An ISP proxy, also sold as a static residential proxy, is one fixed IP address registered to an internet provider and run from a server. A website that looks up the address sees a provider such as AT&T or Comcast. You get a server's uptime and one address that stays yours for the whole plan. One IP cost a median $3.50 for 30 days at 31 sellers in our October 2026 price records. We sell ISP proxies too, so weigh our view with that in mind.
Every public address has three records: who holds it, who routes it, and what data companies call it. An ISP proxy works in the gap between those records and the machine that answers. This page explains that gap from the registry rules upward. It covers why one address costs money, how a provider's name reaches a server, what a website can check, and what research says about an address's history. It rests on IETF standards, the registries' own statements and peer-reviewed studies, all listed under Sources.
What is an ISP proxy?
An ISP proxy is a single static IP address that carries an internet provider's name and answers from server hardware. To a website it looks like a customer of that provider. To you it behaves like a server: always on, fast and the same address every day. It is the proxy to use when a job needs residential trust and must not change IP.
ISP proxy
static proxy
One fixed IP address registered to an internet provider and run from a server. A website that looks up the address sees the provider, and you keep a server's uptime and the same address for the whole plan.
- Also called
- static residential proxy
The term comes from the market. We found no IETF standard, registry policy or peer-reviewed paper that defines it, so each seller draws its own line. Residential proxies have been studied since at least 2019, when Mi and colleagues measured them for the IEEE Symposium on Security and Privacy. We found no study of ISP proxies as a class. The evidence on this page comes from the layers underneath: addresses, routes and leases.
Is a static residential proxy the same as an ISP proxy?
In most sellers' usage, yes. Webshare's help center opens with "Static Residential Proxies, also known as ISP Proxies". "ISP" names who the address belongs to, and "static residential" names how it behaves. Three things still differ from seller to seller:
- Whether the IP stays fixed. Most sellers pin one IP for the plan. Shifter keeps the city and network fixed while its ISP addresses refresh over time.
- Where the address comes from. IPOasis sells "Native" IPs registered to providers in the target country and cheaper "Broadcast" IPs announced through ISP partnerships.
- Whether you share it. A dedicated IP serves only you, and a shared one serves a few customers at once.
Ask which one a seller means before you buy.
Why does one IP address cost money?
Because IPv4 addresses ran out. On February 3, 2011, IANA gave each regional registry its last block. The registries then emptied one by one. APNIC reached its final block on April 15, 2011, ARIN's free pool ran out in September 2015, and RIPE NCC made its final /22 allocation on November 25, 2019. At three of the five registries, new addresses now come through a waiting list. In February 2025 the wait stood at 498 days at RIPE NCC, 615 at ARIN and 1,417 at LACNIC, according to Bernhard Degen and colleagues.
Feb 2011
IANA hands out its last blocks
Apr 2011
APNIC reaches its final block
Sep 2015
ARIN's free pool runs out
Nov 2019
RIPE NCC makes its final /22 allocation
Feb 2024
AWS starts charging for public IPv4
Scarcity turned addresses into goods. Between 2008 and 2017, every registry passed rules that allow transfers, and over 6,184 transfers moved 30.2 million addresses in 2024. A block can also be rented. Ben Du and colleagues give the formal test: an organization leases addresses "when it provides its customer with IP addresses but does not provide the customer connectivity to the Internet".
Prices followed. Degen's team recorded the asking prices at two brokers from December 21, 2024 to March 12, 2025. The median lease was $0.53 a month at one broker and $0.47 at the other. Buying one address cost a median $33.93, so a purchase equals 68 months of rent at $0.50. AWS has charged $0.005 an hour for every public IPv4 address since February 1, 2024, which is $3.60 over 30 days. In 2023 it said the cost of one address "has risen more than 300% over the past 5 years".
That explains why the product is sold per IP: the address is the scarce part. It also shows how much of the price lies elsewhere. A leased address brings no connection, so a seller still pays for a route, a server and the traffic.
How does an ISP proxy get a provider's name?
Three records describe an address, and three different parties keep them.
- The registry record. A regional registry records who holds each block. RDAP, the modern lookup, returns that "registration information" as JSON, and WHOIS is the older way to read it. A holder can pass part of a block to another organization, and that part gets its own entry.
- The route. An address carries traffic only after a network announces it through BGP, the internet's routing protocol (RFC 4271). That network belongs to an autonomous system, a group of address blocks under one routing policy (RFC 1930). Its public number, the ASN, names the origin of the route.
