Unmetered Bandwidth Is Not Unlimited. It Is a Different Meter.

Every "unmetered bandwidth" badge on a VPS or cloud plan page means exactly one thing: the provider has stopped counting your bytes. It does not mean the bytes are unlimited, and it does not mean the pipe is wide. The limit did not disappear — it moved. It left the invoice, where you could read it, and relocated to the port configuration, the shared uplink behind your instance, and an acceptable use policy clause, where most buyers never look.

That is because unmetered is a billing dimension, not a capacity claim. Four different products are hiding behind the two words. A metered plan sells you a volume of bytes per month. An unmetered plan sells you a pipe diameter and stops weighing what flows through it. An unmetered plan with a stated speed cap sells you a narrower pipe. And "unlimited bandwidth" — the phrase that dominates the budget end of the market — usually sells you nothing measurable at all, because the operative limit is a fair-use clause that activates only when a human at the provider decides you are "excessive." Comparing these as if they were one product is the single most common bandwidth shopping mistake.

This article separates the four models using live November 2026 prices, converts port speeds into the monthly transfer ceilings they actually imply, exposes the three places an unmetered limit hides, walks through a fifteen-minute procedure to measure what you really bought, and closes with a six-question checklist to run before prepaying any plan that advertises its traffic terms.

The Direct Answer: What Unmetered Bandwidth Means

Unmetered bandwidth means the provider bills you for the network port — its speed, in Mbps or Gbps — rather than for the gigabytes you transfer, so your ceiling is the port capacity and whatever uplink backs it, not a byte quota. As of November 2026, every unmetered plan on the VPS market is a capped pipe: what varies is whether the cap is the port speed itself, the shared uplink behind it, or a policy clause. The bytes are not counted; they are absolutely still limited.

The practical difference from a metered plan is what the invoice does at the end of the month. On a metered plan, exceed your quota and a per-gigabyte overage line appears — covered with real rates later in this article. On an unmetered plan, there is no overage line, ever. Instead, performance is the enforcement mechanism: saturate the port, or trip the oversubscription on the shared uplink, or attract the attention of the acceptable use policy, and your throughput degrades or your service is suspended. The meter did not go away. It turned into a speedometer.

One more definitional edge matters in November 2026: inbound versus outbound. Unmetered and metered clauses usually apply to outbound traffic only, and many providers do not count inbound at all. A backup destination and a video streaming origin have opposite traffic shapes, and only one of them cares about the outbound meter. Always confirm which direction the plan's traffic terms govern before comparing prices.

Four Traffic Models Are Hiding Behind Two Words

The market's traffic vocabulary collapses four structurally different products into the two words "unmetered" and "unlimited," and the conclusion to draw from the table below is that the wording on the plan card tells you which product you are holding only if you read it as a contract, not as a feature. The model is decided by what you pay for and where the limit physically lives.

The first model is honest arithmetic: you can compute your worst-case bill to the cent. The second is the classic dedicated-server product transplanted into VPS form — it works when the provider sizes the uplink realistically. The third is the most common model at the budget end of the VPS market as of November 2026, and it is a perfectly good product when the cap is printed on the card; it becomes a trap only when the cap is omitted and discovered by the customer as mysterious slowness. The fourth is a marketing term — industry guidance is blunt that "unlimited bandwidth" is primarily a marketing term and that such plans almost always include a fair-use policy that kicks in when your consumption is deemed excessive. The limit exists; it is simply written in prose instead of megabits.

The Math: A Port Speed Is a Monthly Volume in Disguise

Every port speed converts directly into a hard monthly transfer ceiling, and the conversion is one multiplication: 1 Gbps is about 324 TB per month of theoretical maximum, 100 Mbps is about 32 TB, 30 Mbps is about 9.7 TB, and 10 Mbps is about 3.2 TB. As of November 2026, this arithmetic settles most "which plan transfers more" arguments before they start.

The conversion is exact at 100 percent utilization: 125 megabytes per second at 1 Gbps, times 2,592,000 seconds in a 30-day month. Nobody sustains 100 percent. Realistic sustained utilization on a busy shared network lands near 60 to 70 percent of the theoretical figure, and single-connection throughput is lower still once protocol overhead and round-trip time are included.

