What Residential Proxies Actually Cost per GB in 2026
One vendor advertises $0.49 per GB. If you use 5 GB a month, you will pay $10.00 per GB. Both numbers are honest, and the vendor is not hiding anything — the entry plan is simply 100 GB, and a 5 GB user buys the whole plan.
That gap is the entire subject of this post. Residential proxy pricing is quoted in dollars per gigabyte, but almost nobody pays the quoted rate, because the quoted rate is a volume price and your bill is set by minimums, plan sizes and expiry terms that sit somewhere else on the page.
Below is what ten providers actually cost at three real monthly volumes, with every input read off the vendor's own pricing page on 13 August 2026. We do not sell proxies and we take no affiliate commission from anyone here.
TL;DR: Advertised rates run $0.49–$4.90/GB. Effective rates for a 5 GB/month buyer run $1.00–$40.00/GB. The worst gap is 47× (SOAX, whose cheapest paid plan is $200/month); the most-quoted is 20× (Evomi's $0.49 becoming $10.00). The ranking of "cheapest" completely reorders between 5 GB and 50 GB — Evomi goes from second-most-expensive to joint-cheapest without changing its price. Six of ten pricing pages do not say whether unused traffic expires.
The advertised rate is a volume price
Every per-GB figure on a proxy pricing page is the rate at some volume. Usually it is the rate at the largest volume the vendor sells, because that is the smallest number they can legally print.
Rayobyte advertises "as low as $0.50/GB" — true at 5,000 GB and above. SOAX's $0.85 is its enterprise tier-1 rate, on a $3,000-a-month plan. Evomi's $0.49 is the rate on its 100 GB entry plan, which is also its smallest plan. IPRoyal's $1.75 requires a 10 TB purchase.
None of that is deceptive. It is just that "as low as" is doing a great deal of work, and the number that survives into your invoice is a different one. So the only fair way to compare these vendors is to fix a volume and ask what the cheapest published route to that volume costs.
That is the rule used throughout this post, and you can re-run it yourself:
Effective cost per GB = (the cheapest route the vendor publishes to the volume you need) ÷ (the gigabytes you actually use). Where a vendor publishes no route to a given volume, this post says "not published" rather than interpolating between tiers.
What ten providers cost at 5 GB, 50 GB and 1 TB
"Advertised" below is the lowest per-GB figure the vendor puts on its own pricing page. Everything else is that page's published plans, divided.
| Provider | Advertised | Effective at 5 GB/mo | Effective at 50 GB/mo | Effective at 1 TB/mo |
|---|---|---|---|---|
| SOAX | $0.85 | $40.00 — $200 Builder floor | $4.00 — same $200 floor | Credit model, not directly comparable |
| Massive | $4.90 | $16.00 — $80 Starter, 10 GB | $7.70 — $385 Scaler, 70 GB | Not published |
| Evomi | $0.49 | $10.00 — $49.99 entry plan, 100 GB | $1.00 — same plan | $0.32 — $320/1,000 GB |
| Oxylabs | $2.50 | $6.00 — $30 Starter, 5 GB | $10.00 — $500 Advanced, 125 GB | $2.50 — $2,500 Corporate |
| IPRoyal | $1.75 | $5.25–$5.95 — between published anchors | Not published | Not published |
| Bright Data | $2.50 | $4.00 pay-as-you-go with coupon ($8.00 list) | $4.00 with coupon ($8.00 list) | Custom quote above 1 TB |
| Decodo | $2.75 | $4.00 — pay-as-you-go | $3.00 — $150/50 GB plan | Not published |
| Webshare | $1.40 | ≤$3.50 — 1 GB rate as upper bound | ≤$2.75 — 10 GB rate as upper bound | ≤$2.75; $1.40 at 3,000 GB |
| Rayobyte | $0.50 | $3.50 — 1–49 GB band | $2.00 — 50–249 GB band | $0.70 — 1,000–4,999 GB band |
| DataImpulse | $1.00 | $1.00 — $5 minimum buys 5 GB | $1.00 — flat | $0.80 — 1 TB+ rate |
A note on two rows, because they are the ones most easily misread. Webshare publishes anchors at 1 GB, 10 GB and 3,000 GB but not at 5 GB or 50 GB, so its figures are shown as upper bounds taken from the next-lowest published tier — the real number is somewhere at or below them. SOAX does not price in gigabytes at all: it sells credits, and the same gigabyte consumes different amounts of credit depending on the country tier of the IP, so its 1 TB cell is genuinely not comparable to a flat per-GB vendor rather than merely unknown.
