InterServer VPS review: $3 monthly slices and the limits that matter

Conditional verdict: monthly scaling with boundaries you must accept

Start with the workload hubs for n8n, local AI or WordPress if you have not defined the application. Return here when the question is specifically InterServer.

InterServer earns a shortlist place when you want monthly billing, a small 2 GB starting point, linear resource scaling, and U.S. hosting. Its current Linux KVM VPS grid charges $3 per slice per month and reaches 64 GB RAM, 1,280 GB SSD and 64 TB monthly transfer at 32 slices. That is easier to reason about than a promotional ladder, and it makes a modest n8n or application-hosting experiment inexpensive to start.

The recommendation is conditional. The three VPS regions are all in the United States. The advertised 10 Gbps port is shared, not a measured or dedicated throughput guarantee. Ordinary VPS pages do not establish CUDA, VRAM or a GPU. There is no routine VPS trial/refund after service starts. A broad “managed support at eight slices” headline does not establish n8n, Ollama or WordPress application administration. And a manual backup control is not an automatic recovery plan.

Choose InterServer for the cash shape and scaling mechanics, not because $3 per month magically makes every workload suitable. If you need a low fixed annual special, compare RackNerd. If you prefer larger predefined tiers and different provider backup controls, compare Contabo.

The Linux slice arithmetic: simple, useful, still finite

The public grid makes each Linux KVM slice add 2 GB RAM, 40 GB SSD and 2 TB monthly transfer for $3 per month. vCPU increases every second slice: for s slices, ceil(s/2) vCPU. The current range is one through 32 slices.

*The last column is multiplication, not an annual contract, prepaid charge or discount. Source: Linux KVM VPS and VPS FAQ, checked 2026-11-05. Confirm currency, tax, region, stock and final total in your own account.

Monthly service auto-renews under the Terms of Service. Prepaid account credit can pay invoices but does not turn the monthly price into an annual discount. Adding slices changes the price by the difference between the old and new configurations. Save the invoice and exact selected template; a public table cannot establish your checkout.

Two comparisons expose the trade. InterServer’s 2 GB tier costs $36 across twelve retained months. RackNerd’s sampled 2 GB annual special is $35.99 upfront with 35 GB SSD and 5 TB transfer. The cash totals happen to be similar; the commitment, disk, transfer and renewal mechanics are not. InterServer’s 4 GB/$6-month tier has no meaningful annual-prepay requirement, while its larger tiers scale far beyond RackNerd’s sampled specials. Neither observation proves better performance.

Pick the product before counting slices

The shared Storage Hosting tiers illustrate why a full cart matters: 200 GB/$3, 1 TB/$5, 2 TB/$8, 4 TB/$10, 10 TB/$25 and 20 TB/$42 per month, with finite transfer allowances. DirectAdmin, SFTP/SCP and Rsync suit a file repository without a root daemon. Storage VPS instead runs from 1 TB/2 GB RAM/$3 through 32 TB/64 GB/$96 monthly. Compare protocol access, transfer, control and independence before assuming the higher raw capacity is the better backup purchase.

An alternate official VPS page also uses “Cloud Remote Backup Service” wording and says “no transfer fees on the storage service.” Treat that as a product-specific statement requiring checkout clarification, not a promise that VPS backups consume no transfer. The current backup-availability guide says Storage Hosting itself runs no backups; RAID redundancy is not a second independent recovery copy.

Storage VPS is not interchangeable with shared Storage Hosting. The former gives root and SATA disks; the latter provides an account-level storage service with no root. Neither, by itself, proves a second independent copy of your repository. If a backup source and destination are both cancelled or compromised together, your recovery design still fails.

Windows needs its own worksheet. The public grid does not explicitly establish the exact Windows Server edition/version or whether the displayed slice price includes licensing. Hyper-V recovery instructions differ. Do not buy it as a cheap Linux substitute.

The ordinary Linux, Storage and Windows pages also do not establish a GPU VPS. GPU/dedicated infrastructure is a separate sales path. More vCPU or RAM does not create CUDA or VRAM.

Three U.S. regions and a shared port

The current locations page lists the NYC metro area (Secaucus and Equinix NY4), Dallas and Los Angeles. That can be useful for U.S. audience latency or residency requirements. It does not serve a requirement for EU or Asian data residency.

Test reachability from your users, not from your office alone. Record packet loss and latency over time; a single ping is not an application benchmark. Exact template and SKU stock by region was not observable without signing into the order flow.

Each Linux slice has finite transfer and the grid advertises a shared 10 Gbps port. Storage and Windows pages make their own shared-port statements. Do not generalize those words to a dedicated 10 Gbps commitment, unlimited transfer or every other product. Public sources reviewed did not establish whether exhausting VPS transfer causes throttling, suspension or paid overage. They also did not establish whether backup traffic counts against your VPS allowance. Ask both questions with the exact SKU before scheduling large exports.

One public IP is advertised. Current FAQ and upgrade information list justified additional IPs at $3/IP/month; older $1 documentation is stale. IPv6 is advertised on the feature page, but allocation size, gateway and availability for your chosen template remain checkout questions. Outbound SMTP/port 25 is subject to review and acceptable-use rules; use a permitted transactional email provider rather than treating firewall changes as permission.