The table's most important row is the last one. A "10 Mbps unmetered" server transfers less data in a month than a metered plan with 20 TB included — the metered plan is, in volume terms, the more generous product. The unmetered advantage is predictability of cost, never automatically a larger allowance. When a plan card says "unmetered," the first question is never "how much traffic is that?" It is "at what speed?" — because the speed is the volume.

The Three Places an Unmetered Limit Actually Hides

On an unmetered plan, the limit lives in exactly three places: the port speed cap, the oversubscribed shared uplink behind it, and the acceptable use policy — listed here in decreasing order of visibility, so the ones most likely to surprise you are at the end. Each one enforces a ceiling without ever sending you an overage invoice.

1. The port speed cap. The cleanest limit, and the only honest one, is a stated cap: "unmetered at 30 Mbps" tells you the ceiling in writing, and the arithmetic in the previous section converts it to roughly 9.7 TB per month. The problem is not the cap; it is the plans that omit it. A card that says "unmetered" with no speed figure is either quoting the physical port of the host machine — which you share — or hiding the cap deliberately. If the plan card does not state a speed, assume a capped port and confirm before paying.

2. The shared uplink and its oversubscription ratio. Your instance sits behind a physical uplink shared with every other tenant on the host and often the whole rack. The distinction between dedicated and shared bandwidth is structural: dedicated bandwidth is a fixed, exclusive pipe reserved for your instance at the virtualization or switch level, while shared bandwidth pools multiple tenants into one total pipe — the standard example being a hundred VPS instances sharing a 1 Gbps pool. Shared pools run at meaningful oversubscription ratios, and the industry's explanation of shared bandwidth is explicit that high overcommit is the normal commercial design, not an aberration. On a shared uplink, the constraint is not your port speed but your neighbors' behavior at 8 p.m.

3. The acceptable use policy clause. The final limit is written in prose. Fair-use policies on "unlimited" plans typically allow the provider to throttle, suspend, or require an upgrade when your consumption is "deemed excessive" — a standard that is defined by the provider, in hindsight, after you have already built on the service. Industry guidance on unlimited bandwidth is unambiguous: no provider can offer truly limitless resources, and such plans almost always carry this clause. The correct way to read "unlimited bandwidth" on a plan card in November 2026 is: "metered by the provider's discretion, at no published rate."

Metered vs Unmetered: Which Model Fits Your Workload

The cheaper model depends entirely on traffic shape: steady high-volume streaming workloads are cheapest on an unmetered capped port, bursty low-volume workloads are cheapest on a metered plan with a fast port, and directionally asymmetric workloads — backups, mirrors, ingest pipelines — are decided by which direction the plan counts. Match the model to the shape, not to the badge.

The one workload that must never buy on the unmetered badge alone is the latency-sensitive one. A game server or an SSH-heavy operations box moves a few megabytes per hour; its performance is decided by round-trip time and route quality, not by how many terabytes the plan permits. For those workloads, the traffic model is nearly irrelevant and the location is everything — the same trade-off explained in detail in how a Hong Kong VPS position trades raw bandwidth for latency and route quality.

The Overage Rates That Make Meters Scary (November 2026)

The reason unmetered plans exist at all is that metered overage rates are genuinely high: as of November 2026, one extra terabyte of outbound traffic costs about $90 on AWS Lightsail, about $10 on DigitalOcean, and about 1 euro on Hetzner — a ninefold spread between the priciest and cheapest mainstream providers for the identical unit. The meter's danger is not the quota; it is the rate that applies after it.

Three structural observations follow from the rates. First, Hetzner's 20 TB included with a 1 euro per-TB overage behaves as effectively unmetered for any small workload — the buffer is nearly impossible to exhaust, and the penalty for doing so is a coffee. Second, the expensive end of the market is not the dedicated-server world but the hyperscalers: the same terabyte that costs 1 euro in Falkenstein costs 90 dollars on a Lightsail instance in Virginia, and Lightsail's own documentation notes that six of its regions — Mumbai, Sydney, Jakarta, Malaysia, Hong Kong, and Sao Paulo — halve the included transfer allowance, doubling effective overage exposure for Asia-Pacific deployments. Third, on providers that meter, the overage rate matters more than the quota: a 1 TB quota at $0.09 per GB is a worse bill-risk product than a 500 GB quota at $0.01 per GB.

How to Measure What You Actually Bought (Fifteen Minutes)

You can empirically determine your real port speed, your real sustained ceiling, and your real monthly volume in about fifteen minutes with three tools — iperf3, vnstat, and the /proc/net/dev counter — and the procedure below works on any Linux VPS regardless of what the plan card claims.

