Proxy providers, and where each one stops working

Fifteen commercial proxy networks, each with the case it fits and the case it does not. Pricing shape matters more than pool size for most jobs, so both are named.

• Bright Data

  Paid

  A large proxy network covering residential, mobile, ISP and datacenter exits, sold alongside its own scraping and unblocking products.

  Suits

  The widest country and city coverage of any single vendor, with session control granular enough to hold one exit IP across a multi-step checkout or login flow.

  Does not suit

  Access to the residential pool goes through a compliance review with identity and use-case checks, so you cannot sign up and start crawling the same hour. The pricing surface is large enough that a small job often lands on a plan built for a much bigger one.

• Oxylabs

  Paid

  An enterprise proxy network sold with a matching set of scraper APIs for search, commerce and general web targets.

  Suits

  Strong documentation and support for teams running continuous, high volume collection, with per-country and per-city targeting on the residential pool.

  Does not suit

  The commercial shape is built for committed monthly spend rather than experiments, and the residential pool requires a use-case review before it opens. If your project is a weekly script, the entry tier is more contract than you need.

• Decodo

  Paid

  The provider formerly called Smartproxy, offering residential, ISP, mobile and datacenter proxies through a self-serve dashboard.

  Suits

  Self-serve signup and a smaller minimum than the enterprise tier vendors, which makes it a common landing spot for teams crawling steadily but not at scale.

  Does not suit

  The rename means a large body of tutorials, code samples and Stack Overflow answers still say Smartproxy, so searching for an integration problem returns results written against the old brand and the old dashboard. Budget time for that.

• SOAX

  Paid

  A residential and mobile proxy network with filtering down to country, region, city and carrier.

  Suits

  The targeting filters are unusually fine, which matters when you need results as they appear on one mobile carrier in one city rather than nationally.

  Does not suit

  Mobile exits rotate on the carrier's schedule, not yours, so a session can die between two steps of a flow you assumed was atomic. Anything stateful needs retry logic that can resume rather than restart.

• NetNut

  Paid

  A residential proxy network built on ISP partnerships rather than a consumer peer-to-peer app.

  Suits

  Static residential exits that stay up for long sessions, which suits crawls that authenticate once and then page through results for hours.

  Does not suit

  An ISP-sourced pool is deeper in the countries where those partnerships exist and thin everywhere else, so the long tail of small markets is weaker than a consumer-sourced pool of the same headline size.

• IPRoyal

  Paid

  Residential, datacenter, ISP and mobile proxies sold with pay-as-you-go bandwidth rather than a monthly floor.

  Suits

  You can buy a small amount of residential bandwidth without a subscription, which makes it practical to test whether a target actually blocks datacenter traffic before committing to anything.

  Does not suit

  The self-serve model comes with self-serve support, so a network-level problem is yours to diagnose. Teams that need an account engineer on a bad day should look at the enterprise vendors instead.

• Webshare

  Free tier, then paid

  A self-serve proxy service concentrated on datacenter and static ISP exits, with residential available separately.

  Suits

  A permanent free plan of ten proxies with no card required, which is enough to wire up and debug proxy handling in a crawler before you pay for anything.

  Does not suit

  Datacenter IPs live in ASNs that bot mitigation blocklists by default, so the cheap tier that makes this attractive is also the tier that gets a 403 from any site with a vendor in front of it. Test your actual targets, not a generic IP echo endpoint.

• Rayobyte

  Paid

  Datacenter, ISP and residential proxies from a US-based provider, sold with per-IP as well as per-gigabyte options.

  Suits

  Buying dedicated datacenter IPs outright, priced per IP rather than per byte, suits a heavy crawl of a site that does not block on network reputation at all.

  Does not suit

  Per-IP pricing is only a saving while the IPs work. On a target that starts blocking, you own a fixed block of burnt addresses rather than a rotating pool, and swapping them is a support ticket rather than a config change.

• Infatica

  Paid

  A residential and mobile proxy network that also sells datacenter exits and a scraping API.

  Suits

  Country-level residential coverage at a lower entry price than the top tier vendors, with rotating and sticky sessions both available.

  Does not suit

  A mid-sized pool means more repeat exits on a long crawl of one host, which is exactly what per-IP rate limiting catches. Spread the request rate or expect a rising share of 429s as the job runs.

• Evomi

  Paid

  A Swiss proxy provider offering residential, mobile, datacenter and static ISP exits.

