Proxy Types: Datacenter, Residential, Mobile, ISP

The IP your scraper appears from is one of the strongest fingerprints anti-bot systems use. Four broad proxy categories exist; each has a niche.

The four types

Type	What it is	Trust level	Approx. cost	Use when...
Datacenter	IPs in commercial hosting (AWS, OVH, Hetzner ranges)	Low, easily detected	Cheap, often $1-3/GB or $0.50-2/IP-month	Easy targets, internal APIs, dev/test
Residential	IPs from real ISP-allocated home connections	High, looks like real users	Medium, $4-15/GB	Most mid-tier consumer sites
Mobile	IPs from cellular carriers (4G/5G)	Very high, shared by many real mobile users	Expensive, $8-40/GB	Mobile-optimized sites, hardest targets
ISP (a.k.a. "static residential")	Server hardware in datacenters but with ISP-issued IPs	Medium-high, looks residential, behaves like datacenter	Medium, $2-6/IP-month	When you need residential-look but stable IPs

These are 2026 ballpark numbers, pricing varies and shifts quarterly.

Datacenter proxies

Cheapest. IPs originate in commercial hosting (AWS, GCP, OVH, etc.). Anti-bot vendors maintain lists of known datacenter ASNs and flag traffic from them.

When they work:

• The target doesn't run sophisticated bot detection.
• You're scraping an internal or partner API where IPs aren't checked.
• Cost matters more than block rate.
• Dev/test (your own scraping infrastructure).

When they fail:

• Cloudflare-protected sites (high block rate).
• E-commerce / SaaS dashboards.
• Any site that proudly advertises "we block bots."

Providers: most rotate every request from a pool of thousands. Often the cheapest entry point.

Residential proxies

Pooled from real consumer ISP connections, Comcast, BT, Deutsche Telekom, etc. Most providers source them via consent-based SDKs in free apps/VPNs (think Hola). Users opt in; their bandwidth gets resold.

Why they work:

• Anti-bot systems can't easily distinguish you from a real visitor on a home network.
• IPs rotate from a pool of millions, often per-request.
• Geographic targeting is precise (city-level common).

Tradeoffs:

• Cost: 5-15x datacenter prices.
• Latency: variable. Some endpoints are slow.
• Reliability: a proxy IP might disappear mid-session (the user's phone went to sleep).

When to use: mid-tier consumer sites that block obvious bots. Most e-commerce. SERP scraping for non-Google sources.

Mobile proxies

Pooled from mobile carriers, Verizon, Vodafone, etc. Multiple real users share the same IP via NAT, so a single mobile IP serves dozens or hundreds of legitimate humans simultaneously.

Why they're powerful:

• Sites can't ban a mobile IP without banning real users.
• High trust score.
• Frequently used for Instagram, TikTok, mobile-app endpoints.

Tradeoffs:

• Most expensive ($8-40/GB).
• Slower than residential.
• Pools are smaller than residential, less geographic variety.

When to use: hardest targets, mobile-app endpoints, social platforms.

ISP proxies

Servers in datacenters but with IP blocks issued by residential ISPs (typically through arrangements with the ISP). Looks "residential" to anti-bot systems while being as fast and reliable as datacenter.

Why they exist: rotating residential proxies are flaky and slow. ISP proxies give you the trust without the flakiness.

Tradeoffs:

• Smaller pools (thousands, not millions of IPs).
• IPs are static, if one gets banned, it stays banned.
• Mid-tier price.

When to use: session-based scraping (logged-in accounts), sites where you want consistent IPs per session.

Decision matrix

Target difficulty	Recommended proxy type
Internal/partner API	Datacenter or no proxy
Static e-commerce, no Cloudflare	Datacenter, fallback to residential
Cloudflare-protected site	Residential
Major e-commerce (Amazon, Walmart)	Residential, sometimes mobile
Social platforms (LinkedIn, Instagram)	Mobile or curated residential
Banking, hard fintech	Often impossible without partnerships

Cost math, order of magnitude

For a site scraped via 100k requests/day, each request averaging 50 KB:

• Total bandwidth: 5 GB/day = 150 GB/month
• Datacenter: 150 GB × $2 = $300/month
• Residential: 150 GB × $8 = $1,200/month
• Mobile: 150 GB × $20 = $3,000/month

Proxy bills can quickly exceed engineering salaries. Two ways to reduce: lower bandwidth per request (don't fetch images you don't need; use HEAD where possible), or move easier targets to datacenter and reserve premium proxies for hard targets.

The "free proxy" trap

There are free proxy lists. They are almost always:

• Compromised devices. Bots, malware, hacked routers. Using them is ethically dubious and operationally unreliable.
• Honeypots. Proxies run by security researchers or hostile actors logging your traffic.
• Dead within hours. No SLA, no rotation.

Do not use free proxies for anything serious. The math doesn't work, and the legal/security risk is real.

Geographic considerations

If the target serves different content per region, proxy origin matters:

• Pricing pages often differ by country.
• Search results vary by locale.
• Some sites geofence content entirely.

Providers expose this as a per-request parameter (?country=DE&city=berlin) or by routing through a regional endpoint. Confirm the actual egress IP with https://api.ipify.org?format=json during testing.

What about Tor?

Tor is a legitimate proxy network, but:

• Exit nodes are widely known and blocked by most commercial sites.
• Slow.
• High abuse rate causes many sites to permanently block Tor exits.

Tor is good for personal privacy, not for production scraping.

Hands-on lab

Pick one target you've scraped earlier in the curriculum. Identify:

1. What proxy type would actually fit?
2. What's the bandwidth per request? Multiply by daily volume.
3. Estimate monthly proxy cost in each tier.

Most projects can stay on datacenter or basic residential. Knowing the cost up front prevents the "oh no, the proxy bill is $4k/month" surprise.