1. Read the plan card for the stated speed first. Write down the exact traffic wording: port speed, cap if stated, quota if metered, and which directions are counted. This is your hypothesis; the next steps test it.
2. Run iperf3 against a known endpoint. Install iperf3, rent or borrow a second server in a different network, and test in both directions for 30 seconds each. This measures your clean single-flow throughput, which is the closest proxy to your real port cap.
3. Do not trust consumer speed tests on a server. Specialist providers warn against speedtest-type tools for VPS bandwidth verification — their test servers frequently have less than 1 Gbps of available capacity, so the tool reports the test server's ceiling, not yours. Use iperf3 to an endpoint you control.
4. Install vnstat and let it accumulate. vnstat counts interface traffic passively and stores monthly totals. After a full billing cycle, compare its outbound figure against your quota — this is the number that predicts your overage risk, and comparing it to the theoretical ceiling from the port-speed math tells you which limit binds.
5. Retest at the peak hour. One iperf3 run at 3 a.m. proves the port; only a run at your region's evening peak exposes the oversubscription on the shared uplink. A 60 percent drop between the two runs is a shared-pool signature, not a fault.

Treat the results the way you would treat any capacity audit. The same discipline applies to every resource on the invoice — the same fifteen-minute skepticism that catches a capped port also catches an oversold vCPU, which is where a guide like measuring the CPU capacity a server actually delivers earns its keep: verify what was sold, not what was advertised.

Why Providers Price It This Way

Providers sell unmetered plans because their own upstream costs are priced per megabit per second, not per byte: IP transit and cloud interconnect are billed on the 95th-percentile model — the provider's monthly bill is computed from its peak sustained traffic after discarding the top five percent of samples — a model used, for example, by Tencent Cloud's cloud-connect network billing as of July 2026 and by the upstream transit market generally. When your upstream bill is computed from your peak width, selling customers a fixed width and unlimited volume is not generosity. It is passing the exact cost structure through.

The oversubscription economics follow. A provider with a 10 Gbps rack uplink can sell one hundred "1 Gbps unmetered" instances profitably only because the statistical average tenant uses a small fraction of the port at any moment. The product is a bet on statistics — the same shared-resource logic that lets a NAT VPS provider serve dozens of machines behind a single IPv4 address, the economics of which are laid out in this breakdown of how NAT VPS plans turn resource sharing into a price. Sharing is why a $5 unmetered gigabit exists at all; it is also why it cannot possibly be dedicated.

The proof is at the dedicated end of the market. Specialist providers that sell unmetered ports sized honestly price them accordingly: unmetered VPS tiers with real bandwidth guarantees list from about $74 per month as of November 2026, and unmetered dedicated servers trade in the hundreds per month. A dedicated 1 Gbps port costs real money because it reserves real upstream capacity. Any unmetered gigabit offer at a fifth of that price is shared by arithmetic necessity, and there is nothing wrong with that — as long as the buyer knows the uplink is a pool and not a reservation.

Six Questions to Ask Before You Prepay

Before paying a year in advance on any plan whose traffic terms matter to you, six questions — answerable from the plan card, the acceptable use policy, and one support ticket — separate the four traffic models and expose every hidden cap. If a provider cannot or will not answer any of them, treat the silence as the answer.

1. "What is the port speed and any cap on it?" A stated number means a capped-port product you can price with the arithmetic above. No stated number means the cap is somewhere you have not been shown.
2. "Is the bandwidth dedicated or shared, and at what oversubscription ratio?" Honest providers answer this directly. A shared 1 Gbps pool at high overcommit behaves nothing like a dedicated 1 Gbps port at peak hours.
3. "Which AUP clause governs traffic, and what number triggers it?" If the plan says unlimited, the fair-use clause is the real quota. Ask for the threshold in writing; "we'll know it when we see it" is a meter with an unpublished rate.
4. "If metered: what is the overage rate per GB, and is it billed per GB or per TB?" The per-unit granularity matters — a $0.09 per-GB rate is a $90 per-TB rate, and the quota you exhaust at 11 p.m. on the last day of the cycle bills at the small number.
5. "Is inbound traffic counted, throttled, or free?" For backup and ingest workloads this single answer decides the entire comparison; many providers do not meter inbound at all, which makes an inbound-heavy workload effectively unmetered on any plan.
6. "Does the port speed or quota change at renewal or on promotion tiers?" Promotional entry prices sometimes carry different traffic terms than the renewal tier — verify the quota and cap on the renewal plan, not just the one you are buying today.

