Proxy pricing explained: per GB vs per IP

Proxy pricing follows two models. Residential and mobile proxies are billed per gigabyte of traffic, so cost scales with how many pages you fetch, how heavy they are and how often you retry. ISP and datacenter proxies are billed per IP address per month, so cost is fixed and the question becomes how many requests each address can carry before a target objects. Choosing well means estimating your gigabytes honestly and knowing where the extra bytes come from.

How per-GB proxy pricing works

On per-GB plans you pay for traffic that crosses the provider's network, regardless of how many addresses it used. Residential and mobile pools are priced this way because the scarce resource is peer bandwidth: every byte goes through someone's home or phone connection.

What counts as a billed byte varies. Ask each provider:

• Are both directions metered, or only responses?
• Are headers and TLS handshakes included?
• Are failed, blocked and timed-out requests billed?
• Is usage rounded per request, per session or per billing period?
• Do prepaid gigabytes expire?

Two providers with the same headline rate can produce different invoices for identical traffic if they answer these differently.

How per-IP proxy pricing works

On per-IP plans you rent specific addresses, usually by the month. ISP and datacenter proxies are priced this way because the address is the resource and server bandwidth is comparatively cheap. Bandwidth may be unmetered or subject to a generous fair-use limit.

Per-IP cost depends on how many addresses you need, which depends on the target, not the provider. If a target starts challenging an address above a certain request rate, you need enough addresses to keep each one below it. Dedicated addresses cost more than shared ones but keep another customer's behavior from affecting your reputation with the target.

How to estimate your monthly gigabytes

Estimate from the workload, not from the provider's calculator:

monthly GB = pages per day × transfer per page × retry factor × overhead factor × days

Transfer per page is what actually crosses the wire, compressed, including every sub-resource your client fetches. Measure it by running a sample of real requests through a local proxy or your browser's network panel and summing transferred bytes.

Here is a worked example with made-up workload numbers. A team monitors 40,000 product pages a day. Their client retries enough that total requests run 20 percent above pages, and they assume 5 percent for headers and handshakes.

The arithmetic for the first row: 40,000 × 120 KB is 4.8 GB, times 1.2 for retries and 1.05 for overhead is about 6 GB a day. The same pages cost eighteen times more in a headless browser that loads everything. The choice of client matters more to a per-GB bill than the per-GB rate.

Where proxy costs hide

Most surprises on a per-GB invoice come from bytes nobody planned for:

• Retries. Every retry is billed traffic. A block rate that doubles on a strict target can double spend on that target without any change in pages collected.
• Headless browser asset loads. Images, fonts, video, analytics scripts and ad calls often outweigh the HTML many times over, and third-party requests travel through the proxy too.
• Compression off. A client that does not send Accept-Encoding, or a custom header set that drops it, receives HTML uncompressed at several times the size.
• Failed requests billed. Challenge pages, error pages and partial downloads before a timeout all count if the provider meters them.
• Redirect chains and consent pages. Every hop is a request. A fresh session that has to pass a consent interstitial on each visit pays for it every time.
• Polling unchanged pages. Re-fetching a page that has not changed since yesterday costs as much as fetching a new one.

Per-IP plans have their own hidden cost: addresses bought for a peak and left idle, and addresses that a target has started to challenge but that are still being paid for.

How to cut bandwidth on per-GB plans

1. Block what you do not parse. In headless browsers, abort requests for images, fonts, media and third-party domains. The Playwright proxy guide shows where routing hooks sit alongside proxy configuration.
2. Keep compression on. Send Accept-Encoding with gzip and, where your client supports it, Brotli.
3. Fetch the data, not the page. Where a site's terms allow it, a JSON endpoint behind a page is often a fraction of the page's size.
4. Reuse sessions. A sticky session keeps cookies and connection state, so consent pages, redirects and handshakes are paid once rather than on every request.
5. Cache static resources that your crawler does need, such as a script it executes, instead of downloading them per page.
6. Use conditional requests. Send If-None-Match with a stored ETag or If-Modified-Since with a stored date, as defined in RFC 9110. An unchanged page returns a small 304 response instead of the full body.
7. Fix the causes of retries. Better session handling and pacing reduce blocks, which reduces billed retries. Measure cost per successful request, as described in how to benchmark proxy providers, so the saving is visible.

Promotions, commitments and spend caps

Read the pricing page for what happens after the first term. Introductory rates that revert to a higher list price at renewal, committed plans whose unused gigabytes expire at the end of the month, and overage rates above the plan rate all change the real cost of a plan. Model twelve months, not one.

When comparing proxy pricing across vendors, send each the same questions and put the answers side by side:

• What is the rate after any introductory period, and when does it apply?
• Is there a monthly minimum or a commitment, and what happens to unused volume?
• What is the overage rate, and is it higher than the plan rate?
• What exactly is metered, and are failed requests billed?
• Can spend be capped per team or project, and is the cap hard or soft?
• What notice is given before a price change?

Spend caps protect against the other failure: a job stuck in a retry loop overnight. Set caps and alerts per team or project rather than per company, so a runaway job stops before it spends another team's budget. Check whether a cap is hard, stopping traffic, or soft, only alerting.

When to move a workload from per-GB to per-IP

Per-IP pricing wins when a workload moves a lot of data through a small number of targets that accept ISP or datacenter addresses. Using the example above: if the target tolerates one request every ten seconds per address, 48,000 requests a day need about six addresses, eight with headroom. The per-GB version of the HTTP-client workload uses about 180 GB a month, so per-IP is cheaper whenever one address-month costs less than roughly 22 GB of per-GB traffic. For the headless workload the threshold rises to over 400 GB per address, which almost any per-IP price clears.

Consider moving a workload when:

• Its volume is steady and its targets are few and known.
• Pages are heavy, particularly when a headless browser is required.
• The targets accept ISP or datacenter addresses at your request rate, confirmed by a test.
• A partner needs to allowlist your egress addresses.
• Finance wants a fixed monthly figure.

Stay on per-GB when traffic is bursty, spread across many countries, aimed at strict targets that challenge non-residential ranges, or too small to fill an address. The comparison of residential vs ISP vs datacenter proxies covers how targets treat each type.

How ProxyForge prices proxies

ProxyForge publishes flat prices with no introductory rates that revert. Residential and mobile are billed per GB, ISP and datacenter per address per month, all from a prepaid wallet with no monthly minimum, starting from 1 GB or a single address. ISP and datacenter have a paid trial, and volume customers can arrange net-30 invoicing through sales.

Current rates are on the pricing page, with detail for residential pricing and ISP pricing.