Leasing IPv4 Addresses vs Buying Proxy Plans

Buying addresses because you want to look like many different places? Do not lease. A proxy plan gives you more prefix diversity per dollar than any block you could lease, and diversity is the property that decides whether you get blocked.

Buying addresses because you need to own your egress identity for the long term, for outbound mail, a partner allowlist you control, a compliance requirement, or infrastructure you are going to run for years? Lease, or buy. That is what the market exists for, and no proxy plan substitutes for it.

Somewhere in between, running a real product that needs both? Most teams that get this right buy proxy capacity for collection and lease or own a small block for the handful of services that need a permanent, reputation-managed address. Those are two separate purchases with two separate justifications.

The mistake worth naming up front: leasing a /24 because proxy plans feel expensive. A /24 is 256 addresses inside a single prefix, and a single prefix is a single point of failure against any target that bans by prefix.

A lease transaction hands you three things and no more.

The right to use a block of addresses. Usually a /24, which is 256 addresses, sometimes a /23 or /22. Smaller than a /24 is generally useless because most transit providers will not accept a route announcement longer than /24.

A Letter of Authorisation. An LOA is a signed document saying you may announce that block. It is a PDF. It is not connectivity.

Usually a Route Origin Authorisation. An ROA is the cryptographic record in the RPKI that says a specific AS number is allowed to originate that prefix. Increasingly, transit providers will drop your announcement without a valid one.

What you do not receive: a router, transit, a server, a proxy daemon, geolocation records, reverse DNS, an abuse desk, or a reputation. Every one of those is your project now. A lease is an entry in a registry plus permission to use it, and the gap between that and a working exit IP is where the entire cost of this option lives.

Pricing works per deal, not per rate card. Lease marketplaces publishing openly in September 2026, IPXO and InterLIR among them, show /24 blocks in a band of roughly $0.35 to $0.60 per address per month, varying by RIR region, block reputation and term, with APNIC space at the top of the range because of scarcity there. Treat that as a market description. Get your own quotes, because the number that matters is the one on your contract, and blocks with clean histories cost more than blocks without.

You have a block and a PDF. Now it has to reach the internet. There are three ways, and picking one determines your reputation, your geolocation and your timeline far more than the lease price does.

Read the right-hand column of the first row carefully, because that is the path most people actually take and it quietly deletes the reason they were leasing. If your hosting provider announces the block from their AS, your leased addresses geolocate to that provider and carry that provider's ASN classification. To a target running ASN-based filtering, they are datacenter addresses from a hosting network, which is exactly what a cheap datacenter proxy is. You have paid ten times more for the same classification. Our explainer on BGP and RIR allocations behind proxy IP origin covers how that classification is derived.

The RIPE figures above are from the RIPE NCC's own billing page, read September 2026, and apply to the RIPE region. ARIN, APNIC, LACNIC and AFRINIC publish their own schedules with different structures. Check the one that covers you.

Normalise everything to one number: dollars per address per month, before any labour.

Three readings fall out of that table, and all three are uncomfortable for the lease option.

A shared datacenter proxy address is roughly one thirteenth the mid-band cost of a raw leased address that does nothing yet. The proxy vendor buys at /16 scale and amortises each address across many customers; you are buying 256 addresses retail.

A dedicated datacenter proxy address at $1 is roughly 2.5 times a mid-band leased address. The difference, about sixty cents per address per month, is what routing, hosting, the proxy daemon, geolocation submissions, abuse handling and a replacement policy cost when somebody else provides them at scale. Almost nobody builds that stack for sixty cents an address.

Amazon's public IPv4 charge is about 120 times Webshare's shared proxy rate, which is the cleanest available illustration that a cloud IP is priced as a scarce resource attached to a server, not as an exit identity. Cloud ranges are also published and trivially classified; see what cloud provider IP ranges are for why that matters before you reach for the obvious shortcut.

For the flat-fee shape, SparkProxy prices by concurrency instead of by address: Starter $75 a month for 100 threads, Core $140 for 250, Boost $240 for 500, Plus $440 for 1000, all unlimited bandwidth and 30 days validity, drawing on 1M+ datacenter IPs across 80+ countries including 50,000+ US addresses. Divide $140 by the pool and per-address arithmetic stops being meaningful, which is the point of that model. How much proxies cost walks through all three shapes.

Reputation. An address's standing is a function of its history and the ASN announcing it, not of who is paying for it this month. Lease a block that a bulk mailer used in 2024 and you inherit their entries on every blocklist that remembers. Before signing, ask for the block's assignment history, then check the range yourself against the major blocklists and fraud-score databases. This is the due diligence step people skip and regret. Our primer on IP reputation and why it matters covers what those checks look for.

Geolocation. Geolocation databases are commercial products maintained by MaxMind, IP2Location, ipinfo and others, plus separate internal datasets at Google, Netflix and the large CDNs. A newly routed block either resolves to nowhere or resolves to wherever it last lived. Correcting it means submitting to each provider separately and waiting out their refresh cycles, which run from days to several weeks, and repeating the exercise whenever you move the announcement. If you lease "US" addresses for geo-specific collection, budget for weeks of them not behaving like US addresses. There is a method for verifying this in our guide to validating IP geolocation accuracy and fraud scores.

Churn. A proxy vendor's pool is not static. Burned ranges get retired and replaced continuously, and the cost of that churn is spread across every customer. Your leased /24 does not churn. When it degrades, it stays degraded until you lease another one, pay for another one, and go through geolocation again.