The November 2026 Market, Labeled Honestly

Applying the four-model framework to real plan cards as of November 2026: Hetzner sells a high-quota metered product whose 20 TB allowance and 1 euro per-TB overage behave as effectively unmetered for small workloads; DigitalOcean and Vultr sell classic metered plans with fast ports and per-GB overage; and the capped-port unmetered model dominates the budget VPS segment, where SurferCloud's ULightHost line is a useful worked example of the hybrid — its plan cards list "30M bandwidth" with 1 to 4 TB of outbound traffic by tier and inbound traffic uncounted, which reads as: a capped 30 Mbps port (about 9.7 TB theoretical monthly ceiling) plus a metered outbound quota below it, with the effective limit being whichever of the two binds first. That is a metered plan whose overage mechanism is the tier ladder rather than a per-GB rate — fully predictable, and fully legible once you know which model you are reading.

Reading plan cards this way turns the entire bandwidth vocabulary transparent. "1 Gbps unmetered" on a $5 VPS is a shared-pool product whose cap is your neighbors' evening usage. "Unmetered at 30 Mbps" is a volume product of about 9.7 TB per month wearing a speed costume. "Unlimited bandwidth" behind a fair-use clause is a discretionary meter. And "1 TB outbound, inbound free" is a directional product whose cost profile depends entirely on which way your bytes flow. None of these is a scam; each is a different product. The entry tier of that ULightHost line, with its 1 TB of outbound transfer on a 30 Mbps port, starts at $2.90 per month for new users as of November 2026 — current tiers and quotas are listed on the promo page — and is correctly compared against other metered 1 TB plans, not against unmetered gigabit offers, because it is one.

FAQ

Is unmetered bandwidth the same as unlimited bandwidth?

No. Unmetered means the provider does not count your transferred bytes for billing; your limit is the port speed and the shared uplink. Unlimited is a marketing claim with no measurable definition, typically enforced through a fair-use policy clause that lets the provider throttle or suspend accounts whose usage is deemed excessive. Unmetered has a real, findable ceiling; unlimited has a policy.

Can I really use as much traffic as I want on an unmetered VPS?

Up to the ceiling your port imposes, yes — and that ceiling is arithmetic, not generosity. A 1 Gbps unmetered port can theoretically move about 324 TB per month; a 30 Mbps cap can move about 9.7 TB; a 10 Mbps cap about 3.2 TB, which is less than many metered 20 TB plans include. You will never see an overage invoice, but you can absolutely hit a wall.

Why is my unmetered VPS slow even though traffic is free?

Three likely causes, in order: your port is speed-capped well below what you assumed, your shared uplink is oversubscribed and congested at peak hours, or your workload is latency-bound and bandwidth was never the constraint. Test with iperf3 at an off-peak hour and again at the evening peak — a large gap between the two results is the shared-pool signature.

How much traffic can 1 Gbps unmetered actually transfer per month?

About 324 TB per month in pure theory (125 MB/s sustained for all 2,592,000 seconds of a 30-day month), and roughly 210 TB at a realistic 65 percent sustained utilization. Single connections will see less, because per-connection throughput also depends on round-trip time and protocol overhead.

Is metered or unmetered better for a small website?

Metered with a fast port, almost always. A small site's traffic is bursty and low-volume — it needs a fast port for snappy page loads and a generous quota it will never touch, not a flat-rate pipe it cannot fill. Unmetered-at-a-cap products are built for steady high-volume streaming and mirroring, not for bursty request/response workloads.

Do I get charged for inbound traffic on a metered plan?

Usually not — most providers meter outbound traffic only and do not count inbound at all, which is why backup destinations and ingest pipelines can run "free" bandwidth on any plan. Confirm the direction in the plan terms, because the handful of providers that count both directions change the economics of asymmetric workloads completely.

What is a fair use policy on unlimited plans?

It is the clause that defines the real limit on an "unlimited" plan. Typically it permits the provider to slow, suspend, or upsell you once your consumption is judged excessive — a standard defined by the provider, after the fact, with no published number. Industry guidance is blunt that no provider offers truly limitless resources and that such plans almost always carry this clause. Read it as a meter with an unpublished rate.