Fair Use Policy

A Fair Use Policy (FUP) defines how much load a proxy service permits and what happens when a customer's usage exceeds that allowance or starts affecting the network. It can govern data volume, transfer speed, simultaneous connections, and sustained activity. It forms part of the Provider Policies that determine how an advertised plan actually works.

In the second half of 2026, unlimited proxy plans include several very different arrangements: continued access at a lower speed, fewer connections after a traffic threshold, additional charges, or a conversation with an administrator about moving to a larger plan. A proxy can remain online while becoming too slow for the job—or while the bill keeps growing.

What unlimited traffic actually buys

For someone buying proxies, unlimited sounds like permission to transfer as much data as the software can handle. The product may instead provide a fixed monthly price for a particular level of use, with a different service regime beyond it.

Published offers include:

• Unmetered access: transferred gigabytes do not determine the bill. Available capacity and other service rules still apply.
• Traffic with a fair-use allowance: a specified volume is available before speed or concurrency is reduced.
• An allowance followed by usage charges: access continues, but additional traffic becomes a separate expense.
• Time and speed packages: the customer buys access for a period at a selected transfer rate, without a separate charge for each gigabyte.

Some services publish no fixed data ceiling and explicitly advertise unmetered, unthrottled access. Others reserve the right to intervene when usage disrupts their network. A secret monthly cap is therefore not a necessary feature of every unlimited product. What matters is which arrangement the purchase includes.

The word bandwidth can add confusion: sellers use it for both data volume and transfer speed. The units reveal the difference.

A plan can allow unlimited data while limiting connections from the first minute. It can also allow thousands of connections while imposing a monthly data threshold. Those promises concern different resources.

Why a large IP package can still trigger fair use

A customer purchases 50,000 IP addresses and starts transferring large amounts of data through the entire package. From the customer's perspective, every address was paid for. From the service's perspective, that account can create an exceptional load on the infrastructure serving those addresses.

This is an operating pattern reported by a proxy-service operator. Administrators identify customers sustaining unusually high throughput, contact them, and ask them to reduce the load or move to a more expensive plan. Immediate blocking is not the first response. Suspension follows if the customer ignores the warnings; according to the operator, most affected customers choose the upgrade.

That model does not need a fixed rule such as “100 GB and you are out.” The decision follows the intensity of the workload and the capacity the provider is prepared to supply at that price.

The explanation is physical as well as commercial. Many IP addresses can share servers and network links. Buying additional addresses increases the available address inventory; the plan determines how much extra processing and transfer capacity comes with them. A dedicated IP likewise gives exclusive use of an address without automatically providing a dedicated server or network port.

This is where expectations collide: a buyer may treat a large package as permission to run every IP at maximum speed continuously, while the price assumes a mixture of busy and idle addresses.

Connection limits can apply before any FUP warning

A normal plan limit and a restriction imposed after heavy usage can produce similar symptoms, but they occur for different reasons.

In the operator example above, packages below 1,000 IPs allow 1,500 simultaneous connections, while a 1,000-IP package allows 3,000. These are examples from that service's schedule. Larger packages follow their own tiers; another provider may assign limits per IP, per order, or per proxy login.

“Threads” is a common sales term for this capacity. It does not necessarily equal the number of browser profiles or worker threads in a program. A page can open multiple connections, and an application can reuse connections for several requests. A sticky session, which concerns retaining an exit IP, is another separate concept.

Consequently, launching more workers does not create more purchased capacity. Once a connection ceiling is reached, additional work may wait, fail, or be rejected, depending on the service. Increasing the number of IPs will only help if the relevant connection allowance also increases.

What changes after a fair-use threshold

Fewer simultaneous connections

Published dedicated datacenter terms include a reduction from 100 to 10 concurrent sessions per IP after 100 GB of monthly traffic on that IP. Other per-IP products use a 50 GB threshold. These are product-specific examples, rather than an industry-wide allowance.

Another published ISP plan gives a more dramatic example: a 1,000-IP package with 5,000 concurrent sessions falls to 10 sessions when its 25 TB fair-use threshold is reached. The reduced allowance lasts for the rest of that subscription period.