- The label. Data companies then sort each network into types such as ISP or hosting, and estimate where its addresses sit.
Nothing forces these records to match the machine that answers. A block can carry a provider's name while the server that uses it sits in a data center. An ISP proxy depends on that split, and sellers describe two routes to it. IPOasis sells "Native" IPs "registered to internet service providers in the target country" and "Broadcast" IPs "announced via ISP partnerships". Shifter's glossary says a seller "partners with consumer ISPs to obtain IPs registered under the ISP's name and serves them from its own infrastructure".
Oxylabs is candid about the server side: its static residential proxies "originate from datacenter proxies, therefore they are just as fast as their counterparts". A plain datacenter proxy shows its hosting company in the same lookup and fails it on defended sites.
How does leasing move a block?
Leasing is one way for a block to reach a company that did not register it. Our glossary entry on IP leasing covers the basics, and Du's team names four roles:
IP holder
stays on the record
Broker
arranges it, if used
Lessee
pays to use it
Originator
announces it in BGP
The holder can then sign a route origin authorization, a record showing that it "has authorized an Autonomous System (AS) to originate routes" for the block (RFC 9582). The lease itself does not have to pass through the registry, because the block stays registered to the holder. Degen's team notes that the real user can stay out of WHOIS altogether. Their review found no registry policy that explicitly allows or forbids leasing. ARIN's own statement says leasing "can be done if the addresses have been legitimately obtained and subsequently become unused", and it asks lessors to record leases in its Whois database.
Du's team inferred 47,318 leased prefixes in April 2024. That is 4.1% of the 1,146,921 prefixes they saw in BGP, and 0.9% of routed address space. Degen's team ran the same method on a later table and counted 62,944 of 1,032,086 prefixes, or 6.10%, in early 2025. We cannot say how many ISP proxies sit on leased blocks. Sellers do not publish it, and a lease is only one route.
Can a website tell it is an ISP proxy?
Sometimes. A lookup of the ASN shows the provider, which is the point of the product. A site can look at much more than that, and most of it is public data:
| Check | What it reads | Who sets it |
|---|---|---|
| Holder and origin | Two records | Registry, router |
| Network label | Type of network | Data firms |
| Location | Stated or measured | Operator, physics |
| Timing | Two round trips | The path |
| Reputation | Past abuse | Blocklists |
| Client | TLS and habits | Your software |
Holder against origin
If a block's holder and its origin network have no business link, the block is probably leased. That is the test Du's team built. On 14,856 prefixes with known answers, it reached 98% precision and 82% recall, and the authors published their code and data. Any defender can run the same comparison.
Network labels
Labels often disagree. Webshare's own help center answers "the most common question" it gets, why a checker tool shows "datacenter":
Tools that check the hosting environment may label the IP as "datacenter" or "hosting", because, technically, that's where the server is.
Research shows how soft that line is. Maya Ziv and colleagues tested seven data sources against hand-checked networks in 2021. Most were "nearly two times worse" at sorting technology organizations, such as ISPs and hosting companies, than at sorting other kinds. All but IPinfo scored below 63% on hosting providers. So two checkers can give different answers for one address, and each is reading its own database.
Location
A stated location is weak evidence. RFC 8805 lets an operator publish its own locations in a geofeed, and it warns that "only out-of-band verification methods are implemented". Ioana Livadariu and colleagues found about 1.5% of allocated IPv4 prefixes covered in 2024. The feeds were wrong for 0.9% of client, 4.0% of router and 8.5% of server addresses they tested, so the authors treat them as a hint.
Measured location is harder to bend, because delay turns into distance. Robert Beverly and colleagues converted round-trip times at two thirds of the speed of light and audited 62,084 IPv4 prefixes for IMC 2026. Of those, 98.2% sat in the world region their records named. Three commercial location databases caught between 7% and 92% of the mismatches, depending on the database and the region. So an address can pass one location database and fail the next.
Timing
A proxy splits the connection. Diwen Xue and colleagues describe it at NDSS 2025: transport sessions "break into two parts at the proxy", while the encrypted session runs from you to the site. So the site's TCP round trip ends at the proxy, and its TLS round trip reaches your machine. Elisa Chiapponi and colleagues compare the two from the server side. In a four-month campaign, a gap above 50 ms marked a residential proxy request.