The cheapest provider changes with your volume
Read the table down a column instead of across a row and the finding falls out.
At 5 GB a month, cheapest to most expensive: DataImpulse $1.00, Rayobyte $3.50, Webshare at or below $3.50, Bright Data and Decodo $4.00, IPRoyal about $5.50, Oxylabs $6.00, Evomi $10.00, Massive $16.00, SOAX $40.00.
At 50 GB a month: DataImpulse and Evomi $1.00, Rayobyte $2.00, Webshare at or below $2.75, Decodo $3.00, Bright Data and SOAX $4.00, Massive $7.70, Oxylabs $10.00.
Three movements are worth naming:
- Evomi goes from ninth to joint-first without changing a single price. Its entry plan is 100 GB, so every gigabyte you do not use is dead weight at 5 GB and irrelevant at 50 GB.
- SOAX improves tenfold, from $40.00 to $4.00, for the same reason in reverse: a $200 floor is punitive at 5 GB and ordinary at 50 GB.
- Oxylabs gets worse as you grow, from $6.00 to $10.00. Its 5 GB Starter plan fits a 5 GB user exactly; the smallest plan that covers 50 GB is a 125 GB plan, so you buy 75 GB you do not need. This is the one movement in the table that runs against intuition, and it is a pure artefact of plan granularity.
And at 1 TB the order changes again: Evomi's $0.32 and Rayobyte's $0.70 lead, and the flat-rate vendors that looked strongest at 5 GB no longer do, because a flat rate by definition gives you no volume discount. There is no "cheapest residential proxy provider." There is only a cheapest provider at your volume, and the answer moves by a factor of ten across the range most teams operate in.
If you are choosing a vendor rather than costing one, the companion piece on who publishes the "best provider" rankings covers the selection question, including why nearly every ranking you will find was written by a seller.
Four mechanisms that move the real number
1. The entry minimum
The single biggest factor below 50 GB, and the one buried furthest from the headline. Minimums across the ten range from $3.50 to $200 a month. Below roughly 10 GB the minimum is your price, and the per-GB rate is decoration.
2. Plan granularity
Related but distinct: even without a minimum, if the smallest plan covering your volume is much larger than your volume, you pay for the gap. That is what produces Evomi's 20× at 5 GB and Oxylabs' unusual reversal at 50 GB. Look for the plan above your volume, not the rate beside it.
3. Expiring traffic
Bandwidth you paid for and did not use before it expired is bandwidth you paid full price for and received nothing from. Of the ten pages checked on 13 August 2026:
- States it never expires: IPRoyal, Rayobyte (pay-as-you-go), DataImpulse.
- States it rolls over: Massive.
- States an expiry: SOAX — credits die after 60 days on monthly billing, 365 days on annual.
- Silent: the remaining six.
SOAX has the least generous published policy and is also the only vendor that tells you plainly, which is worth more than it sounds. Silence is not a promise, and a support-chat assurance is not a term. Get it in writing before you prepay.
4. Country tiering and credit models
SOAX bills credits against country tiers, so a gigabyte from a tier-1 country costs more than one from tier 3. This is defensible — those IPs genuinely cost more to source — but it means no single per-GB figure describes your bill until you know your own country mix. If most of your targets are US or EU, price the tier-1 rate and ignore the headline.
Per GB is the wrong unit anyway
Here is the part that makes all of the above a floor rather than an answer: you are billed for bytes, but you care about results.
A blocked request still moves traffic. A CAPTCHA page, a challenge interstitial, a 429 response — all of them consume metered bytes and return nothing you can use. So two pools at an identical $2.00 per GB can differ by a factor of several in cost per usable result, and no rate card will ever show you that, because it depends on your targets rather than on the network.
That the underlying pools are noisier than the marketing suggests is now measurable from outside. GreyNoise's Invisible Army report, published 2 April 2026, analysed 4.02 billion malicious sessions from 5.72 million unique source IPs against edge infrastructure over 90 days. It found 39% of unique IPs came from residential address space, and that 78% of residential IPs appear in only one or two sessions and are never seen again. A pool with that much churn is a pool where reputation-based blocking is fighting a moving target — which cuts both ways for a buyer, and is a good reason to test rather than assume.
