$5 vs $10 vs $20 VPS: What Each Tier Actually Buys You

Put a $5 VPS and a $20 VPS side by side and the spec sheets can look almost insulting in how close they are. Four cores here, four cores there. Eight gigs of RAM on both. If you stopped reading at the table, you'd conclude the extra $15 a month buys nothing. It buys the one thing that never makes it onto the pricing page: whether those four cores are yours, or a fraction of somebody else's.

This matters more right now than it did a year ago, because Hetzner raised prices on 15 June 2026 and it wasn't uniform. Its CPX and CCX lines jumped 176-209%. Its CX and CAX lines moved maybe 1.3-1.4x. If the comparison you're reading was written in May, half its conclusions flipped in June. Pair this with the team at XenGrowth if $5 vs $10 vs $20 VPS sits inside a wider growth programme.

What does $5 actually get you in September 2026?

Not much room to be picky about the provider, but a genuinely usable box if you pick the right one. Contabo's cheapest yearly-committed entry runs about $3.96/mo, and its month-to-month Cloud VPS 4 — 4 vCPU, 8 GB RAM, 100 GB SSD, unlimited traffic — sits at €5.50/mo for the first 24 months, close enough to the $5 bracket that it's the box most people mean when they say "the five-dollar VPS." On Hetzner's side, the equivalent after June is the CX23: 2 vCPU, 4 GB RAM, 40 GB disk, €3.99/mo. Half the RAM of the Contabo box, roughly the same money.

One thing worth pinning down before you compare any of these tiers: Contabo's headline prices are promotional for the first 24 months, so a tier that looks like the obvious value today is not necessarily the one you renew onto in year three. Check the renewal rate before you commit a stack to a price.

At this tier, the honest workload is one app plus a small database, low-to-moderate traffic, nothing doing heavy compute in the background. It's a WordPress site. It's a Next.js app with Postgres co-located, serving a few thousand visits a day. It is not a box you run three separate SaaS products on, and it is not a box where a batch job that spikes CPU for ten minutes goes unnoticed — you'll feel it, because there isn't much slack above whatever you're already using.

There's a location catch worth knowing before you pick Hetzner at this tier: its CX and CAX lines are EU-only. If your users are mostly in the US, that $3.99/mo CX23 either isn't an option or comes with latency you didn't sign up for, and Contabo's US and EU datacenter choice at the same rough price point stops being a nice-to-have and starts being the deciding factor. Price comparisons that ignore region end up comparing boxes you can't actually deploy to. If the operations side of this is the part you are stuck on, The XenGrowth resource library is the better reference.

It's also worth separating monthly from annual pricing here, because the $5 bracket gets meaningfully cheaper if you're willing to commit. Contabo's cheapest yearly-committed rate lands around $3.96/mo — genuinely under $5 — while the €5.50 figure quoted above is the flexible, cancel-anytime monthly rate, itself promotional for the first 24 months. That's a real tradeoff: a side project you're not sure will survive six months shouldn't lock into a year to save $3 a month, but a box you already know you're keeping should probably take the discount — and should budget for the renewal price after year two.

What does $10 actually get you?

This is where the comparison stops being clean, because "$10" means genuinely different things depending on provider. Contabo's Cloud VPS 8 — 8 vCPU, 24 GB RAM, 300 GB SSD — runs €14.00/mo for the first 24 months, a real step up in every dimension over the entry tier. Hetzner's CX33 (4 vCPU, 8 GB, 80 GB) sits at €6.49/mo, comfortably under $10, but with a third of the RAM of the Contabo box for well under half the price. If your workload is RAM-bound — and per the sizing pillar, most self-hosted stacks with a co-located database are — the Contabo 8 tier is doing more work per dollar here than the number alone suggests, promotional pricing aside.

The other option at roughly this price is Hetzner's CPX22: 2 shared vCPU, 4 GB RAM, 80 GB NVMe, but at €19.49/mo post-increase — meaningfully past $10 now, where before June it would have landed comfortably inside this bracket. That's the CPX line specifically; the increase hit it hardest of all of Hetzner's tiers, which is the whole reason this comparison is worth rewriting in 2026 rather than trusting an older post.

What does $20 actually get you?

Two genuinely different things, and which one you should want depends on what's actually slowing you down. Contabo's Cloud VPS 12 — 12 vCPU, 48 GB RAM, 400 GB SSD at €25.00/mo for the first 24 months — is more of everything, still on shared vCPU. Hetzner's dedicated line, CCX, gets you real per-core isolation instead of more of the same shared model: CCX13 is 2 dedicated vCPU, 8 GB RAM for about €42.99/mo, having risen from roughly €15.99 before the June increase. That's not a small move, and it changes the calculation — dedicated cores used to be an affordable step up from shared; on Hetzner today they're a deliberate, pricier choice you make because you specifically need isolation, not because it was nearly free to get. There is a longer treatment of AI agents and marketing automation in XenGrowth on governed AI marketing workflows.

