ISP Proxy Pricing: What One Static IP Costs a Month

Budget between $1.30 and $2.80 per dedicated ISP IP per month if the address must be exclusively yours, and between $0.30 and $1.50 if you can accept a shared or semi-dedicated address. Those bands come from the published ladders in the next section, all read on 23 September 2026, and all of them will have moved by the time you read this.

Two rules fall out of the ladders, and they matter more than picking a vendor.

Exclusivity roughly doubles the price. On Bright Data's own ISP pricing page the shared tier at 100 IPs lists $1.45 per IP and the dedicated tier at the same count lists $2.75 per IP, a factor of 1.9. Oxylabs prices its ISP product at $1.30 per IP at the 100 IP tier and states on the same page that those addresses are "shared with up to 3 users", while its separately listed Dedicated ISP product starts from $2.50 per IP. The premium for sole tenancy is consistent enough across vendors to treat as a rule of thumb.

The address count is your lever, and it is the wrong one. Per-IP billing is flat in bytes and flat in requests. It moves only when you buy more addresses, which you do when a target tightens its per-IP tolerance. That is a decision your target makes, not you. Understand it before you sign, because it is the failure mode of the whole model.

An ISP proxy is a static address hosted on datacenter hardware but registered in a netblock belonging to a consumer internet provider. The mechanics are in what ISP proxies are. The price reflects five things, and vendors bundle them differently:

1. The ASN lease. The provider pays an internet service provider for addresses inside that ISP's registered range. Comcast, Lumen, BT Group, AT&T, Orange and Verizon all appear by name on the vendor pages read for this piece. That lease is the largest input cost, and it is why ISP addresses cost multiples of a datacenter IP.
2. Exclusivity. Sole tenancy means no other customer can burn your address on the target you care about. The biggest single price lever.
3. Tenure. How long the address stays assigned to you. Bright Data's ISP page advertises "keep your IPs for life" alongside its per-IP model. Others reassign at the end of a term.
4. Bandwidth terms. Usually described as unlimited, usually with a fair usage clause that does something specific at a threshold. Section five has one vendor's exact wording.
5. Replacement rights. What happens when an address gets burned: free swap, charged swap, or a cap per cycle. Almost nobody puts it on the pricing page.

If a quote does not answer points two, three and five, it is not a quote, it is a number.

Every figure below was read off that vendor's own ISP or static residential pricing page on 23 September 2026. The tax notes are the vendors' own wording. Confirm the current rate on their site before you buy, because these ladders change quarterly.

The cheapest and the dearest dedicated options differ by a factor of about 9 ($2.75 divided by $0.30 is 9.2), and both describe the address as exclusively the buyer's. Price cannot be distinguishing exclusivity here, so something else is: ASN quality, subnet spread, replacement policy, and how hard neighbouring customers have hit your targets. None of that appears on a pricing page. You find it by testing, which is what how to spot a fake proxy provider is for.

Semi-dedicated tiers sit in the middle, roughly halfway between budget and premium dedicated. If you are buying ISP addresses to hold logged-in sessions that is a strange purchase: residential-grade money for an address three strangers can also burn. The same argument one tier down is in shared versus dedicated datacenter proxies.

Most buyers size the order and never look at the term selector. On IPRoyal's ISP page, read the same day, four terms are offered: 24 hours from $1.80 per proxy, 30 days from $2.70, 60 days from $2.55 and 90 days from $2.40. The headline rate falls about 11% from the 30 day option to the 90 day one, a bigger move than most volume ladders give you below 500 addresses.

One caution on reading any term ladder. A page showing a term and a per-proxy price side by side often does not say whether that figure is the total for the term or a per-30-day equivalent, and those two readings differ by a factor of three on a 90 day row. Put one proxy in the cart and read the total before you extrapolate to 200 of them. A per-proxy figure with no period attached is not a price.

The buying rule is unglamorous: shortest term for evaluation, then the longest term you are confident you will still want in three months. The commitment discount is real, and so is the cost of holding addresses for a project that ended in week two. Our roundup of proxy free trials and refund policies covers what you can get out of before committing.

"Unlimited bandwidth" is the standard claim on per-IP plans and it is mostly honest. What varies is the far end, and one vendor publishes the mechanism precisely enough to quote.

Oxylabs states its ISP fair usage policy on the product page: up to 50 GB of usage per ISP proxy in a month you get 100 concurrent sessions per purchased proxy, and once 50 GB per proxy is reached, concurrency drops to 10 sessions per proxy for the rest of the billing cycle. Their own example: 100 ISP proxies gives 10,000 concurrent sessions until 5,000 GB of usage, after which 1,000 sessions remain until the cycle ends.

