Proxy Pricing Explained: Per-GB, Per-IP, Per-Port, and More

Per-GB pricing is the dominant model for residential proxies. You pay for the total data transferred through the proxy — both request data sent and response data received. Every byte counts toward your allocation.

How it works: You purchase a plan with a bandwidth allocation (e.g., 50GB/month at $8/GB = $400/month). Each request you send through the proxy consumes bandwidth proportional to the size of the response. A typical HTML page is 50-200KB. A page with images might be 2-5MB. API responses are usually 1-50KB.

Typical pricing ranges:

• Low volume (under 10GB/month) — $10-15/GB. Small plans carry a premium because the provider's fixed costs are spread across less usage.
• Mid volume (10-100GB/month) — $6-10/GB. The sweet spot for most businesses. Volume discounts start appearing.
• High volume (100GB+/month) — $2-6/GB. Enterprise-scale pricing with significant volume discounts. Often requires annual commitment.

Best for: Variable workloads where request volume fluctuates week to week. You pay for what you use (within your plan tier), and bandwidth consumption correlates roughly with work accomplished.

Watch out for: Failed requests still consume bandwidth. If a request returns a block page or error, you paid for that data transfer without getting useful results. A provider with a 70% success rate effectively costs 30% more than their stated per-GB rate because nearly a third of your bandwidth produces nothing usable.

Per-IP pricing charges a fixed monthly fee for each proxy IP address you control. This model dominates the datacenter proxy market because datacenter IPs are static, dedicated resources that the provider allocates specifically to you.

How it works: You purchase a set number of IPs (e.g., 100 IPs at $2/IP/month = $200/month). Each IP is yours exclusively for the billing period. You can send unlimited requests through each IP — there is no bandwidth cap. The "per-port" variation is functionally identical; each "port" maps to a unique IP address.

Typical pricing ranges:

• Shared datacenter proxies — $0.50-1.50/IP/month. Multiple customers share the same IPs. Lower trust, higher risk of pre-existing blocks.
• Dedicated datacenter proxies — $1.50-3.00/IP/month. The IP is exclusively yours. Clean reputation, no other users causing blocks.
• Premium datacenter (ISP proxies) — $3-10/IP/month. Datacenter-hosted IPs registered under residential ISP ranges. Higher trust than standard datacenter.

Best for: Use cases requiring consistent IP identity — account management, social media profiles, long-running sessions where you need the same IP repeatedly. Also excellent when you transfer large data volumes per request, since there is no bandwidth charge.

Watch out for: You pay the same whether you use the IP heavily or barely at all. If you purchase 200 IPs but consistently use only 50, you are paying for idle capacity. Also, datacenter IPs are more easily detected than residential, so a low per-IP cost may come with lower success rates.

Per-request pricing is a newer model where you pay a fixed amount per successful request rather than per gigabyte of data or per IP. It is most common with proxy APIs and scraping-as-a-service products that handle proxy rotation, retries, and rendering for you.

How it works: You submit a URL to the provider's API. They handle the proxy selection, rotation, retries, and optional JavaScript rendering. You pay only when you receive a successful response. Typical pricing: $0.001-$0.01 per request for standard pages, $0.01-$0.05 for JavaScript-rendered pages, and $0.005-$0.02 for search engine results pages.

Advantages:

• Predictable per-unit cost — You know exactly what each data point costs. No hidden bandwidth consumption from retries, CAPTCHAs, or failed requests.
• No infrastructure management — The provider manages proxy rotation, fingerprinting, and retry logic. You send a URL, you get data back.
• Success-based billing — You are not charged for failed requests (in most implementations). This eliminates the hidden tax of failed requests consuming bandwidth.

Disadvantages:

• Higher per-unit cost at scale — Compared to self-managed proxies with per-GB billing, per-request pricing is typically 3-10x more expensive at high volumes.
• Less control — You cannot choose which IP to use, configure session behavior, or optimize the request pipeline. The provider's black-box approach works or it does not.
• Vendor lock-in — Switching from a per-request API to self-managed proxies requires rebuilding your scraping infrastructure.