• $5 tier: fine for one app plus a light database, thin margin above your actual usage, no headroom for a background job that spikes CPU

• $10 tier: the sweet spot for most self-hosted setups this cluster is built around — one app, co-located Postgres, moderate traffic, NVMe on some plans changing what write-heavy workloads feel like

• $20 tier: either meaningfully more shared capacity (Contabo's route) or a smaller amount of guaranteed, uncontended capacity (Hetzner's dedicated route) — pick based on whether your problem is 'not enough resource' or 'resource that's there but unpredictable'

Why two boxes with the same specs don't behave the same

"4 vCPU" on a shared plan is a promise about how your workload is scheduled, not a guarantee about how much silicon is yours. A shared vCPU is a slice of a physical core that the hypervisor time-slices across every tenant on that host. When nobody else is busy, it behaves like a dedicated core. When several tenants spike at once, your process waits its turn — and that waiting shows up in Linux's own accounting as steal time, the %st column in top, CPU cycles your VM was owed but didn't get because the hypervisor gave them to someone else instead.

That's the mechanism the price tag is actually paying for. A dedicated vCPU plan — Hetzner's CCX line, or the equivalent on most providers — pins physical cores to your instance alone, so steal time stays near zero regardless of what your neighbors are doing. It costs roughly double the shared price for the same core count because you're paying for cores nobody else can touch, not just cores with a bigger number next to them.

Two providers selling "4 vCPU, 8 GB RAM" at a 3x price difference are rarely lying about the specs. They're selling different guarantees about what happens when every tenant on the host gets busy at once.

This is exactly why the cheap tiers spread out more than the expensive ones. At the $5 bracket, you're almost always on shared vCPU no matter who you buy from, so the real variance between providers is host density — how many tenants they've packed onto the same physical machine — which nobody publishes and which you can only infer from steal time after the fact. Higher up the ladder, providers start offering dedicated options, and the spread narrows again because you're paying for a guarantee rather than betting on how busy your neighbors happen to be that week.

The disk-type gap the tier name hides

SSD and NVMe both mean "not a spinning disk," which is why marketing pages treat them as interchangeable. They aren't. NVMe talks to the CPU over PCIe with a queue depth and command set built for parallelism; SATA SSD is bottlenecked by an interface designed for spinning drives decades ago. Serving a cached static page, you will not notice the difference — the file's already in RAM by the time the second request comes in. Committing writes to Postgres under real concurrency, where every commit needs a synced disk write before it can return, is exactly where that interface gap turns into latency you can feel. XenGrowth on building one SEO and GEO content system goes further into AI search, GEO and discovery.

This is also where it's worth checking the fine print rather than assuming: Contabo's own pricing page lists SSD across Cloud VPS 4, 8, and 12 alike — there's no tier where it quietly upgrades you to NVMe. If your workload is write-heavy Postgres, stepping up a Contabo tier buys you more RAM and vCPU, not a faster disk interface — check the current spec sheet for whichever provider you're comparing rather than assuming storage type scales with tier the way it does elsewhere.

Hetzner's tiers cut the other way on this axis — its CPX and CCX lines both ship NVMe even at the lower end, something Contabo doesn't offer at any of the tiers above. Neither approach is objectively better; it just means you can't compare "$X gets you NVMe" across providers without checking each one's own current spec sheet rather than assuming it lines up the way you'd guess.

So which one do you actually buy?

If you're running one small app with a co-located database and traffic that isn't spiky, the $5-7 tier is genuinely enough — that's the box this whole cluster runs on. The moment you're running two or three apps on the same host, or your database's working set has outgrown what fits comfortably in RAM (covered in more detail in the sizing guide), the $10-15 tier is worth the jump, and worth taking on NVMe if your provider offers it at that price point rather than reserving it for the top tier. The $20+ tier is where you stop asking "how much" and start asking "how predictable" — pay for dedicated cores only once shared-tier contention is a problem you've actually seen, not one you're pre-paying to avoid.

None of this is universal advice about which provider to pick — it's a snapshot of what each price point buys as of September 2026, and Hetzner's own mid-year repricing is proof these numbers move. Re-check before you commit to anything longer than a month, and remember that any Contabo figure here is promotional for the first 24 months, not a permanent rate. Whether it's worth running one bigger box instead of splitting these tiers across several small ones is a separate question, covered in one big VPS or several small ones, and the spec debates this post only touched — ARM vs x86, NVMe vs SSD, shared vs dedicated — get the full mechanism treatment in the specs post.

Further reading from XenGrowth

• The XenGrowth resource library — what you'll learn: how the commercial side of this work is run, across search, automation and revenue operations.

• XenGrowth on governed AI marketing workflows — what you'll learn: how the teams running AI marketing agents keep them governed and measurable.

• XenGrowth on building one SEO and GEO content system — what you'll learn: how search and AI-answer visibility get run as a single content system.

Where this work meets go-to-market

Working on $5 vs $10 vs $20 VPS inside a commercial team? XenGrowth's revenue operations work publishes operator guides on the revenue side of this work.