The addresses can still connect, and a simple proxy checker may report them as working. Individual requests can remain fast; the problem is how few can now run together. For a job built around thousands of simultaneous requests, the original completion schedule may become impossible.

Lower transfer speed

Speed throttling lets requests continue while reducing how much data moves each second. One published policy describes normal speeds of 300–400 Mbps, an initial reduction to 100 Mbps, and further reductions reaching 10 Kbps with continued heavy usage. These are the provider's stated operating figures, not a guaranteed speed test result.

At such a low final rate, a proxy may remain technically reachable while ordinary pages and downloads become impractical. The distinction matters when an unlimited offer promises continued access without promising unchanged performance.

Additional charges

Some plans keep the original service running and charge for traffic beyond the allowance. One published policy sends alerts at 85% and 100%, then bills the excess at the applicable pay-as-you-go rate. Adding more IPs later does not cancel overage already incurred under that policy.

Continued operation can therefore be the expensive outcome. A script left running overnight may finish its work and also produce an unexpected charge. This is usage billing; automatic replenishment of the account balance is a separate payment setting.

Whose allowance is being consumed?

“100 GB per IP” can describe different accounting rules.

With an individual allowance, one heavily used address can be restricted while the others retain their original capacity. With an order-wide pool, the number of purchased addresses determines the total allowance, which can be distributed unevenly.

A documented pooled model gives an order containing 2 IPs a total of 200 GB. Using 100 GB through one address and nothing through the other leaves another 100 GB available to the order. An account-wide or product-wide pool can work across several orders or zones instead. The scope has to come from that product's rules.

Shared addresses need not share a traffic allowance: usage can be attributed separately to each buyer. Network congestion and billing attribution are separate matters.

Some thresholds have to be reached together

A further published model combines 10 TB of total traffic with an allowance calculated at 25 GB for each purchased IP. Both conditions apply. Its examples put a 100-IP package at the 10 TB threshold and a 1,000-IP package at 25 TB. Reading only the smaller number produces the wrong expectation.

The same provider publishes a 100 GB-per-IP figure for certain dedicated products alongside the 10 TB condition. Even within one service, a number taken from a shared plan can misdescribe a dedicated plan.

Rotating or replacing an exit address does not create new allowance when the meter belongs to the order or account. Similarly, creating another proxy login does not multiply a shared subscription limit.

Why the traffic counter grows faster than the results folder

Proxy traffic can include both the request and the response: headers, uploaded data, page content, scripts, images, and other downloaded resources. A scraper might save a few price fields while its browser has transferred an entire product page.

The arithmetic shows how much the workload matters. At an illustrative average of 100 KB transferred per exchange, 1 million exchanges amount to 100 GB. At 3 MB per exchange, the same number amounts to 3 TB. These are calculations using assumed average sizes, not measurements of particular websites.

Repeated requests and downloaded error or CAPTCHA pages can add traffic without producing useful records. Their billing treatment depends on the product: a raw byte-metered proxy and an API that charges for successful results can use different rules. Failed work is therefore not automatically free traffic.

Long downloads, video, unnecessary page resources, and repeated retries can dominate usage. Loading fewer resources when the task permits it and avoiding endless retries directly reduce the amount transferred. Those changes address the cause of consumption; changing the exit IP does not.

A speed limit also sets a physical ceiling

Even without a monthly data cap, a fixed transfer rate limits what can pass in a month.

These calculations assume continuous transfer, decimal units, and no downtime or protocol overhead. Real useful throughput will be lower. If a speed allowance applies to the whole package, adding addresses does not multiply that allowance.

Units matter here too: 100 Mbps equals 12.5 MB/s before overhead. GB and GiB are different volume units: 1 GB is 1,000,000,000 bytes, while 1 GiB is 1,073,741,824 bytes. A counter using a different unit can disagree numerically without traffic having disappeared.

Mobile and residential plans can treat sustained activity differently

A mobile proxy depends on a modem, the cellular network, and the upstream operator's conditions. A service may be protecting its SIM connections and the capacity of a cell tower as well as its own servers.

One mobile policy sets 20 GB per day for Germany and a 50 GB-per-day guideline for the UK and US. The latter permits occasional heavier days but can restrict sustained excess.

Some mobile plans also prohibit continuous automated traffic, including 24/7 scraping. A low monthly total does not cancel that condition.