  Suits

  Low per-gigabyte entry pricing and Swiss data handling, which is the deciding factor for some European teams regardless of anything technical.

  Does not suit

  A younger and smaller network than the established vendors, so coverage outside the main markets is thinner and there is less public writing to consult when an integration misbehaves.

• Massive

  Paid

  A residential proxy network sourced through an SDK that pays app publishers for opted-in bandwidth.

  Suits

  The consent model is documented in public, which matters if your legal review asks where the exit IPs come from and expects an answer you can show someone.

  Does not suit

  Bandwidth sourced from consumer devices means throughput and latency vary with whatever else that device is doing. Anything latency sensitive, such as a browser lane with a hard page-load budget, will see a wider spread than a datacenter route.

• PacketStream

  Paid

  A bandwidth marketplace where residential users sell spare capacity and buyers route traffic through it.

  Suits

  Among the cheapest residential bandwidth available, which suits bulk fetches of static pages where a failure just means retrying the URL later.

  Does not suit

  Exit quality is whatever the marketplace has online at that moment, with no targeting precision or uptime commitment. Do not build a scheduled job with a deadline on it, and do not expect a support answer inside an hour.

• Proxy-Cheap

  Paid

  A low-cost provider selling residential, mobile, ISP and datacenter proxies, mostly in small units.

  Suits

  Small purchases are genuinely small, so testing whether a mobile exit changes a site's response costs about the price of lunch rather than a monthly plan.

  Does not suit

  Cheap capacity is shared capacity, and the same address ranges reach targets through many customers. On a popular target, some of your pool is already rate limited before your first request.

• spider.cloud

  Free tier, then paid

  Residential, mobile and ISP exits from its own peer pools, peers.spider.cloud and spider.cloud/shield.

  Suits

  One balance covers the proxy, the browser and the crawl API.

  Does not suit

  Targeting stops at the country, and pricing is in credits rather than per gigabyte.

• Nimble

  Paid

  A data collection platform that bundles its residential proxy network with browser and parsing services rather than selling the proxy alone.

  Suits

  The proxy, the browser and the extraction step are one product, so a site that needs all three does not need three integrations and three sets of credentials.

  Does not suit

  You are buying the platform. If you already run your own browser fleet and only want exit IPs, most of what you pay for is machinery you will not call.

How to read this list

A proxy changes one thing about your request: the address it arrives from. That fixes blocks that are decided by network reputation and nothing else. It does not fix a 403 driven by your TLS fingerprint, a 429 driven by request rate, or a page that renders empty because it needs JavaScript you never executed.

So the first question is not which provider. It is whether the target blocks you on the IP at all. Fetch the URL from a plain server and from a home connection, compare the status codes and the byte counts, and only reach for a paid network when those two differ. How a crawler resolves a page sets out the order those steps run in, including the cases where the answer is a server rule rather than a network block.

The two pricing shapes

Residential and mobile bandwidth is priced per gigabyte because the provider is paying for a borrowed consumer connection. Datacenter capacity is priced per IP or per month because the provider owns the hardware. That difference decides more crawl budgets than pool size does.

Per-gigabyte pricing means your bill tracks page weight. A product page with six hero images costs many times its own HTML, and none of those bytes are the data you came for. Blocking images, fonts, stylesheets and analytics at the browser level is the largest single saving available on a residential crawl, and it usually costs nothing in output quality.

Per-IP pricing means your bill is fixed and your risk is not. The addresses are yours until a target blocks them, and then they stay yours.

What a proxy will not do

Bot mitigation reads the TLS handshake, the ordering of HTTP/2 frames, the header casing and the browser fingerprint. All of that is identical whether your request leaves from a rack in Virginia or a flat in Lisbon. A crawl that fails on fingerprint fails from every IP you can buy.

Worth separating from the technical question: identifying honestly is a different activity from evading detection, and the crawlers that do the first have documented user agents you can allow or deny on purpose. See Googlebot and GPTBot for what that looks like, and cloaking for what happens when a site serves one thing to a verified crawler and another to everyone else.

Related

Crawlers covers the software that walks a site. Scraping APIs covers the services that bundle proxy, browser and parsing into one call, which is usually the cheaper answer when a target is hard. Browser APIs covers the rendering layer. Scraping routers covers the layer where a service chooses a backend per request, which is also where you stop knowing which address served a page. Search APIs covers querying an index rather than fetching pages yourself, and the SERP group there runs on this same network layer.