ISP proxies were not in those tests. The logic should still hold for any proxy that relays a connection this way, and the gap should grow with the distance between you and the proxy. The signal also has false alarms. Some mobile carriers run their own proxies, and a phone on a French 4G network measured 51.5 ms with no VPN or proxy service.
Reputation and client
Per-IP defenses remember what an address did before you arrived, and the next section covers that record. Your client gives you away too. A TLS fingerprint describes how your software opens a connection. Cloudflare's JA3 and JA4 documentation says these fingerprints "identify TLS clients based on how they initiate connections". A trusted IP carrying an automation library's fingerprint still looks like automation. Timing and headers count as well, and our guide to how websites detect proxies covers them.
Defenders know the product well. GeeTest, a bot-defense company, calls static residential proxies "a critical challenge for modern bot detection". A good ISP IP gets you past the first check, and your own behavior has to carry you past the rest.
What it gives you
A provider's name
the ASN lookup shows an ISP
Server uptime
runs in a data center, always on
One fixed identity
the IP stays the same under a login
Unmetered traffic
some sellers cap fair use at 100 GB
What it costs you
One blast radius
flag the IP and all of it is flagged
Inherited history
the address had tenants before you
Hosting checks
some tools may label it datacenter
Monthly rent per IP
costly when you need many identities
Does a leased address carry a bad history?
More often than other addresses, and two peer-reviewed studies measured it. Du's team found 1.1% of leased prefixes announced by networks on Spamhaus's ASN-DROP list in April 2024, against 0.2% of other prefixes. That is about five times as often. Degen's team checked address blocklists against a matched sample from January 11 to February 10, 2025. On the strict FireHOL Level 1 list, 0.518% of leased prefixes appeared against 0.179% of the others, or 2.89 times. On the wider FireHOL set it was 32.0% against 17.4%, or 1.84 times.
Among the types of abuse, the gap was widest on the lists of open proxies and Tor exits, at 2.59 times. Spam ran the other way, with leased prefixes 1.55 times less likely to be flagged. The authors warn that these groups differ in size, so compare them with care.
Two more findings help a buyer:
- Price says nothing about history. Across 930 /24 blocks at one broker in February 2025, the lease price had no significant link to past blocklist entries (r = -0.007, p = 0.83). The authors advise against using the price as a sign of reputation.
- Reputation moves during a lease. The rate of abuse listings rose 60.2% in the first 30 days and 269% by 150 days. Among blocks that got listed, the median wait was 27 days. And 44.4% of 20,351 observed leases lasted one month, so a block can change hands often.
These figures cover every leased block, including those of VPN and hosting companies. They describe the market an ISP address may come from. They cannot tell you about any one seller, ours included.
Why does a login hold on an ISP proxy?
When you sign in, the site ties your session to a cookie and often to the IP the session started on. OWASP's session guidance recommends binding the session ID to "the client IP address". A change of IP in the middle of a session, it says, "is a very good indicator of session manipulation and hijacking attempts".
A rotating residential proxy changes the IP by design, per request or when a sticky window ends. Sellers' sticky windows default to 10 or 15 minutes and stretch to 24 hours at most. Our guide to sticky vs rotating proxy sessions has the details. An account you keep for weeks outlives any of them. An ISP proxy keeps one IP for the whole plan, so the site sees the same provider's address every day. That is why account management, storefronts and logged-in bots run on static IPs.
ISP proxy or rotating residential?
Both read as residential to a site. They differ in what you hold and how you pay:
| ISP (static residential) | Rotating residential | |
|---|---|---|
| Network a site sees | A home provider | A home provider |
| Runs on | A server | Home devices |
| Your IP | One, fixed | New per request or session |
| Login over weeks | Holds | Breaks when the IP moves |
| After a ban | Replace the IP | The next request moves on |
| Priced | Per IP per month | Per GB of traffic |
Our page on rotating vs static residential proxies prices that choice at eight sellers. One static IP a month costs about as much as 1 GB of rotating traffic. Datacenter proxies, the third option, are compared in datacenter vs residential proxies.
How much does a static residential proxy cost?
One IP for 30 days cost a median $3.50 in our ISP proxy price study, which read 31 sellers' price lists between September 26 and October 26, 2026. The rate falls as the order grows:
Three things move that price:
- Shared or dedicated. Shared static IPs cost far less: Decodo lists shared ISP IPs from $0.27, Evomi charges $1.00 and Bright Data $1.80, in packs of 10.