Per-request pricing makes sense for teams without proxy engineering expertise, low-to-moderate volume operations (under 100,000 requests/month), and use cases where development time is more expensive than per-request premiums.

Some providers offer flat-rate subscription plans with "unlimited" bandwidth or requests for a fixed monthly fee. The appeal is obvious: predictable costs regardless of usage. The reality is more nuanced.

How subscription plans actually work:

No provider can offer truly unlimited residential proxy bandwidth at a flat rate. Residential bandwidth has a real cost — the provider pays their SDK partners for every gigabyte routed through their network. "Unlimited" plans always have hidden constraints:

• Concurrent connection limits — You can use unlimited bandwidth, but only through 50-200 simultaneous connections. This effectively caps your throughput.
• Fair use policies — Terms of service include a "reasonable use" clause that lets the provider throttle or terminate accounts that exceed undisclosed thresholds.
• Speed throttling — After a certain usage level, connection speeds degrade. You technically have unlimited bandwidth, but at speeds that make high-volume scraping impractical.
• IP pool restrictions — Unlimited plans may access a smaller subset of the provider's IP pool, reducing geographic coverage and increasing detection risk.

When subscription plans work: If your usage is genuinely moderate and consistent — say 30-80GB/month — and the subscription price is less than the equivalent per-GB cost, the flat rate saves money and simplifies budgeting.

When they do not: If you plan to use 500GB+ per month, the provider will enforce limits. You will either hit connection caps, experience throttling, or receive a notification that your usage exceeds fair use. For heavy workloads, transparent per-GB pricing with volume discounts is more reliable than "unlimited" promises.

Most proxy providers offer a choice between flexible monthly plans and committed (annual or quarterly) plans with volume discounts. The discount for commitment is typically 20-40%, which is significant at scale — but the tradeoff is reduced flexibility.

Pay-as-you-go (monthly) plans:

• Cancel or change plan tier any month
• Scale up or down based on seasonal demand
• Higher per-GB or per-IP rates (no volume discount)
• Best for: new projects with uncertain usage, seasonal businesses, teams still evaluating providers

Committed (annual) plans:

• 20-40% lower per-unit pricing than monthly
• Locked into a minimum spend for the contract duration
• May include additional perks: dedicated support, custom IP pools, priority routing
• Best for: established workflows with predictable monthly usage, teams that have already validated the provider through a trial

Hybrid approach: Some providers allow you to commit to a base volume at discounted rates and pay overages at standard rates. This gives you the discount on predictable usage while maintaining flexibility for spikes. If offered, this is usually the optimal structure.

Decision framework: If your monthly proxy spend has been consistent (within 20% variance) for 3+ months and you are satisfied with the provider, switch to an annual commitment for the discount. If your usage varies by more than 50% month-to-month, or you have been with the provider for less than 3 months, stay on monthly billing until your usage stabilizes.

Proxy pricing reflects the underlying cost structure and trust value of each proxy type. Understanding why prices differ helps you evaluate whether a price is fair or inflated.

Cost hierarchy (most to least expensive per GB):

Mobile > Residential > ISP > Datacenter

Why residential proxies cost $5-15/GB:

• Sourcing cost — Residential IPs come from real users who opt in through SDK partnerships. The provider pays revenue share to these app partners for every gigabyte routed through their users' connections. This is a per-GB cost that scales linearly with usage.
• Infrastructure cost — Managing millions of peer connections from residential devices (which go online and offline unpredictably) requires sophisticated network infrastructure.
• Trust premium — Residential IPs access websites that block other proxy types. You pay for this access advantage.

Why datacenter proxies cost $1-3/IP:

• Low sourcing cost — IPs are purchased or leased in bulk from data centers. No revenue share to end users.
• Simple infrastructure — Static servers in data centers are easy to manage compared to millions of residential peer connections.
• Lower trust value — Websites can identify and block datacenter IP ranges, reducing their access utility.