This is a question of product suitability. Other proxy offers explicitly support continuous workloads, including time-based residential plans sold with selected speeds. Automated activity or 24/7 operation is not universally a FUP violation, and residential access does not have one universal 10 TB ceiling.

For the buyer, the distinction is substantial: a plan suitable for periodic account activity may be unsuitable for uninterrupted collection, even if both tasks reach the same websites.

How the provider detects heavy usage

Fair-use enforcement can be automatic, manual, or a combination. Useful measurements include transferred bytes, current throughput, open connections, connection duration, and how long an account maintains its load. A short burst and hours of sustained transfer place different demands on capacity.

The 50,000-IP example shows that an administrator can make the decision after observing usage. Other services apply a published threshold automatically. There is no need for every provider to use deep packet inspection to count traffic or identify a heavily loaded account.

For HTTPS carried through a normal CONNECT tunnel, application content remains encrypted between the client and destination. The proxy can count bytes and connections without reading page contents or the destination's HTTP status codes. A service that terminates TLS or fetches pages through its own scraping API has different visibility and can use application results, such as 4xx and 5xx responses, in its monitoring.

A website block is a different event

A destination can reject requests while the proxy subscription remains within all its allowances. Conversely, the provider can throttle an account even when the destination accepts every request.

Slow transfers alone do not establish that FUP has been applied. Neither does an HTTP 429 response identify the responsible service or the exact limit. A destination's request-rate control, a proxy connection ceiling, and a monthly traffic restriction are different explanations.

Repeatedly retrying a rejected workload can still create additional load. This can turn a website-access problem into a separate usage problem, without making website rejection itself proof of a FUP breach.

When normal capacity returns—and what an upgrade changes

Restrictions tied to accumulated traffic commonly last until the allowance resets or additional capacity is purchased. The reset can follow a calendar month, a 30-day period beginning on the purchase date, or another subscription term. A daily guideline can follow a different schedule again.

An overnight pause therefore has different effects in different models. It reduces current load, which may resolve a manual capacity concern. It does not erase the traffic already counted in the current billing period.

Top-ups also have details that affect their value. A published option adds 100 GB for every IP in the order: choosing it on a 10-IP order adds 1 TB, rather than 100 GB shared across that order. The extra data expires with the current cycle and does not roll over. A late-cycle purchase can consequently have a short useful life.

An upgrade can buy more IPs, a larger data allowance, more connections, a higher transfer rate, or a negotiated service arrangement. Its usefulness depends on which resource caused the restriction. More addresses alone will not resolve a speed ceiling that remains unchanged.

For plans with overage billing, an email alert reports usage; it does not stop the software. A configured spending or traffic stop serves a different purpose. Some controls are checked periodically—for example, a published zone limit is evaluated every 15 minutes—so even a stop setting can allow additional usage before it takes effect.

Warnings, suspension, and refunds

There is no universal enforcement sequence. A service can warn and negotiate an upgrade, apply a lower service tier automatically, temporarily suspend access, or end the commercial relationship. Published terms also differ on whether advance notice is promised.

Refund outcomes differ just as sharply. Some policies state that breaches can result in termination without a refund. Another published mobile policy provides for communicating with the customer, returning payment for unused days, and withdrawing access when excessive use threatens the network. A FUP restriction does not have one industry-wide rule under which every remaining balance disappears.

FUP also overlaps with, but has a different purpose from, an Acceptable Use Policy. Fair use addresses resource consumption; acceptable use addresses what customers are permitted to do. An otherwise permitted collection task can exceed its plan's capacity. A prohibited activity can breach service rules while consuming very little data.

Clear terms make the purchased service understandable: the measured resource, the scope of the counter, the trigger for intervention, the consequence, and the route back to normal service. Phrases such as “excessive use” leave more of that decision to the provider, especially when no numeric threshold is offered. An undisclosed threshold makes the usable service harder to price: the headline monthly fee does not show how much work can be completed at the original performance level.

A short trial may confirm that the IPs connect and reach the required websites. It cannot, by itself, show what happens after weeks of sustained traffic. The long-term value of an unlimited proxy plan depends on the capacity it preserves as usage grows—and the price of continuing beyond it.