Best hosting for a Node.js API in 2026: the honest guide

A Node.js API is one of the friendlier workloads to host cheaply. Node's single-threaded event loop is lightweight per request, and a typical REST or GraphQL API with 100-500 requests per second can run comfortably on a $5-$20/mo VPS. The trap most Node developers fall into is buying too much hosting for the workload (paying for AWS Lambda when a $5 VPS would do) or too little (running a production API on a 512 MB Lightsail plan that OOMs under any real traffic).

This guide gives the honest picks. Prices as of 2026-09-22.

What a Node.js API actually needs

• RAM: at least 1 GB, 2 GB comfortable. Node's baseline heap usage is small (~50-150 MB), but any real API with Postgres client pooling, Redis, and moderate concurrency needs headroom for spikes. 2 GB is the tier where nothing surprises you.
• vCPU: 1-2 for early-stage, 2-4 for scaled. Node is single-threaded; a single API process pins one core. If you cluster (using pm2 or Node's built-in cluster module), you can use multiple cores.
• NVMe storage. Only matters for logs and any local caching; the database is usually external.
• Fast network for external service calls. Most APIs spend more time waiting on other services than computing anything.
• Managed database or Redis, optionally. Depends on your data layer.

Recommended picks

Early-stage API ($5-8/mo)

Hetzner CX22 at $4.60/mo — 2 vCPU, 4 GB RAM, 40 GB NVMe, 20 TB transfer.

Massive headroom for a launch-tier API. Deploy with Docker or PM2. Add a systemd service and Let's Encrypt SSL. Postgres and Redis run on the same VPS at this size without competition. This is the correct default for a Node.js API you're not sure will succeed.

Alternative: Vultr High Frequency 1 GB at $6/mo if you specifically need US or Asia regions Hetzner doesn't cover, and your API is CPU-bound (fast-executing sync request handlers where single-thread speed matters).

Steady API traffic ($10-20/mo)

Vultr High Frequency 2 GB at $12/mo (approx; not in our dataset, verify at vultr.com/pricing). Real gains for CPU-bound Node handlers with 3.7 GHz+ single-thread performance.

Or Hetzner CX32 at $7.60/mo for maximum resources per dollar if regions aren't a constraint. 4 vCPU / 8 GB is more than most Node APIs need at this stage.

Consider adding DigitalOcean Managed Redis at $15/mo or a managed Postgres if your data layer is starting to feel operational at scale.

Scaled API ($30-100/mo)

DigitalOcean Basic 8 GB at $48/mo — 4 vCPU, 8 GB RAM, 160 GB storage. Cluster Node across all 4 cores using PM2. Externalize Postgres to a DO Managed Postgres at $15-60/mo based on your DB size.

Or Vultr Regular 8 GB at $40/mo for the same shape with slightly more storage transfer.

At this tier, the single-thread performance difference between DO and Vultr HF matters less because you're clustered anyway.

Alternative architectures worth considering

Serverless (Vercel, Netlify Functions, AWS Lambda): if your API is genuinely bursty (0-2000 req/s peak with long idle periods), serverless can be cheaper than a VPS. Not our category so we don't compare pricing here, but worth mentioning. Rule of thumb: below ~200k requests/day, a $5 VPS is usually cheaper than Lambda; above 5M requests/day, Lambda is often cheaper due to auto-scaling economics.

Managed container platforms (Fly.io, Railway, Render): convenient developer experience, prices in the $5-40/mo range for equivalent Node workloads. StackRates doesn't publish live pricing on these because their model doesn't fit our dataset shape. Worth checking directly if the deploy-from-git story matters to you.

Kubernetes: overkill for a single Node API. Legitimate at multi-service SaaS scale but you should not adopt K8s just to run one API.

Deployment patterns per host

Hetzner / Vultr / Linode / Contabo (unmanaged)

```

1. Provision Ubuntu 24.04.

2. Install Node.js via nvm or nodesource.

3. Deploy your app; run under PM2 or a systemd service.

4. Front with Nginx as a reverse proxy for TLS termination.

5. Certbot for Let's Encrypt SSL.

6. UFW for basic firewalling.

7. Cron a nightly pg_dump to S3-compatible object storage.

```

Setup time: 2-4 hours the first time. Repeatable in 15 minutes after that with a bash script or Ansible playbook.

DigitalOcean App Platform, Vultr Cloud Compute + Managed Database

Cleaner git-push deploy story. Small setup premium, less operational maintenance.

Things that fail Node APIs on hosting

• Running on 512 MB Lightsail. Node's heap will grow to 400 MB with any moderate work, OOM killer will restart you at 3 AM.
• Not clustering. A 4 vCPU box running one Node process uses one core. PM2 cluster mode or Node's built-in cluster module is free performance.
• No connection pooling on Postgres. Every request opens a fresh Postgres connection → connection storm → DB unavailable. Use pg-pool or a pool at your DB level.
• Cron running WP-Cron equivalents on the request path. Move background work to bullmq + Redis with a separate worker process.

Which one you should actually pick

• Prototyping / launched last week → Hetzner CX22 at $4.60/mo. Overkill in a good way.
• Real traffic, CPU-heavy handlers → Vultr HF 2 GB at ~$12/mo.
• Real traffic, cost-first → Hetzner CX32 at $7.60/mo.
• Scale, cluster across cores, external DB → DigitalOcean Basic 8 GB at $48/mo + DO Managed Postgres.
• Bursty, sub-second execution → serverless (out of our category).
• You value deploy-from-git DX above cost → Fly.io or Railway (out of our category).

Related guides

• Best hosting for a Django app — the Python parallel.
• Best hosting for a Laravel app — the PHP parallel.
• Best hosting for a small SaaS MVP — the multi-service architecture.
• How to host a Next.js app on a $5 VPS — the SSR-adjacent version.
• How to calculate cloud hosting cost — the arithmetic that makes the above prices real.

Prices as of 2026-09-22.