Why mobile proxies cost $20-40/GB:

• Limited supply — Mobile IPs are scarcer than residential or datacenter IPs.
• Highest trust level — Mobile carrier IPs are shared via CGNAT, making them very difficult to block without affecting real users.
• Higher sourcing cost — Mobile bandwidth is expensive, and carriers monitor for abuse.

When a provider charges significantly below market rates for a proxy type, investigate how they achieve the discount. Often, it comes at the expense of pool size, success rate, ethical sourcing, or support quality.

The sticker price on a proxy plan is not your actual cost. Several hidden factors inflate what you truly pay per unit of useful data collected.

Failed requests consuming bandwidth: This is the largest hidden cost on per-GB plans. When a target website serves a CAPTCHA page, a block message, or a redirect instead of the requested content, you still paid for the data transferred. A 50KB CAPTCHA page costs the same bandwidth as a 50KB data page — but delivers zero value. At a 10% failure rate, you are effectively paying 10% more per GB of useful data.

Minimum commitments and rollover policies: Many plans require a minimum monthly spend or a minimum bandwidth purchase. If you buy 50GB but only use 30GB, you paid for 20GB of unused capacity. Check whether unused bandwidth rolls over to the next month — most providers do not offer rollover, which means your minimum commitment is a "use it or lose it" cost.

Overage charges: Exceeding your plan allocation triggers overage rates that are typically 1.5-3x the base per-GB rate. A plan at $8/GB might charge $15/GB for overages. If your usage is unpredictable, overages can spike your bill significantly. Some providers auto-upgrade your plan, others charge overages, and others cut off access when you hit your limit. Know your provider's policy before you get surprised.

Retries and redirect chains: If your scraping workflow follows redirects or retries failed requests, each attempt consumes additional bandwidth. A request that redirects three times before reaching the target page consumes 3-4x the bandwidth of a direct request. Optimize your request pipeline to minimize unnecessary data transfer.

Full page loads vs HTML only: Loading images, CSS, JavaScript, and fonts through the proxy dramatically increases bandwidth consumption per page. If you only need HTML content, configure your client to request only the HTML document. This can reduce per-page bandwidth by 80-90%.

The metric that matters is cost per successful request — the total amount you spend divided by the number of requests that returned usable data. This single number lets you compare providers on equal footing regardless of their pricing model.

The formula:

Cost per successful request = Total monthly proxy spend / Number of successful requests

Example calculation:

You spend $500/month on residential proxies at $10/GB (50GB plan). Your scraping operation sends 250,000 requests per month. Average successful response size is 150KB. Success rate is 92%.

• Successful requests: 250,000 x 0.92 = 230,000
• Bandwidth from successful requests: 230,000 x 150KB = 34.5GB
• Bandwidth from failed requests: 20,000 x 30KB (avg error page) = 0.6GB
• Total bandwidth used: 35.1GB of your 50GB plan
• Cost per successful request: $500 / 230,000 = $0.0022

Now compare against a different provider: $600/month at $8/GB (75GB plan), 97% success rate, same request volume.

• Successful requests: 250,000 x 0.97 = 242,500
• Cost per successful request: $600 / 242,500 = $0.0025

The "cheaper" per-GB provider ($8 vs $10) actually costs more per successful request ($0.0025 vs $0.0022) in this example because of the higher success rate and plan structure of the first provider. Always run this calculation with your actual data before choosing based on per-GB price alone.

Proxy spend is an investment, not a raw expense. The data you collect through proxies has a quantifiable business value that should far exceed the proxy cost. Framing proxy costs as ROI helps justify the investment to budget holders.