- Where the address comes from. IPOasis charges $5.49 for Native IPs against $3.99 for Broadcast ones, and Evomi charges $4.50 for "Virgin" IPs nobody has used.
- Pack sizes. Nine of the 31 sellers sell no single IP, so one IP costs their smallest plan. Decodo sells at least 3 IPs, ProxyOmega 5 and Bright Data 10.
"Unlimited" traffic has fine print too. MarsProxies, Fastproxy, IPOasis and ProxyOmega list unmetered bandwidth, and Evomi says its static IP has "no data meter". Bright Data and ProxyEmpire cap fair use at 100 GB per IP a month. Oxylabs offers "unlimited bandwidth with fair usage" without a figure.
Our ISP proxies cost $2.70 a month for one dedicated static IP with unmetered traffic, on US and EU ranges. Per-IP plans are where our traffic is unmetered, and one balance, which never expires, pays for every product. The ISP price table also lists our rotating ISP pool, billed per gigabyte. For every proxy type side by side, see how much do proxies cost.
What are the catches?
One identity means one blast radius
A rotating pool shrugs off a ban, because the next request takes another IP. An ISP proxy is one address, so if it gets flagged, everything on it is flagged together. Keep one account, or one narrow purpose, per IP.
A static IP inherits its history
An ISP address had tenants before you, and per-IP defenses remember them. The leasing research above shows that a price cannot tell you about that past. Check an IP before you build on it.
"Static" and "unlimited" need reading
Some sellers refresh ISP addresses over time, and some cap "unlimited" traffic at 100 GB per IP a month. Read the plan terms before you put an account on an IP.
How do you check an ISP proxy before you trust it?
Five checks take a few minutes, and two of them use free public services. We ran those two on Google's public DNS address on November 2, 2026, so you can repeat them and compare.
1. Find the exit IP. Connect through the proxy and open our proxy checker. It shows what a site sees when you connect.
2. Read the registry record.
curl -sL https://rdap.arin.net/registry/ip/8.8.8.8
ARIN answers for its own addresses and redirects the others to the right registry. Here it returned block NET-8-8-8-0-2, type DIRECT ALLOCATION, registrant Google LLC.
3. Read the route.
curl -s "https://stat.ripe.net/data/prefix-overview/data.json?resource=8.8.8.8"
RIPEstat returned prefix 8.8.8.0/24, announced by AS15169, Google LLC. Here the holder and the origin match. On an ISP proxy, a provider's name in both records is the cleanest result. A provider in the registry and an unrelated network in the route points to a lease.
4. Compare two or three IP databases. A difference between them is normal, as the label research shows. Read it as a preview of what different sites will see.
5. Check the record and the promise. FFraud runs a free IP abuse check. Note the IP and test it again a week later, to confirm that it stays fixed.
Do you need an ISP proxy?
Walk your job down this list and stop at the first match:
- Does anything log in and need to stay logged in past a sticky window? A static ISP proxy, one IP per account.
- Are you running accounts, a storefront or posts as one steady identity? An ISP proxy. Each account should come from one consistent address.
- Is it high-volume collection across many pages? Rotating residential. One static IP doing a pool's job gets itself banned.
- Does the target barely check (open data, most APIs)? A datacenter proxy, which is cheaper still.
- Both collection and accounts in one project? Split them: rotating residential for the crawling, ISP for the logins.
The common mistake is buying a rotating pool for account work because "residential" sounded right, then fighting logouts for a week. If a task has to be the same person tomorrow as today, it is an ISP job. Our guide to what a residential proxy is covers the rotating side in depth.
What this page could not check
- No study of the product itself. We found no peer-reviewed measurement of ISP proxies as a class. The research here covers leased address space and residential proxies, and we say where we extend it.
- Leases are inferred. Both leasing studies read leases from public records. Du's method missed 18% of known leases, and some counted leases may be something else.
- Older label data. The study of network labels dates from 2021, and the databases have changed since.
- One abstract. We read the abstract of the geofeed paper, because the full text refused automated reads.
- No performance data. We measured no speed, success rate or block rate, for any seller, ours included.
- How sellers get their addresses. We know only what each seller says about it.
- Prices. 18 of the 31 price lists date from September 26 and 27, 2026, and seller terms come from pages read on October 22, 2026.