That is not a bandwidth charge. It is a tenfold concurrency cut, which is worse for a scraper than an overage invoice, because it silently stretches your job instead of billing you. Nobody spots it in the dashboard. They spot throughput halving on the 22nd of the month.

Work out where that line sits for your workload before you commit. Taking 1 GB as 1,000 MB:

For account management the clamp is irrelevant, because a human-paced session moves a few hundred megabytes a month. For scraping at volume it is the real ceiling on the plan, and it is expressed in concurrency rather than dollars. Sizing that side of the problem is covered in concurrent connections in proxies, and the general case in whether unlimited bandwidth is worth it.

Ask every vendor the same question in the same words: what specifically changes when I cross the fair usage line, and is it a charge, a throttle, or a concurrency reduction? If they will not answer in writing, assume the worst of the three.

Webshare publishes a full ladder on its static residential page, and the shape of it is instructive. Read on 23 September 2026: 20, 50, 75, 100 and 250 proxies all price at $0.30 per proxy. The first discount appears at 500 proxies, $0.285, marked 5% off. Then 1,000 at $0.27, 2,000 at $0.255, 3,000 and 5,000 at $0.24, with 10,000 listed at $0.225.

Read that again. Going from 20 addresses to 250, a 12.5x increase in order size, earns a 0% discount. The entire discount curve lives above 500 addresses.

The shape repeats across the other ladders read for this article. Oxylabs moves $1.60 to $1.30 to $1.20 across its 10, 100 and 500 IP tiers. Decodo's dedicated ladder runs $3.33 at 3 IPs down to $2.50 at 200. Bright Data's shared ladder runs $1.80 at 10 IPs down to $1.30 at 1,000. The interesting savings always sit at counts most buyers never reach, and the range where real teams operate is nearly flat.

So do not inflate an order to chase a discount tier. Between 20 and 250 addresses you pay list either way, so buy the count your target's per-IP tolerance requires and nothing more. Measuring that tolerance is the subject of how to measure proxy success rate.

Three costs sit outside the per-IP figure, and all three get missed in budgets.

Tax. Two of the five pages read for this article state tax separately. Decodo's ladder reads "+ VAT billed monthly" and Oxylabs' reads "VAT may apply". For an EU buyer at a 20% rate, a $260 order is $312. That is not a hidden fee, it is printed on the page, but it is not the number people quote to finance.

Replacement churn. Addresses get burned. If your provider charges for swaps, or caps free swaps per cycle, your effective monthly cost is rent plus replacements, and the replacement rate depends on your targets rather than on the vendor's marketing. None of the five pages read for this piece published a replacement fee schedule. Ask before you buy, and ask what evidence they require that an address really is burned.

Renewal. Introductory pricing is normal here, and none of the pages read for this article published a renewal rate distinct from the list rate. That is not evidence there is no uplift, it is evidence the pages do not address it. Check the renewal price in the dashboard after your first purchase and before your second billing date. We deliberately print no renewal-uplift percentage, because we could not verify one from a vendor's own page on the day of writing, and a figure carried over from memory is worth less than nothing in a pricing article.

The per-IP price is an input, not an answer. The output you care about is cost per successful request, and it is easy to compute once you have a block rate.

Take a workload of 300,000 page fetches in a month spread across 100 addresses. That is 3,000 requests per address per month, or 100 per address per day, a sane pace for most targets. These are model assumptions for illustration, not measurements from any test we ran.

Note what that table does not say. For pure scraping the cheap pool wins the arithmetic even at a 30% failure rate, because a failed request on a per-IP plan costs nothing beyond the wait. If your only measure is cost per extracted record, the $30 option is defensible and you should test it first.

The table stops being useful the moment failures are not free. If a block means a logged-in account gets flagged, the cost of the failure is the cost of rebuilding that account and it dwarfs the rent. At $2.75 per IP per month a dedicated address costs $33 a year. The shared alternative at $1.45 costs $17.40 a year, so sole tenancy costs $15.60 per address per year, or $1,560 across 100 addresses. If avoiding one account loss a year is worth more than $1,560, the dedicated tier is cheap insurance rather than an expense.

Price the failure, then pick the tier. Buyers who skip that step either overspend on dedicated addresses for a scraper that does not care, or underspend on shared addresses holding accounts that do.

A team runs 100 long-lived logged-in sessions, one per address, mostly browsing at human pace. Traffic is around 2 GB per address per month, far under any fair usage line. Blocks mean account review, so failures are expensive. Twelve month horizon.

Step one: rule out shared. At 2 GB per address the bandwidth terms are irrelevant, and the whole value of the purchase is that nobody else touches the address. Oxylabs' ISP tier is published as shared with up to 3 users, so it is out regardless of price, though the Dedicated ISP product they list separately would qualify.

