38,000 IOPS is fine for most production work — but that you pay a premium for the footprint, not for the hardware, and buyers whose users sit near a Hetzner or UpCloud region should do the arithmetic.

Performance & infrastructure

The hardware story is consistent across the catalog: current-generation Intel and AMD CPUs on the newer lines, NVMe on all Optimized tiers, 10 Gbps+ uplinks with localized peering that the network team actively optimizes per region, and per-region test IPs published openly so you can measure latency before buying. DDoS mitigation covering layer 3 and layer 4 attacks is part of the platform. Snapshot-based image management, custom ISOs, one-click Marketplace apps, cloud firewalls, VPC networking, and 50 Gbps networking options on VX1 round out a platform that has spent a decade accumulating the boring features that production teams actually need.

Reliability is backed by the strongest SLA language we have seen from a provider at this price: a 100% uptime guarantee covering network and host-node availability, with credits requested via support ticket. That is a bold commitment, and while no provider hits 100% in practice, writing it into the SLA sets a real contractual floor rather than an aspiration. Vultr's decade-long run, 80M+ deployments, and standardized platform across all regions suggest operational maturity; the company's status history and public incident handling are better documented than most of the mid-market.

We benchmarked one instance — High Frequency 2GB in Amsterdam — and the numbers were respectable on CPU (4,100 sysbench) and unremarkable on storage (38,000 IOPS, 1,800 MB/s) for an NVMe-branded tier. One instance is one instance; treat it as a data point, not a verdict. What we can say editorially is that Vultr's platform consistency is the real performance feature: identical provisioning, images, and billing behavior across 34 regions means fewer surprises when you spread a fleet across continents.

Where it falls short

Price-performance is the recurring theme. Across the board — entry VMs, high-frequency tiers, optimized shapes — comparable European providers deliver more compute and faster storage per euro in their home regions. If your workload lives in one region and does not need Vultr's geography, you are paying for map coverage you do not use.

Then the nickel-and-dime list. Backups at 20% of instance price scale awkwardly: on a $96/month High Frequency plan that is nearly $20/month for automation you could script with snapshots. Snapshots themselves bill per GB and keep accruing until deleted. Windows is extra. Egress beyond plan allowances is metered. And the terms allow no refunds — a real friction point for a platform that otherwise courts experimentation.

Support is ticket-based with no phone line; for a platform serving 1.5 million customers, response quality varies and the enterprise experience is a different product from the self-serve one. The catalog's rapid evolution also creates churn — the VX1 line's introduction alongside Cloud Compute means buyers must compare several overlapping instance families to know if they are getting the current best deal. Our own database, and even Vultr's own pages, disagree on the region count (32, 33, and 34 all appear across its site); we use the locations page's 34.

Finally, managed means infrastructure, not operations: managed databases and Kafka exist, but nobody patches your VMs. Teams needing managed OS support must look at managed-cloud providers instead.

Who should pick this

Our pick when geographic reach is the requirement. If you need low latency in São Paulo and Seoul from one bill, one API, and one console — or you need a region nobody else in this comparison covers, like Santiago, Honolulu, Tel Aviv, or Johannesburg — Vultr is the default answer, and the per-region test IPs let you verify before you commit. It is also a strong pick for API-driven fleets: mature REST API, Terraform provider, one-click apps, bring-your-own-IP via BGP, and consistent behavior across regions.

GPU and bare-metal buyers get one of the broadest self-serve accelerator catalogs anywhere (NVIDIA and AMD, VM or dedicated, hourly), which makes Vultr a practical alternative to hyperscaler AI pricing for inference and training at small-to-medium scale. Developers who value an established, well-documented platform with a 100% network-and-node SLA over rock-bottom pricing will be comfortable here.

We would choose Hetzner Cloud in Europe, or UpCloud where storage performance dominates, when pure price-performance is the goal — our /benchmarks numbers make that trade explicit. And teams needing hands-on managed support or refunds as a safety net should weigh those policies before deploying production.

Key takeaways

• Vultr (The Constant Company, LLC; launched February 2014) is one of the largest independent clouds: 80M+ instances, 1.5M+ customers, 34 regions on six continents, all sold through one console, API, and price list.
• Pricing is hourly and monthly and fully published: $2.50/month IPv6-only entry, $5/month for 1 vCPU/1 GB/25 GB/1 TB, High Frequency from $6/month, optimized/dedicated shapes up to $3,840/month, plus GPU and bare metal.
• SLA is a 100% network and host-node uptime guarantee with ticket-requested credits — the strongest written commitment in its price class.
• Our 2026-04-01 test of a High Frequency 2GB (Amsterdam): CPU 4,100 / 38,000 IOPS / 1,800 MB/s — solid CPU, the slowest NVMe storage of the six providers we measured, at €12/month.
• Watch the extras: automatic backups add 20% of instance price, snapshots bill per GB, Windows and extra egress cost more, and the terms allow no refunds.
• In our view the pick for global footprint, API-first fleets, and self-serve GPU; for single-region value in Europe, Hetzner Cloud wins on the numbers (see /benchmarks).

Sources

• https://www.vultr.com/pricing/
• https://www.vultr.com/locations/
• https://www.vultr.com/company/about-us/
• https://www.vultr.com/legal/sla/
• https://www.vultr.com/legal/tos/
• https://www.vultr.com/products/bare-metal/
• https://www.vultr.com/kubernetes/
• https://docs.vultr.com/support/platform/billing/how-much-does-it-cost-to-enable-automatic-backups
• https://docs.vultr.com/products/storage/backups/faq
• https://platform.tracxn.com/a/d/company/53326a87e4b0398cae238fe8/vultr