Proxy Pricing by Use Case: Datacenter vs Residential vs Scraping APIs, and What to Pay (Aug 2026)
A cost comparison of proxies and scraping services by use case: datacenter proxy pools, residential per-GB, pay-per-request scraping APIs, and agent-friendly scrape APIs (Firecrawl, Crawlbase). With per-case recommendations and worked cost examples.
A companion to “Polite Web Scraping: Rate-Limit Design Patterns, Whether You Need Proxies, and the AI-Agent-Friendly Open Source Stack” (polite-web-scraping-design-patterns.md). That article answers whether you need a proxy; this one answers how much it costs when you do. All prices are list prices from official provider pages fetched Aug 5, 2026 — pricing changes; verify before committing.
TL;DR
There are three different pricing models, and picking the wrong one can cost 10-100x more for the same job:
| Model | What you pay for | Who it fits | Entry price (Aug 2026) |
|---|---|---|---|
| Datacenter proxy pool | per IP / month | You run your own crawler, site is robots-friendly, moderate volume | Webshare 100 IPs = $2.99/mo (10 free) |
| Residential proxy | per GB of traffic | Geo-targeting or sites that block datacenter IPs; you run your own crawler | $2.25-$3.75/GB at modest volume (Webshare/Decodo) |
| Managed scrape API | per successful request/1k | You want proxies+retries+rendering+CAPTCHA handled; fastest to ship | Zyte from $0.13/1k; ScraperAPI $0.49/1k; Firecrawl $0.83/1k |
Cheapest per use case:
- Polite scraping of a robots-allowed site (any volume): $0 — no proxy. Reuse the article’s tiered approach (Crawlee starts at
[null]). Only add proxies if actually blocked. - Datacenter rotation for cheap scaling: Webshare — 100 proxies for $2.99/mo (https://www.webshare.io/pricing). At $0.018-$0.03/IP/mo, datacenter IPs are ~100-300x cheaper than residential bandwidth.
- Managed API at scale, simple sites: Zyte PAYG, $0.13-$0.23/1k for easy tiers (https://www.zyte.com/pricing/) — the cheapest on-paper per-request. ScrapeOps Aggregator is the flexible aggregator at $0.36/1k (https://scrapeops.io/proxy-aggregator/).
- Residential for geo/anti-bot, per-GB: Webshare $2.25/GB @100GB or Decodo (ex-Smartproxy) $3.75/GB entry; IPRoyal if you want pay-as-you-go with no expiry ($7.35/GB small, $5.51/GB @10GB); Bright Data is the premium option (list $8/GB, ~$4/GB with promo).
- Hardest anti-bot targets: Bright Data Web Unlocker $1.50/1k (5k free/mo) bundles unblocking+CAPTCHA (https://brightdata.com/pricing/web-unlocker); Zyte’s highest tiers handle advanced sites.
- AI-agent friendly: Firecrawl hosted — free 1,000/mo, $0.83/1k at Standard (https://firecrawl.dev/pricing.md) — or self-host Firecrawl/Crawl4AI for infra cost only. Crawlbase gives 20,000 free requests (https://crawlbase.com/pricing).
The headline insight: managed scrape APIs charge per successful response, raw proxies charge per GB. For ~1MB HTML pages, Zyte’s $0.13/1k = roughly $0.13/GB-equivalent, while residential bandwidth runs $2.25-$8/GB — so the managed API is often 10-30x cheaper than raw residential bandwidth and does more work for you. Raw residential only wins when you need full control of your own Crawlee/Scrapy pipeline, or pages are tiny.
Part 1 — The three pricing models, and what’s actually cheap
1.1 Datacenter proxy pools (per IP/month) — the “cheap rotation” option
Datacenter (DC) proxies come from cloud/data-center IP ranges. They’re easily detected by anti-bot systems, but for a robots-allowed site they’re all you’d ever need, and the price is tiny:
- Webshare: 10 proxies free; 100 IPs $2.99/mo ($0.0299/IP), 1,000 IPs $26.91 ($0.0269/IP), 5,000 $119.60 ($0.0239/IP), 60,000 $1,076 ($0.0179/IP). Yearly billing cuts up to 60%. (https://www.webshare.io/pricing)
- Webshare static residential (ISP): $0.30/IP → $0.225/IP @10,000 (https://www.webshare.io/pricing).
