Residential Proxy Pricing Explained: What You Pay For Per GB

Residential proxy price lists are built to resist comparison. One provider quotes a rate that only exists at a terabyte of monthly commitment. Another quotes a monthly subscription with an included allowance and a punitive overage rate behind it. A third sells unlimited bandwidth and charges by the thread. All three can honestly claim to be the cheapest, because they are all describing different products.

The underlying model is simple enough. Everything confusing sits in the terms wrapped around it.

Why bandwidth became the billing unit

A residential provider does not own its exit IPs. It rents access to households, usually by paying an app publisher for the bandwidth their users contribute, and then pays transit on top of that. Its own cost is measured in bytes moved. Requests are close to free; gigabytes are not. That cost structure is passed straight through to you.

Which means two teams issuing an identical number of requests can be two orders of magnitude apart on spend, purely because of what they download.

That is a 640-fold spread between the top and bottom rows of the same price list. Before negotiating a rate, find out which row you are on. If you are on the bottom row and do not need to be, fixing that beats any discount you could negotiate, and the techniques are collected in reduce proxy bandwidth costs.

Static ISP and datacenter proxies are usually sold per IP per month with bandwidth included, which is a completely different shape of bill and often far cheaper for high-volume work against tolerant targets. The comparison is in residential vs datacenter proxies.

Tiers make headline prices incomparable

The number on the pricing page is nearly always the floor of the deepest tier. "From $2 per GB" means $2 per GB at a terabyte a month on an annual commitment. At 40 GB a month you might pay three or four times that, and the provider has not misled anyone.

Discounts run along two axes at once: how much you buy per month, and how long you commit. Providers place their break points in different places, so the ranking at 1 TB tells you nothing about the ranking at 50 GB. A provider that is mid-pack on the marketing page can be the cheapest option at your volume, and the reverse happens just as often.

The contract terms that decide what you actually pay

Three clauses do most of the damage, and none of them appear in the big number on the pricing page.

Commitment and rollover

A monthly commitment is a floor on your spend, not a budget. Unused gigabytes usually vanish at the end of the billing period. Where rollover exists it is often capped at one month, restricted to accounts that renew at the same tier or higher, and forfeited entirely on a downgrade. Ask for the rollover rule in writing before signing anything longer than a month, because sales answers and billing behaviour diverge here more than anywhere else.

Expiring prepaid bandwidth

Pay-as-you-go packs frequently carry an expiry: 30, 60 or 90 days are all common, and some providers offer packs that never expire at a higher rate. This clause is worth real money. Buy a 100 GB pack at $3.00 per GB, spend $300, then use 20 GB before it expires, and you paid $15.00 per gigabyte. The non-expiring pack at $4.00 per GB would have been a bargain by comparison.

Minimum top-ups and auto-recharge

Many accounts carry a minimum purchase, and some carry a minimum monthly spend that applies whether you send traffic or not. Auto-recharge is the related trap: a threshold that triggers another pack the moment your balance drops, which turns a runaway retry loop into a four-figure bill overnight. Set the recharge cap deliberately and alert on it from your own side.

What an unlimited plan actually limits

Unlimited bandwidth plans are real, and the constraint moves rather than disappears. What you buy is concurrency: a fixed number of threads, ports or simultaneous sessions. Bandwidth is then whatever those threads can pull.

GB per day = threads / avg_seconds_per_request * avg_KB_per_request * 86400 / 1e6

# 50 threads, 1.5s average round trip, 400 KB responses
#   50 / 1.5 = 33.3 requests per second
#   33.3 * 400 KB = 13,333 KB/s  ->  about 1,150 GB per day

Run that arithmetic with the thread count you are being offered and the latency you actually measure, and the plan either looks generous or looks like a metered plan in disguise. Residential round trips of one to two seconds are normal, so latency has as much influence on the ceiling as the thread count does.

My honest read: unlimited plans are good value for sustained, latency-tolerant, bandwidth-heavy work, and poor value for bursty jobs that need hundreds of parallel connections for twenty minutes a day. Check three things in the terms before believing the word. Whether the pool behind an unlimited plan is rotating residential or static ISP, since these are frequently sold under the same label and behave differently, as described in ISP proxies explained. Whether a per-thread speed cap exists. Whether a fair-use clause lets the provider throttle you at their discretion.

Sub-users, seats and overage rates

Sub-user accounts split one contract into separately authenticated credentials with their own bandwidth caps. Agencies need them to bill clients. Engineering teams need them to stop a runaway job in staging from consuming the production allowance. Providers treat them very differently: included and unlimited, capped at a small number, priced per seat, or gated behind an API that only exists on higher tiers.

Overage is the other place where a good rate turns bad. Exceeding an allowance can bill at your tier rate, at the pay-as-you-go list rate, or at a penalty rate above both. Some providers stop serving traffic instead. Neither behaviour is wrong, and you need to know which one you bought, because one costs money and the other costs a night of on-call. Cap the spend, alert at 80 percent of the allowance, and make the crawler degrade gracefully when the cap is hit.

Effective cost: price per GB divided by success rate

A blocked request still moves bytes. The TLS handshake happens, the challenge page downloads, the retry runs, and all of it crosses the meter. So the price you are quoted is a lower bound on the price you pay per useful result.

Beta advertises a rate 38 percent below Alpha and delivers results at a higher price. The assumption in that table is worth stating plainly: it treats a failed attempt as costing the same bandwidth as a successful one. In reality a challenge page is smaller than a real page, so the true penalty usually lands somewhere between the list price and the effective price. Measure the split yourself with the logging schema in how to benchmark proxy providers, and read proxy success rate explained for why the denominator moves so much between targets.

Why we normalise every provider to 100GB

Comparing price lists directly is impossible when every provider draws its tier boundaries somewhere else. The fix is to pick one volume and quote everybody there. We use 100 GB per month because it sits above the pay-as-you-go cliff on most price lists and below the point where deals become bespoke, which makes it the widest band where published numbers are still real numbers.

Every provider we track carries that normalised figure alongside its measured performance on the live benchmark table, ranked directly in the cheapest residential proxies comparison, and broken down per network on pages like Webshare. How the price component feeds the overall score is set out in our methodology.

Estimating your bandwidth before you sign

Most teams guess this number and guess low. You can measure it in an afternoon by logging response sizes on your existing crawl, proxied or not, and multiplying:

1. Count requests per month, including retries, health checks and anything a scheduler fires while you sleep.
2. Take the mean response size in kilobytes, weighted across target types rather than sampled from the one page you had open.
3. Multiply and divide by one million to get gigabytes. Three million requests at 180 KB is 540 GB a month.
4. Add the request side. Headers, cookies and POST bodies are metered too, and a fat cookie jar on millions of requests is not a rounding error.
5. Reconcile against the provider dashboard during the first week. Their meter is the one that bills you, and a persistent gap above a few percent deserves a support ticket rather than a shrug.

With a real number in hand, the remaining work is commercial: which tier you land in, what the terms cost, and which provider is credible at that volume. The selection framework is in how to choose a proxy provider, and the underlying economics of the product are in what are residential proxies.