Step two: price the qualifiers at the 100 IP tier. Bright Data dedicated at $2.75 per IP is $275 a month, $3,300 a year. Decodo dedicated at $2.60 is $260 a month, $3,120 a year plus VAT. Webshare static residential at $0.30 is $30 a month, $360 a year, and their page states the addresses are assigned only to the buyer.

Step three: do not buy on that spread. Buy 10 addresses from two of them, hold real sessions for three weeks, and count challenge prompts and re-verification events per address. The annual gap between cheapest and dearest here is $2,940, real money and also less than most teams spend recovering from a bad month of account flags. The test costs roughly $30.

Step four: budget the second-order costs. Tax where it applies, replacement swaps, and one address in reserve per ten in production so a burned IP does not idle a worker. At 100 production addresses, buy 110.

Total at the premium tier with that reserve: 110 at $2.75 is $302.50 a month, $3,630 a year before tax. At the budget tier: 110 at $0.30 is $33 a month, $396 a year. Both are real answers, and the three week test decides which one is yours.

Per-IP billing suits identity work. It suits high-volume scraping badly for one reason: the bill is driven by address count, and address count is driven by your target's per-IP tolerance, which is outside your control. If a target drops from tolerating 500 requests per address per day to 50, a per-IP plan needs ten times the addresses to do the same work.

Three alternatives, with the honest case for each.

Datacenter proxies on a flat concurrency plan. If your target does not inspect ASN, ISP addresses are money spent on a check nobody is running. SparkProxy sells datacenter capacity this way: Starter $75 a month for 100 threads, Core $140 for 250, Boost $240 for 500, Plus $440 for 1000, all unlimited bandwidth with 30 days validity, across 1M+ datacenter IPs in 80+ countries including 50,000+ US addresses. Higher Pro and Pro+ thread tiers exist in the fair usage policy without a published price. The gateway is gateway.sparkproxy.io, port 11000 for HTTP and HTTPS, 11002 for sticky sessions, 13000 for SOCKS5. A flat plan is right when volume and page weight are both unpredictable, and wrong when a target rejects hosting ASNs outright. ISP proxies vs datacenter proxies has the test for telling those two cases apart.

Per-GB residential. Better when you need many distinct addresses briefly rather than a few forever. The arithmetic differs enough to need its own treatment, in residential proxy pricing per GB.

A managed API that hides the address layer. SparkProxy's Scraping API bills credits rather than addresses: 1,000 free credits with no card, then Starter $49 for 250,000 credits a month at 50 concurrent, Growth $99 for 1,000,000 at 100 concurrent, Pro $249 for 3,000,000 at 200, Scale $599 for 8,000,000 at 400. A plain fetch through the rotating pool is 1 credit and JavaScript rendering is 5. Residential exits come from premium_proxy=true, which costs 10 credits without rendering and 25 with it, and country_code adds 5 credits.

curl -G "https://scrape.sparkproxy.io/api/v1" \
  -H "X-API-Key: YOUR_API_KEY" \
  --data-urlencode "url=https://example.com/account/summary" \
  --data-urlencode "render_js=false" \
  --data-urlencode "premium_proxy=true" \
  --data-urlencode "country_code=US"

That call costs 15 credits: 10 for the premium exit without rendering, plus 5 for country targeting. On the Growth plan at $99 for 1,000,000 credits, 15 credits per request works out to 66,666 requests a month, about $1.49 per 1,000. Compare that against the $0.92 per 1,000 attempts in the cost table above and the ranking depends on what you are counting: the API figure includes the unblocking and the maintenance, the ISP figure does not. Credits are deducted before the request runs and refunded automatically on failure, and the response carries X-Credits-Used and X-Duration-Ms so you can cost a job before scaling it. Full mechanics are in scraping API credit pricing.

Put these to support in writing before you pay, not to the marketing page.

1. Is each address assigned exclusively to my account for the full billing period, or shared with other customers? If shared, with how many?
2. What exactly happens when I cross the fair usage line: a charge, a throttle, or a concurrency reduction, and at what threshold?
3. How many distinct /24 subnets and how many distinct ASNs will my allocation span in the country I need?
4. What is the replacement policy for a burned address: free or charged, capped per cycle, and what evidence do you require?
5. What is the renewal rate after my first term, and is the current price promotional?
6. Do I keep the same addresses across renewals, or are they reassigned?

Questions one, two and four change the real cost. Question three predicts whether the plan still works at three times the size. A vendor who answers all six clearly is worth more than a 20% price difference. Check the answers against the contract, since the marketing page and the SLA often disagree, and note that whitelist-based authentication has its own gotchas, covered in IP whitelisting for proxies.