- Apify: 5 DC IPs on Free, then ~$1/IP on Starter down to ~$0.60/IP (https://apify.com/pricing).
Rule of thumb: ~$0.02-0.03/IP/mo for shared datacenter IPs. If your crawler is polite (per the companion article), a 100-1,000 IP pool at $3-$27/mo is already massive headroom.
1.2 Residential proxies (per GB of traffic) — the expensive axis
Residential IPs ride on real end-user connections (Bright Data claims 400M+; Oxylabs 175M+), so they pass most anti-bot checks — and that’s why they’re priced per-GB at an order of magnitude more. Verified list prices (Aug 2026):
| Provider | Entry | Effective price/GB | Notes |
|---|---|---|---|
| Bright Data | PAYG | $8.00 list (→ ~$4.00 with 50% promo); $3.50/GB @ $499/mo, $2.50/GB @ $1,999/mo | 400M+ IPs, free trial (no card); KYC required |
| Oxylabs | 5 GB / $30/mo | $6.00 → $2.50/GB @ 1 TB | 175M+ IPs |
| IPRoyal | PAYG, $10 deposit | $7.35 (1 GB) → $5.51 (10 GB) → $1.75 @ 10 TB | no expiry, no contracts |
| Smartproxy → Decodo | 3 GB / $11.25/mo | $3.75 → $2.75 @ 100 GB, from $2.00 enterprise | 115M+ IPs |
| SOAX | Builder $200/mo | $3.00 tier-1 (US/EU) → from $0.85 enterprise | credit-based (1 credit = $1) |
| Webshare | 1 GB / $3.50 | $3.50 (1 GB) → $2.25 @ 100 GB → $1.40 @ 3,000 GB | monthly; yearly: $1.58/GB @100GB, $0.98/GB @3TB |
| ScrapeOps resi aggregator | $15/mo / 3 GB | $5.00 → $2.49 @ 100 GB → $2.00 @ 500 GB | routes to many pools |
| Apify resi add-on | — | $8.00 (Starter) → $7.00 (Business) | bundles with actor platform |
Sources: https://brightdata.com/proxy-types/residential-proxies · https://oxylabs.io/pricing · https://iproyal.com/pricing/residential-proxies/ · https://decodo.com/proxies/residential-proxies/pricing · https://soax.com/pricing · https://www.webshare.io/pricing · https://scrapeops.io/proxy-aggregator/ · https://apify.com/pricing
What a GB buys: providers don’t publish “pages per GB,” and bandwidth accounting includes request headers + response (Bright Data FAQ, https://docs.brightdata.com/proxy-networks/residential/introduction). A reasonable planning estimate is ~1,000 HTML pages (≈0.5-1 MB each) per GB — so $2.25/GB ≈ $2.25 per 1,000 pages. That’s the number to compare against the per-request APIs below.
1.3 Managed scraping APIs (per successful request) — the “everything included” axis
These bundle proxies + retries + JS rendering + CAPTCHA solving + geo-targeting into one HTTP call, charging only for successful responses. That’s a fundamentally different cost unit than bandwidth.