The 75-template catalog is a provisioning menu

The marketplace capture lists 75 cloud-init templates in nine groups: Docker, Coolify, NodeJS, databases, monitoring, HermesAgent, Nextcloud and Immich among them. Three distinctions matter for automation buyers:

• Docker template 7781 exists. No n8n or Ollama template was found in the reviewed catalog. The n8n guide therefore provides a custom Compose baseline rather than an imaginary one-click button.
• HermesAgent template 7900 is categorized under Monitoring & Security. Its page does not establish an AI runtime, model architecture or RAM floor. Identify the delivered application before operating it.
• OpenClaw template 7576 is described as a web/server control panel for web, database, DNS and email. Do not assume it is the similarly named AI agent from another provider’s catalog.

Template pages advertise cloud-init, random credentials, UFW, fail2ban and automatic security updates on relevant products. Those are vendor statements, not AiXVPS inspection results. Application configuration, DNS, TLS, backups, monitoring and updates remain operational work. Reinstall/rebuild replaces server contents; it is never a routine updater.

WordPress has two visibly different stories. The dedicated WordPress page advertises OpenLiteSpeed, FastCGI, LSPHP, LSCache, OPcache and tuning. Its order-link selection does not independently verify that this is the actual installed default. Marketplace WordPress template 7583 instead describes WP-CLI, Nginx, Redis, MariaDB and Certbot. Webuzo is included on its specific order path; DirectAdmin lists an additional $8/month VPS license; cPanel’s final license/add-on total is checkout-dependent. Record the actual stack before following any optimization guide.

Scenario fit: start from RAM, state and failure cost

These are editorial starting hypotheses, not InterServer benchmarks or application minima. Measure the real model, workflows, database and concurrency before committing to a larger monthly bill.

The AI app guide turns these hypotheses into measurement and stop conditions. The n8n guide starts at two slices/4 GB for its custom Postgres reference stack. Neither claims a model or production workload is guaranteed to fit.

Managed support: eight slices is a question, not an answer

The sales grid and FAQ advertise managed support at eight or more slices. Eight Linux slices cost $24/month, but that arithmetic does not make n8n or Ollama managed.

The linked managed-support guidelines are older and narrower: eight slices with a control panel, particular panel-support exclusions, and repair of certain non-functioning web, mail, DNS, SQL and FTP services. Third-party applications, plugins, themes, custom code, integrations and data remain customer responsibilities. The same page lists extended support at $160/hour in 15-minute increments and paid reinstall/password-reset services without a panel. Recheck those dated fees and current eligibility before relying on them.

The practical dividing line is server service versus application outcome. Restarting a failed supported SQL service is different from repairing an n8n workflow, choosing an Ollama model, resolving a WordPress plugin conflict, or proving that your encryption key and custom nodes survive a migration. If you cannot own those application tasks, price a competent administrator or managed application service rather than relying on a compute threshold.

Self-managing does not mean refusing support. It means keeping a private topology/version inventory, reproducible configuration, logs, restore evidence and a precise problem statement. Those artifacts make a paid support request more useful and reduce the chance that a well-intentioned reinstall destroys evidence. The setup runbook establishes that inventory on day zero.

Send a concrete ticket before buying for support: name the selected template, OS, panel, n8n/Postgres/Ollama or WordPress versions, and the exact repair or update task. Ask whether diagnosis, restoration, application upgrades and custom configuration are included or billable. Preserve the answer. The Terms describe a response-time goal, not a guaranteed resolution time; no support-speed claim is made here.

The SLA is network/power-focused with exclusions and capped case-by-case credits. It is not application availability, data-loss or recovery-time insurance.

Recovery, renewal and exit change the value calculation

The Terms make customers responsible for frequent backups and say VPS/dedicated services do not run backups by default. A 2025-12-19 backup guide describes manual KVM/Rapid Deploy snapshots and manual portal backups stored on cloud storage for 45 days for eligible services up to 1 TB. Hyper-V is excluded. Account quota, restore target and behavior above 1 TB remain unknown.

Snapshots may not survive an underlying array failure and can require restoring a full disk to extract one database. Keep an independent application-consistent export and encrypted off-host repository. The lifecycle guide provides that runbook.

VPS has no trial/money-back guarantee; the Terms exclude routine VPS refunds after service starts. Cancellation should be requested at least three days before automatic renewal, otherwise the Terms place cancellation after the next renewal. An older tutorial describes immediate shutdown within five minutes and possible reactivation with data for up to 30 days. That is not a retention guarantee or backup. Export and restore-test before choosing any destructive option.

InterServer versus RackNerd and Contabo

Choose RackNerd when a suitable annual special and its app path outweigh the upfront commitment. Choose Contabo when its capacity and recovery controls match the complete cart. Choose InterServer when monthly slice scaling, U.S. locations and manual KVM controls outweigh its narrower geography and support/transfer unknowns. There is no first-party performance benchmark here and no universal winner.

Procurement checklist and decision

Before payment, record: product/virtualization; slice count and exact monthly total; renewal date and auto-payment; tax/currency; region and IP allocation; OS/template/application versions; panel and license total; transfer exhaustion; support scope in writing; snapshot/backup eligibility; refund/cancellation terms; and independent backup destination.

Reject the purchase if the answer depends on a non-U.S. region, guaranteed application management, CUDA in ordinary KVM, routine refunds, automatic VPS backups, or dedicated 10 Gbps throughput. If those are not requirements, start with the smallest defensible hypothesis, keep the monthly invoice visible, and prove recovery before adding critical work.

Next: safe first-boot setup, AI app fit, complete n8n operations, or billing, backup and cancellation.

Primary sources

Official InterServer Linux VPS, FAQ, Storage VPS, Storage Hosting, Windows VPS, marketplace, locations, Terms, backup availability and managed-support guidelines. Public pages checked 2026-11-05. Editorial boundaries (including what we do not invent) live on About.