Who actually owns the hardware? A map of budget VPS provider infrastructure

Three tiers of ownership

The distinction isn't binary. There's a spectrum between "we built the building and racked every server" and "we rent someone else's VPS and put our logo on it," and most budget providers sit somewhere in between.

Tier 1: Own the datacenter and the hardware. The provider operates the physical facility — power, cooling, security, network — and owns every server inside it. They control the overselling ratio, the hardware refresh cycle, the network topology, and can physically walk to a machine when it fails. Hetzner, Contabo, BuyVM, and DartNode sit here. The upside is total control. The downside is total responsibility: when a disk dies at 3am, there's no upstream provider to page.

Tier 2: Own the hardware, colocate the facility. The provider buys and manages its own servers — choosing the CPU, RAM, drives, and network cards — but racks them inside a third-party datacenter facility (Equinix, Interxion, Telepoint, etc.). They control the software stack and the metal, but not the building. HostHatch, GreenCloudVPS, AlphaVPS, and Netcup operate this way. The upside is premium facility access (Equinix AM3, Interxion ZUR1) without the capital cost of building a datacenter. The downside is less control over power, cooling, and physical security.

Tier 3: Reseller / upstream capacity. The provider rents server capacity from one or more upstream hosts, deploys VPS instances on it, and sells them under its own brand. RackNerd and SpeedyPage are examples. The upside is flexibility — a reseller can switch upstream providers, add new locations fast, and doesn't bear hardware failure risk. The downside is that they can't control what they don't own: the overselling ratio, the disk I/O policy, and the hardware quality are set by whoever's upstream.

The ownership map: who has the receipts

Every claim below is sourced from provider disclosures, LowEndTalk community verification, datacenter directory listings, or public ASN records. "Owns hardware" means the provider has publicly stated or demonstrated that they purchase and manage their own physical servers, not just that they have a nice-looking dashboard.

Own datacenter + hardware

Hetzner — operates three datacenter parks: Nuremberg and Falkenstein in Germany, and a facility in Tuusula (near Helsinki), Finland. These are purpose-built facilities that Hetzner designed and constructed, not leased space. The German parks alone house tens of thousands of servers. Hetzner also has colocation presence in Ashburn, Virginia and Hillsboro, Oregon for US coverage, but the European facilities are fully owned. ISO 27001 certified. This is the largest infrastructure owner in the budget tier by a wide margin.

Contabo — built two datacenters from scratch in Germany: Munich (2009, capacity for 4,000 servers) and Nuremberg (2014, capacity for 10,000 servers). Contabo has publicly stated they "build everything themselves, not relying on external contractors." They've since expanded to 12 global locations, though it's not clear whether all expansion sites are owned facilities or colocation. The German core is definitely theirs. This is why Contabo can offer 4 vCPU / 8GB RAM for $4.50/month — they're not paying someone else's margin on the hardware.

BuyVM / FranTech — owns hardware across four locations: Las Vegas, New York, Miami, and Luxembourg. Founded by Francisco Dias, BuyVM has been operating since 2010 and built its own control panel (Stallion), its own DDoS filtering infrastructure (500+ Gbps, L4-L7), and its own InfiniBand RDMA fabric for Block Storage Slabs. The level of in-house engineering is unusual: most budget providers buy off-the-shelf panels and outsource DDoS to Cloudflare or Path.net. BuyVM builds the stack.

DartNode — owns its primary datacenter in Spring, Texas, near NASA's Johnson Space Center, with a 24/7 on-site NOC. Additional locations (Kansas City, Orlando, Los Angeles, Netherlands, Germany, UK, Singapore, Japan) are likely colocation. DartNode is the youngest infrastructure owner on this list, and the growing pains showed in 2025: servers offline for extended periods, backup restores that failed, support tickets waiting a week. Owning the building doesn't mean running it flawlessly — but it does mean DartNode controls the hardware that makes its $2/month dedicated-core VDS Slices possible.

Own hardware, colocated

HostHatch — owns Dell EMC servers running AMD EPYC 7R13 processors with Samsung Enterprise NVMe drives. The hardware is colocated in premium facilities: Equinix AM3 (Amsterdam), Interxion ZUR1 (Zurich), GleSYS (Stockholm), plus facilities in London (Iron Mountain), Oslo (Blix), Vienna, and locations across the US, Singapore, Hong Kong, Japan, and Australia. HostHatch doesn't build datacenters — it buys enterprise-grade hardware and places it in Tier III+ facilities. The trade-off is that you get hardware quality that matches providers at twice the price, but support is understaffed and promotional plan customers are explicitly deprioritized.

GreenCloudVPS — owns all of its hardware, operating as a registered LLC in Delaware since 2013. The infrastructure runs AMD EPYC Milan and Genoa processors with NVMe 4.0 storage across 30+ datacenter locations globally. GreenCloudVPS manages its own network and was voted the #1 LowEndTalk provider in both 2024 and 2025. The company has publicly stated it's debt-free — an unusual detail that suggests the hardware is paid for, not leased. The Trustpilot score (around 2.9) reflects KYC and refund policy friction, not hardware quality.