| Provider | Free tier | Entry paid | Effective per 1k | Notes |
|---|---|---|---|---|
| Zyte | $5 credit trial | PAYG | $0.13 (Simple) - $1.27 (Advanced) HTTP body; $1.01-$16.08 browser-rendered | 5 difficulty tiers by site; DC+resi+mobile auto-rotated; $500/mo commitment → from $0.06/1k |
| ScrapeOps Aggregator | 1,000 credits | $9/mo | $0.36 (25k credits) → $0.083 @ 3M | routes to 15+ smart-proxy APIs; resi routing = 10x credits, render_js = 10x |
| ScraperAPI | 1,000 credits (5k trial) | Hobby $49/mo | $0.49 (100k credits) → $0.095 @ 5M | premium resi+mobile by default; domain multipliers (Amazon 5x, Google 25x) |
| Crawlbase | 20,000 requests | Smart AI Proxy $149/mo | ~$0.75 (200k credits, DC); resi at $229/mo | 1 credit = 1 request; JS = 2 credits |
| Bright Data Web Unlocker | 5,000/mo | PAYG | $1.50 → $1.30 @ Scale | residential-grade unblocking + CAPTCHA + rendering |
| Firecrawl | 1,000/mo | Hobby $16/mo (annual) | $3.20 (5k) → $0.83 @ 100k (Standard $83/mo) | agent-friendly: markdown/JSON, MCP; self-hostable (AGPL) |
| Apify | $5 usage/mo | Starter $29/mo | ~$0.20/CU-h compute + resi $8/GB | pay for compute + proxies separately |
Sources: https://www.zyte.com/pricing/ · https://scrapeops.io/proxy-aggregator/ · https://www.scraperapi.com/pricing · https://crawlbase.com/pricing · https://brightdata.com/pricing/web-unlocker · https://firecrawl.dev/pricing.md · https://apify.com/pricing
Why the unit matters: Zyte’s $0.13/1k for a “Simple” site is roughly $0.13 per GB of response body — versus $2.25-$8/GB for raw residential. Even Bright Data’s residential-grade unlocker ($1.50/1k) beats its own raw residential bandwidth ($2.50-$8/GB) on page-sized responses. Managed APIs win on cost-per-page and save you the engineering of proxies, retries, and rendering.
Part 2 — Worked example: 100,000 pages/month of a robots-allowed site
Assume ~1 MB HTML per page ≈ 100 GB/month of raw traffic, polite crawler, no anti-bot wall.
| Option | Monthly cost | Effort |
|---|---|---|
| No proxy (Crawlee/Scrapy, polite defaults) | $0 | write own crawler |
| Datacenter rotation (Webshare 1,000 IPs) | $26.91 | write own crawler |
| Zyte API (Simple tier, PAYG) | $13 (100k × $0.13/1k) | API call per page |
| ScrapeOps Aggregator | $29 (250k credits) | API call per page |
| ScraperAPI Hobby | $49 (100k credits) | API call per page |
| Firecrawl Standard | $83 (100k credits) | API call + markdown/JSON for agents |
| Webshare rotating residential (100 GB) | $225 | write own crawler |
| Bright Data residential PAYG | $800 ($8/GB, or ~$400 w/ promo) | write own crawler |
For a robots-allowed site the answer is unambiguous: $0 first (you likely never get blocked if you follow the companion article’s rate-limit pattern), then Webshare DC if you want IP diversity cheaply, then a per-request API if you’d rather not run infrastructure. Residential bandwidth is the most expensive option here — reserve it for sites that actually block datacenter/automated traffic or need geo-targeting.
For a hard site (Cloudflare/DataDome, geo-locked): budget the residential-grade managed tier instead — Bright Data Web Unlocker $1.50/1k or Zyte’s higher difficulty tiers — because raw residential bandwidth still leaves you to build retries, rendering, and CAPTCHA handling yourself.
Part 3 — Recommendations per use case
1. Polite scraping, robots-allowed, any volume → $0.
No proxy. Tiered escalation built into your crawler (Crawlee tieredProxyUrls: [[null], ...] https://crawlee.dev/docs/guides/proxy-management) means you only touch a proxy if you’re actually blocked.
2. Cheapest insurance / modest scale-up → Webshare datacenter. 10 free proxies to test; $2.99/mo for 100 IPs. Perfect for a polite crawler that just needs more than one egress IP. Don’t expect it to defeat real anti-bot systems (the companion article’s caveat).
3. Build vs buy at scale → Zyte or ScrapeOps. Zyte is the cheapest per-request at easy tiers and handles everything (IP rotation across DC/residential/mobile automatically); ScrapeOps gives you an aggregator + a cheap datacenter default with optional per-request “residential” upgrade. Both free-tier start small.
4. Agent/LLM data extraction → Firecrawl (or self-host). 1,000 scrapes/mo free, $0.83/1k at Standard, official MCP server + agent skill (https://github.com/firecrawl/firecrawl-mcp-server), respects robots.txt by default. For $0-at-scale, self-host Firecrawl (AGPL, docker-compose) or Crawl4AI (Apache-2.0, Docker + MCP) per the companion article.
