How many proxies do I need?
Count two things separately: the IP addresses that must stay distinct (one per account or identity you keep apart) and the requests that must run at the same time (sessions, threads or ports). The box below shows that plans state the first, the second, both or neither, so match each number to the field the plan actually publishes.
What the 863 plans with a price say about capacity, checked up to Nov 17, 2026:
- 162 say how many IP addresses you get.
- 412 say how much can run at once: concurrent sessions, threads or ports.
- 95 say both, and 384 say neither; 51 of those are sold by traffic, where the provider's pool supplies the addresses, and 312 by the IP, one address for each you buy.
A limit a plan does not state is unknown, not unlimited: ask the provider before you size a job on it.
Answer · The comparison uses prices checked up to Nov 17, 2026; every figure in the text keeps the date it was checked.
On this page
“Proxy” means different units on different price lists. A provider may sell you assigned IP addresses, a rotating gateway, ports, threads or simultaneous sessions, and a number on one list is not the same thing as the same number on another.
Count fixed IPs by the identities you keep apart
If each account, profile or identity must keep its own stable address, start from the peak number that must not share one:
IPs needed = identities that must never share an address at the same time
Add a few spares if the provider replaces addresses slowly. This tells you how many addresses to buy, not how many requests each can carry: check the plan’s session or thread limit for that.
Size rotating access by parallel work
With a rotating gateway you do not buy addresses; the provider picks one per request or session. Size it by your peak parallel requests, then compare that with the plan’s published concurrency, threads or ports.
parallel requests = requests per second × seconds each request takes
For example, 100,000 pages a day at 2 seconds a request is about 100,000 × 2 ÷ 86,400 ≈ 2.3 requests at once, if the work is spread over the whole day. Fetch the same pages in an 8-hour window and it is three times that, about 7; leave room for retries and slow pages, so a plan allowing 10 concurrent sessions or threads covers it. A plan of “100 threads” may say nothing about how many addresses sit behind it, and a pool of millions of addresses says nothing about how many requests you may run at once.
Most rotating plans are priced by traffic. Estimate the gigabytes your job moves in a month and compare those in the residential comparison; how much bandwidth does web scraping use helps with the estimate.
If the plan does not say, ask
Before buying, write down four numbers: identities that need their own address, peak simultaneous requests, how long a session must keep the same address, and monthly traffic. Match each to the plan’s IP count, concurrency or threads, sticky session length and traffic allowance. A limit the plan does not publish is unknown, not unlimited: ask the provider, or choose a plan that publishes it. Concurrency limits and thread limits explain the difference.
Price it for your amount
Checkout totals, every billing type competing. Change the amount to recalculate.
Geonode
$34.65Residential monthly subscription · Monthly
Checked Nov 17, 2026 · $0.69 per GB used
Buy at Geonode →GB and IP packages use separate amounts and billing units.