- Check by February 2, 2027. Prices, caps, lease rates and waiting lists change, so every number here needs a fresh read by then.
Sources
Standards
- Hawkinson, J. and Bates, T. RFC 1930, Guidelines for creation, selection, and registration of an Autonomous System (AS). IETF, March 1996.
- Rekhter, Y., Li, T. and Hares, S. RFC 4271, A Border Gateway Protocol 4 (BGP-4). IETF, January 2006.
- Kline, E., Duleba, K., Szamonek, Z., Moser, S. and Kumari, W. RFC 8805, A Format for Self-Published IP Geolocation Feeds. IETF, August 2020.
- Hollenbeck, S. and Newton, A. RFC 9083, JSON Responses for the Registration Data Access Protocol (RDAP). IETF, June 2021.
- Snijders, J., Maddison, B., Lepinski, M., Kong, D. and Kent, S. RFC 9582, A Profile for Route Origin Authorizations (ROAs). IETF, May 2024.
Papers
- Beverly, R., Phokeer, A. and Gasser, O. "WHEREIS: IP Address Registration Geo-Consistency." ACM Internet Measurement Conference, October 2026.
- Chiapponi, E., Dacier, M. and Thonnard, O. "Poster: The Impact of the Client Environment on Residential IP Proxies Detection." ACM Internet Measurement Conference, October 2023.
- Degen, B., Du, B., Mok, R. K. P., Sommese, R., Jonker, M., van Rijswijk-Deij, R. and claffy, kc. "From Scarcity to Opportunity: Examining Abuse of the IPv4 Leasing Market." Network Traffic Measurement and Analysis Conference, 2025.
- Du, B., Fontugne, R., Testart, C., Snoeren, A. C. and claffy, kc. "Sublet Your Subnet: Inferring IP Leasing in the Wild." ACM Internet Measurement Conference, November 2024.
- Livadariu, I., Vermeulen, K., Mouchet, M. and Giotsas, V. "Geofeeds: Revolutionizing IP Geolocation or Illusionary Promises?" Proceedings of the ACM on Networking 2, CoNEXT3, August 2024.
- Mi, X. et al. "Resident Evil: Understanding Residential IP Proxy as a Dark Service." IEEE Symposium on Security and Privacy, 2019.
- Xue, D., Stanley, R., Kumar, P. and Ensafi, R. "The Discriminative Power of Cross-layer RTTs in Fingerprinting Proxy Traffic." Network and Distributed System Security Symposium, February 2025.
- Ziv, M., Izhikevich, L., Ruth, K., Izhikevich, K. and Durumeric, Z. "ASdb: A System for Classifying Owners of Autonomous Systems." ACM Internet Measurement Conference, November 2021.
Registries and platforms
- Pan, G. "Keep calm and carry on: The status of IPv4 address allocation." APNIC Blog, December 13, 2019.
- Sweeting, J. "Where Does Waiting List Address Space Come From?" ARIN, September 23, 2019.
- Sweeting, J. "ARIN and IPv4 Address Leasing." ARIN, February 23, 2023.
- RIPE NCC. "The RIPE NCC has run out of IPv4 Addresses." November 25, 2019.
- Barr, J. "New: AWS Public IPv4 Address Charge + Public IP Insights." AWS News Blog, July 28, 2023.
- OWASP, "Session Management Cheat Sheet", read October 22, 2026.
- Cloudflare, "JA3/JA4 fingerprint", bot management documentation, read October 22, 2026.
Sellers and defenders
- Webshare, "What is a Static Residential Proxy (ISP Proxy)?", help center, updated June 17, 2026.
- IPOasis, static residential proxies page, read October 22, 2026.
- Shifter, "What Is an ISP Proxy?", glossary, read October 22, 2026.
- Oxylabs, static residential proxies product page and dedicated ISP pricing, read October 22, 2026.
- MarsProxies, Fastproxy, Decodo, Bright Data, ProxyEmpire, Evomi, Proxy Cheap and ProxyOmega, ISP pricing pages, read October 22, 2026.
- GeeTest, "What Is a Static Residential Proxy?", December 26, 2025.
Our own data
- HProxy price records: 31 sellers' own price pages, read September 26 to October 26, 2026.
- HProxy ISP product pages and price list, read October 22, 2026.
- RDAP and RIPEstat lookups of 8.8.8.8, November 2, 2026.