5. Geo-targeting or genuinely blocked sites → residential per-GB or managed unblocker. Cheapest per-GB at modest volume: Decodo $3.75/GB entry or Webshare $2.25/GB @100GB; IPRoyal if you want no-expiry PAYG and dislike commitments. For the hardest targets, Bright Data Web Unlocker $1.50/1k (5,000 free/mo) is the “it just works” price point; Zyte’s Advanced tier ($1.27/1k PAYG) is the alternative.
6. Always avoid → free public proxy lists. Rejected in the companion article: “easy for websites to detect, throttle, or block,” only for tests (https://www.scraperapi.com/blog/best-10-free-proxies-and-free-proxy-lists-for-web-scraping/).
Part 4 — Caveats that change the numbers
- Pricing is volatile. Several providers run standing promotions (Bright Data’s 50%-off coupon, Webshare’s 50%-off banners); list price ≠ street price. Verify on the day you buy.
- Bandwidth accounting bites. Residential “per GB” counts headers + responses both directions, and IPRoyal’s FAQ warns traffic “runs out fast” partly due to rotation/keep-alive overhead (https://help.iproyal.com/en/articles/12782215-why-your-traffic-seems-to-run-out-quickly).
- Site difficulty tiers. Zyte/ScraperAPI/Crawlbase price by how hard a specific site is to scrape (5 tiers, or domain multipliers like Amazon=5x, Google=25x). A cheap “Simple” price only applies if your target is actually simple.
- Credits vs requests. “1 credit” isn’t uniform: Crawlbase JS = 2 credits, Firecrawl JSON/enhanced mode = +4 credits, ScrapeOps residential = 10x credits. Always compute your effective per-request cost.
- Commitment requirements. SOAX’s cheap per-GB tiers require $200-$3,000/mo plans; Oxylabs’ $5/GB needs a $100/mo plan. Pay-as-you-go (Bright Data, IPRoyal, Zyte, ScraperAPI) costs more per unit but has no minimum.
- Self-host vs hosted. For a robots-allowed site, self-hosting Firecrawl/Crawl4AI or running Crawlee/Scrapy on a $5 VPS means the only cost is your compute — no per-request or per-GB fees at all.
Sources
- Bright Data residential: https://brightdata.com/proxy-types/residential-proxies · https://brightdata.com/pricing/proxy-network
- Bright Data Web Unlocker: https://brightdata.com/pricing/web-unlocker · Web Scraper API: https://brightdata.com/pricing/web-scraper
- Oxylabs residential: https://oxylabs.io/pricing · https://oxylabs.io/pricing/residential-proxy-pool
- IPRoyal residential: https://iproyal.com/pricing/residential-proxies/ · trial: https://help.iproyal.com/en/articles/7222620-do-you-offer-a-free-trial · traffic: https://help.iproyal.com/en/articles/12782215-why-your-traffic-seems-to-run-out-quickly
- Smartproxy/Decodo residential: https://smartproxy.com/pricing → https://decodo.com/proxies/residential-proxies/pricing
- SOAX: https://soax.com/pricing · https://soax.com/proxies/residential
- Webshare: https://www.webshare.io/pricing
- ScrapeOps: https://scrapeops.io/proxy-aggregator/ · request costs: https://scrapeops.io/docs/web-scraping-proxy-api-aggregator/account/request-costs/
- Zyte API: https://www.zyte.com/pricing/
- ScraperAPI: https://www.scraperapi.com/pricing · free proxies: https://www.scraperapi.com/blog/best-10-free-proxies-and-free-proxy-lists-for-web-scraping/
- Crawlbase: https://crawlbase.com/pricing
- Apify: https://apify.com/pricing
- Firecrawl: https://firecrawl.dev/pricing.md · MCP: https://github.com/firecrawl/firecrawl-mcp-server
- Crawlee proxy management (tiered, starts at no proxy): https://crawlee.dev/docs/guides/proxy-management
- Companion article on this site: Polite Web Scraping: Rate-Limit Design Patterns