Direct value calculation examples:

• Price monitoring — If competitive pricing intelligence helps you optimize prices by 2% on $5M in annual revenue, that is $100,000 in value. Proxy cost for the monitoring might be $6,000-12,000/year. ROI: 8-16x.
• SEO rank tracking — If monitoring 10,000 keywords identifies ranking drops that would have cost $50,000 in organic traffic before you fixed them, the proxy cost of $3,000-6,000/year delivered 8-16x return.
• Ad intelligence — If competitor ad monitoring lets you improve your ad CTR by 15%, and your annual ad spend is $500,000, the improved efficiency is worth $75,000. Proxy cost: $2,000-5,000/year.
• Lead generation data — If web scraping produces 10,000 qualified leads per month and your cost per lead from other channels is $15, the proxy-sourced leads represent $150,000/month in equivalent value.

The comparison framework: What would it cost to get the same data without proxies? Manual research, third-party data providers, SaaS tools for rank tracking or price monitoring — all have costs. In nearly every case, self-managed proxy infrastructure costs 50-80% less than equivalent SaaS tools, with the added benefit of full data ownership and customization.

Present proxy costs alongside the value they enable, not in isolation. A $1,000/month proxy bill sounds expensive until you show it enables $15,000/month in data value.

If your monthly proxy spend exceeds $2,000-3,000, you have leverage to negotiate beyond published plan pricing. Proxy providers have healthy margins on enterprise accounts, and their cost of acquiring a high-value customer justifies discounting.

Negotiation levers that work:

• Volume commitment — Committing to a minimum annual spend gets the largest discounts. A provider might offer $6/GB on a monthly plan but $4/GB for a 12-month commitment of 100GB+/month.
• Multi-year deals — If you are willing to commit for 2 years, ask for an additional 10-15% discount over the annual rate. The provider gains revenue predictability.
• Competitive quotes — Get pricing from 2-3 providers and share them (or describe them generally) during negotiation. Providers will match or beat competitor pricing to win the deal.
• Overage rate reduction — Even if the base rate is firm, negotiate overage rates down to within 10-20% of the base rate instead of the default 50-200% premium.
• Rollover bandwidth — Ask for unused bandwidth to roll over month-to-month within the annual contract. This is low-cost for the provider and high-value for you.

Timing matters: End of quarter and end of year are the best times to negotiate, as sales teams push to hit targets. January and July (common fiscal year starts) are also good — providers may offer launch discounts on new plans.

What not to negotiate on: Do not sacrifice support quality, IP pool access, or success rate guarantees for a lower price. A 10% discount is worthless if it comes with a smaller IP pool that drops your success rate by 15%. Negotiate on price while holding firm on performance specifications.

Regardless of which pricing model you use, these strategies reduce your effective proxy cost without reducing data quality:

Mix proxy types by task: Use residential proxies only for targets that require them (heavily protected sites, geo-specific tasks). Route simpler requests through cheaper datacenter proxies. A mixed approach can cut total spend by 30-50% compared to using residential proxies for everything.

Cache aggressively: If you scrape the same pages regularly, cache results and check for changes before re-scraping. A product page that updates daily does not need hourly checks. Reducing redundant requests directly reduces bandwidth consumption and cost.

Fail fast on blocked requests: When a request returns a CAPTCHA or block indicator, do not follow redirects or load additional resources. Detect the block, discard the response, and queue a retry with a different IP. Every kilobyte of block-page data you load is wasted bandwidth.

Request only what you need: Disable image loading, font loading, and CSS rendering if you only need HTML content. Configure your HTTP client to reject responses above a certain size if you know your target pages should be small. Strip unnecessary headers from your requests to reduce outbound bandwidth.

Optimize retry logic: Do not retry immediately with the same parameters. If a request failed, rotate IP, change the user agent, and add a delay before retrying. Rapid retries without changes waste bandwidth and often fail again.

Monitor and audit usage: Review your proxy usage dashboard weekly. Look for unexpected bandwidth spikes, unusually high failure rates, or IP pools being depleted faster than expected. Early detection of problems prevents runaway costs.

Right-size your plan: Review actual usage against your plan allocation monthly. If you consistently use 60% of your allocation, downgrade. If you regularly hit 90%+, upgrade before overage charges kick in. The cheapest gigabyte is the one you do not pay for but did not need.