An abuse desk. Once the block is announced on your behalf or by you, complaints route to the contact on the registry record. Someone has to read them, act inside the upstream's stated window, and keep the relationship intact. Miss that window and the upstream withdraws the route. That is not a theoretical risk; it is the normal operating condition of running address space.

This is the part that decides the comparison for collection workloads, and it is arithmetic rather than opinion.

A /24 is one prefix. Targets that block datacenter traffic almost never ban single addresses; they ban the /24, and frequently the whole originating ASN. So the question is not how many addresses you hold, it is how many independent prefixes you hold.

Lease one /24 and you hold exactly one. A single block decision at a single target takes out 100% of your capacity at that target, instantly, and no amount of rotation inside the block helps because every address in it shares the prefix. Lease four /24s from different sources, which quadruples your cost and your admin, and you hold four.

Compare that against what a proxy vendor's pool looks like when the same ban lands: you lose one prefix out of however many they operate, the health checks route around it, and you very likely never notice. That is the actual product. You are not buying addresses, you are buying dispersion and the vendor's ongoing replacement of burned space. Mixed subnet datacenter proxies covers why dispersion outperforms raw pool size, and proxy pool size claims explained covers how to read the numbers vendors publish.

A blunt way to put it: leasing converts a recurring, amortised, vendor-absorbed risk into a concentrated one you hold alone. For long-lived infrastructure that is fine, because nobody is trying to ban your mail server's prefix. For collection it is the whole problem.

Some teams skip the lease market and go to the source: become a Local Internet Registry, get an AS number, and hold resources directly. It is a legitimate path and the fees are public.

The RIPE NCC's billing page, read September 2026, lists a EUR 1,000 sign-up fee, a EUR 1,800 annual service fee per LIR account, EUR 50 per ASN assignment and EUR 75 per independent assignment or legacy resource. Those are the registry's charges only. They do not include IPv4 addresses, which RIPE no longer has to allocate at any meaningful volume, so you are back in the transfer or lease market for the actual space. They do not include transit, hardware, or the network engineer who runs the BGP sessions.

Realistic timeline from decision to first working exit on your own AS: months, not weeks. You need a legal entity in the service region, the membership approved, the ASN assigned, at least one upstream willing to peer with you, ROAs published, and then the geolocation and reputation work still ahead of you. Against that, a proxy plan issues credentials in minutes.

That timeline is the argument, not the money. EUR 1,800 a year is roughly $160 a month, which is in the same range as a mid-tier proxy plan. The reason most teams should not do it is that the registry fee is the smallest line on the project.

Proxy plans do not cover everything, and there are jobs where leasing or owning is clearly correct.

• Outbound email at volume. You need addresses whose sending reputation you build and control, with your own reverse DNS, SPF alignment and warm-up history. A shared or rotating proxy address is the exact opposite of what deliverability needs.
• A partner or regulator requires a fixed, attributable range. Financial and healthcare integrations frequently demand that you disclose and hold your egress range. You cannot do that on a rotating pool.
• You are already running network infrastructure. If you have an AS, transit and an engineer on staff, the marginal cost of adding a /24 is low and the answer changes.
• Long-lived services, not collection. Anything where the same address should answer for the same service for years benefits from ownership.
• Very large scale. Past a certain size, amortising your own space across your own workloads beats paying a per-address or per-thread margin. That threshold is much higher than most teams estimate.

Notice what is not on that list: scraping, price monitoring, rank tracking, availability checks, ad verification, geo testing. Every one of those wants many short-lived identities across many prefixes, which is the opposite of what leasing produces.

Before you commit either way, three concrete steps.

One: price both sides on your real prefix requirement, not your address requirement. Decide how many independent /24s your workload needs to survive a ban at your worst target. Then price that many leases, plus routing, plus hosting, against a proxy plan. Most people discover their address count was never the constraint.

Two: test whether your target even discriminates. Run a few hundred requests through a cheap datacenter pool and measure the success rate on real content, not on status codes. If a datacenter address works, leasing buys you nothing a datacenter proxy does not already give you at a thirteenth of the price. If it does not work, leased space announced from a hosting AS will not work either, and your answer is residential exits rather than owned addresses.

# Baseline: what does the target return through a rotating datacenter exit?
curl -x http://USERNAME:PASSWORD@gateway.sparkproxy.io:11000 \
     -s -o /dev/null -w "%{http_code}\n" https://example.com/product/123

# Same check with a residential exit, via the Scraping API
curl -s "https://scrape.sparkproxy.io/api/v1?url=https%3A%2F%2Fexample.com%2Fproduct%2F123&premium_proxy=true" \
     -H "X-API-Key: YOUR_API_KEY" -o /dev/null -w "%{http_code}\n"

The first command uses the rotating pool on port 11000, with 11002 for sticky sessions and 13000 for SOCKS5. The second requests a residential exit through the Scraping API using premium_proxy=true, which is the correct parameter for that. A plain fetch costs 1 credit, JavaScript rendering 5, a screenshot or PDF 10, and the free tier is 1,000 credits with no card, which is enough to answer the question.

Three: cost your own time at a real rate. Routing, geolocation submissions, abuse handling and monitoring are not one-time tasks. If the honest ongoing figure is even four hours a month, that alone exceeds the price difference between a leased /24 and a mid-tier proxy plan.

Run those three and the decision is usually obvious. It is also usually not the one people expect when they start reading about the IPv4 lease market.