The practical move is to measure cost per successful request on your own targets during a trial, not cost per GB. Our write-ups on why proxies get banned by Google and what a 429 "unusual traffic" response really means cover what to count as a failure, and rotating versus sticky sessions usually moves that number more than switching vendor does.
The same trap exists one layer up, on the data side, which is why we keep a whole post on whether failed requests get billed.
Work out your own number in five minutes
- Estimate your monthly gigabytes, honestly. Page weight × pages per month. The tier table in residential vs datacenter proxies for SERP scraping includes the pages-per-GB math, which is the part people skip.
- Find the smallest plan that covers that number on each vendor's page — the plan above your volume, not the rate beside it.
- Divide the plan price by your gigabytes, not the plan's. That is your effective rate.
- Check the expiry term and, if it is silent, ask and keep the answer.
- Multiply by your failure rate. If 30% of requests return a block page, your real cost per usable result is about 1.4× the number from step 3.
Do that for three vendors and you will usually find the answer is not the one the rate cards suggest. For the wider budget — bandwidth is rarely the largest line — the true cost of running your own Google scraper and the SERP API cost calculator put it next to the engineering time, and the hidden costs of SERP APIs covers the same class of trap on the data side. If your volumes are small enough, the free limits available without any proxy at all may make this whole calculation unnecessary.
Or price the data instead of the bandwidth
If what you need is search results rather than infrastructure, Serpent bills per request, not per byte — no minimum, no expiring allowance, no plan you have to grow into. 10 free searches, then from $0.60 per 1,000. See full pricing or the Google SERP API.
Get Your Free API KeyExplore: All SERP APIs · Docs
FAQ
What is the real cost per GB of residential proxies?
It depends almost entirely on how much you buy, and the gap between the advertised rate and your rate is largest for small buyers. Using each vendor's own published pricing on 13 August 2026, a buyer using 5 GB a month pays an effective $1.00 per GB at the cheapest end and $40.00 per GB at the most expensive, against advertised rates that range from $0.49 to $4.90. At 50 GB a month the same set of providers runs from $1.00 to $10.00 per GB. The advertised number is a volume price. Your number is the cheapest route the vendor publishes to the volume you actually need, divided by the gigabytes you actually use.
Why is the advertised residential proxy price so different from what you pay?
Four mechanisms, and they stack. Entry minimums: SOAX's cheapest paid plan is $200 a month, so a 5 GB user pays $40 per GB no matter what the rate card says. Plan granularity: if the smallest plan that covers your volume is much bigger than your volume, you pay for the whole plan, which is why Evomi's $0.49 per GB becomes $10.00 per GB for a 5 GB user on its 100 GB entry plan. Expiring traffic: bandwidth that dies at the end of the period is bandwidth you paid for and did not use. And country tiering: on credit-based pricing the same gigabyte costs different amounts depending on where the IP is, so a rate card figure does not describe your mix.
Which residential proxy is cheapest?
The answer reorders completely depending on volume, which is why the question has no single answer. On published prices verified 13 August 2026, Evomi is one of the two most expensive options for a 5 GB buyer at an effective $10.00 per GB, and one of the two cheapest for a 50 GB buyer at $1.00 per GB, purely because its entry plan is 100 GB. Oxylabs gets more expensive as you scale into the middle, from $6.00 per GB at 5 GB to $10.00 per GB at 50 GB, because the smallest plan covering 50 GB is a 125 GB plan. Above about 1 TB the order changes again. Work out your monthly volume first; the cheapest provider is a function of that number, not a fixed fact.
Do residential proxy providers charge for failed requests?
Bandwidth-metered providers bill for traffic, and a blocked request still moves traffic. A CAPTCHA page, a challenge interstitial and a 429 response all consume bytes that count against your allowance, so a pool with a poor success rate on your targets costs more per usable result than its rate card suggests, even though its per-GB price is unchanged. This is the reason cost per successful request is a more useful metric than cost per GB, and it is not a number any vendor publishes, because it depends on your targets rather than on their network.
Does unused proxy bandwidth expire?
It depends on the vendor, and most of them do not say. Of ten providers checked on 13 August 2026, only three state on the pricing page that traffic never expires: IPRoyal, Rayobyte on pay-as-you-go, and DataImpulse. Massive states that unused bandwidth rolls over. SOAX states that credits expire after 60 days on monthly billing, or 365 days on annual. The remaining pages are silent, which is the answer you should like least, because silence is not a guarantee. Get the expiry term in writing before you prepay for a block of traffic.