AlphaVPS / DA International — owns its hardware and operates its own network (AS203380) from Sofia, Bulgaria, where the company offices sit directly inside the Telepoint-East datacenter — literally above their own equipment. Founded in 2013, AlphaVPS has expanded to five locations across Europe and the US. The network runs on 500 Gbps Juniper hardware with Tier-1 upstreams (Lumen, GTT, TI Sparkle). Like GreenCloudVPS, this is a provider that owns the metal but not the building.

Netcup — operates its own infrastructure in Nuremberg (Germany) and Vienna (Austria), connected to the Anexia Backbone Europe with direct peering at DE-CIX, N-IX, and AMS-IX. Uses Dell and HP servers with Juniper networking. Netcup keeps replacement hardware in stock at each facility for same-day component swaps. They've expanded to Manassas, Virginia and Singapore.

Resellers (upstream capacity)

RackNerd — uses colocation from upstream providers, historically including MultaCom for its Los Angeles DC02 location (which RackNerd shut down in 2026, migrating customers elsewhere). RackNerd operates across multiple third-party datacenters in the US. Being a reseller is what enables RackNerd's aggressive pricing and flash-sale model — they can provision capacity fast and switch upstream when a facility doesn't work out. The LA DC02 shutdown is actually an example of the reseller model working as designed: bad upstream, exit, move customers.

SpeedyPage — runs AMD Ryzen 9950X and 7900 processors with DDR5 RAM and Gen4 NVMe across locations in London, Los Angeles, Ashburn, Singapore, Tokyo, Sydney, Amsterdam, and Stockholm. The hardware choices are premium — Ryzen 9950X is a current-generation desktop processor rarely seen in VPS hosting. But SpeedyPage hasn't disclosed datacenter ownership, and the breadth of locations suggests colocation rather than owned facilities. What SpeedyPage controls is hardware selection and support quality (UK-based, sub-hour response times documented by users) — which may matter more than building ownership for most buyers.

What ownership gives you

This isn't abstract. Here are the specific things that exist in the budget tier because the provider owns the hardware:

• Unthrottled disk I/O. HostHatch's NVMe line delivers 3,000+ MiB/s and uncapped IOPS because HostHatch controls the hypervisor policy on its own Dell EMC servers. Contabo uses NVMe branding but throttles IOPS to roughly 1,300 on its SSD line because it has different priorities for its owned hardware: density over speed. Ownership enables the choice; what the provider chooses varies.
• No-overselling policies. BuyVM deliberately limits the number of VPS instances per physical node, which is why it's perpetually sold out. A reseller can't make this promise — they don't control how many tenants share the upstream node.
• Dedicated CPU cores at budget prices. DartNode's VDS Slices ($2/month for a dedicated physical core) exist because DartNode controls the CPU allocation on its own hardware. A reseller selling shared upstream capacity can't guarantee zero CPU steal.
• Custom infrastructure. BuyVM's Block Storage Slabs run on an InfiniBand RDMA fabric — a custom storage network that BuyVM designed and deployed on its own equipment. You don't see resellers building InfiniBand fabrics.
• Rare locations without rare pricing. HostHatch can offer Stockholm, Zurich, and Vienna at $4/month because it owns the servers placed in those premium facilities. A reseller in Equinix Zurich would need to add the facility's margin on top.

What ownership costs you

The budget VPS community tends to treat "owns their own hardware" as an unqualified positive. It isn't. Ownership comes with structural costs that affect you as the customer:

• Growing pains hit harder. DartNode's 2025 is the clearest example. When you own the datacenter and something breaks — power, cooling, network, a storage array — there's no upstream to absorb the failure. DartNode had servers offline for over two weeks, backups that failed to restore, and support tickets waiting 7 days per response. A reseller in the same situation would have migrated affected customers to a different upstream provider.
• Availability constraints. BuyVM's refusal to oversell means the product is frequently unavailable. If you need a VPS today and BuyVM is sold out, the ownership philosophy that makes it good also makes it inaccessible.
• Slower geographic expansion. Building or colocating hardware in a new location requires capital and lead time. RackNerd can spin up a new DC partnership in weeks. HostHatch adding a new city means buying servers, negotiating colocation terms, deploying networking, and testing — months of work.
• Support stretches thin. HostHatch openly stated that promotional plan customers are deprioritized in the support queue. The reason is structural: a small team that owns and maintains hardware has less bandwidth for tickets than a reseller that outsources infrastructure management to its upstream. Owning the metal means the same engineers fixing hardware also answer your ticket.

How to use this map

This isn't a ranking. It's a filter. Use ownership tier as a constraint alongside price and specs, not instead of them.

You probably want a hardware owner if:

• Your workload depends on specific hardware behavior — unthrottled NVMe, dedicated cores, or unmetered bandwidth with no hidden soft cap
• You're deploying in a rare location (Stockholm, Zurich, Vienna) where resellers either don't exist or charge cloud-platform prices
• You're running a set-and-forget workload that doesn't need responsive support but does need consistent performance
• You care about European data residency and want a provider that can point to a specific building (Hetzner Falkenstein, Contabo Munich, Netcup Nuremberg)

You probably want a reseller if:

• You need a VPS today — resellers have stock, hardware owners often don't
• You want aggressive promotional pricing (RackNerd's flash sales are enabled by the reseller model's lower capital requirements)
• You value operational flexibility over hardware control — if a facility has problems, a reseller can move you
• Your workload doesn't need specific hardware guarantees and you're optimizing for price

FAQ