Technical Deep Dive

Mobile IP Pool Math: How Many Unique IPv4s Can Each Port Really Generate?

Understanding carrier-grade NAT pools, rotation mathematics, and why each mobile proxy port can generate thousands of unique IPs daily—without owning a city-specific pool. Scale with multiple ports for higher throughput.

12 min read
Last updated: Oct 30, 2025
Research-backed

TL;DR: The Core Truth About Mobile IP Pools

Mobile carriers (T-Mobile, Verizon, AT&T) front their subscribers with very large CGNAT pools. Your mobile device doesn't "own" a city-local IP pool—it samples from carrier NAT egress clusters that can serve huge regions (multi-state).

With fast rotations (10–60s), each mobile proxy port can see thousands of unique IPv4 addresses per day until you hit natural duplicates from the birthday paradox. Rotation math shows 2,000-8,500 unique IPs/day/port is typical with aggressive cadences. Users can rent multiple ports and scale throughput while sharing a common prepaid bandwidth pool.

Exact per-city pool sizes aren't public. Carriers don't publish per-gateway CGNAT pool sizes, and IP geolocation at city-level is too imprecise on mobile (50-100+ km accuracy radius). The only robust approach: measure empirically with 7-day logging of unique /24s and /32s.

What We Can Know (Publicly) About Carrier CGNAT Pools

Mobile carriers operate Carrier-Grade NAT (CGNAT) at massive scale. A single public IPv4 address can represent hundreds or thousands of end users simultaneously. This is why mobile proxy egress pools are enormous and highly shared—it's an economic necessity driven by IPv4 address exhaustion.

KEY INSIGHT FROM CLOUDFLARE

"CGNAT makes it possible for one IP address to serve hundreds or even thousands of users, which is why detecting shared IPs is critical for security and analytics."

Source: Cloudflare Blog

T-Mobile US (AS21928): IPv6-First with Massive CGNAT

T-Mobile operates an IPv6-only mobile core with 464XLAT for IPv4 backward compatibility. This means all IPv4 connectivity goes through NAT64/CGNAT infrastructure. The carrier originates millions of IPv4 addresses across numerous prefixes:

These blocks confirm heavy address sharing and frequent egress changes. When you rotate a mobile proxy on T-Mobile, you're sampling from these regional CGNAT pools, not a phone-specific allocation.

⚠️ IMPORTANT CAVEAT

These ASN prefix totals are upper bounds and include non-CGNAT uses (enterprise, IoT, Wi-Fi calling). Carriers don't publish exact NAT pool sizes per PGW (Packet Gateway) or UPF (User Plane Function), and pool allocations change dynamically based on traffic.

Verizon Wireless: Multi-ASN Giant Pools

Verizon Wireless mobile egress commonly appears under AS6167 (CELLCO-PART) andAS22394 (CELLCO), both operated by Verizon/Cellco, with truly massive IPv4 allocations:

  • 97.128.0.0/9 - 8.3 million addresses
  • 75.192.0.0/10 - 4.2 million addresses
  • 174.192.0.0/10 - 4.2 million addresses

This demonstrates the scale of national mobile CGNAT pools. While not every subnet is used for mobile egress (some serve enterprise, FiOS, etc.), the sheer address budget means your mobile modem can sample many different /24 blocks over hours or days.

Key Takeaway: Pools Are Regional, Not City-Scoped

Both T-Mobile and Verizon route mobile traffic through regional CGNAT clusters, not tower-by-tower NAT boxes. A device in New York City might egress via a New Jersey data center; a San Francisco device could use a Los Angeles or Nevada router. This is intentional for load balancing, redundancy, and cost optimization.

Why "NYC vs SF vs Miami" Pool Sizes Is the Wrong Mental Model

The question "How big is T-Mobile's New York pool?" assumes mobile egress is geographically localized at the city level. This assumption is incorrect for how modern mobile networks actually operate.

Egress Is Centralized and Region-Wide

Mobile carriers architect their networks for scale and efficiency, not fine-grained geographic distribution:

What Actually Happens

  • • Traffic aggregates at regional PGW/UPF nodes
  • • One NYC device may egress via New Jersey
  • • SF devices can route through SoCal/Nevada
  • • CGNAT pools serve multi-state regions
  • • Load balancing shifts egress points dynamically

Common Misconceptions

  • • "Each city has a dedicated NAT pool"
  • • "Tower location = egress IP location"
  • • "IP geolocation is city-accurate"
  • • "Pool size is stable and public"
  • • "Nearby devices share the same /24"

IP Geolocation Accuracy Is Poor on Mobile

Commercial geolocation databases (MaxMind, IP2Location, IPinfo) provide an accuracy radius with each lookup. For mobile IPs, this radius is frequently 50-100+ kilometers, making city-level precision meaningless.

FROM MAXMIND DOCUMENTATION

"The accuracy radius indicates the geolocation precision. Mobile carrier IPs often have radii of 50-200km due to CGNAT and routing complexities. City-level accuracy should not be assumed."

Source: MaxMind Support

Academic research confirms this limitation. Studies analyzing mobile network geolocation show that tower handoffs, CGNAT egress points, and VPN use all degrade city-level accuracy. Therefore, "New York pool size" isn't a stable or measurable quantity.

Carrier-Specific Data You Can Cite (Without Over-Promising)

T-Mobile US (AS21928)

Confirmed IPv4 Blocks for Mobile Services

  • 208.54.0.0/16 - Used for Wi-Fi calling (documented in T-Mobile's enterprise firewall guidance)
  • 66.94.0.0/19 - Mobile data egress
  • 172.58.0.0/15 - Large mobile CGNAT allocation

T-Mobile's IPv6-only core with 464XLAT means all IPv4 traffic undergoes translation, confirming these blocks are heavily shared via CGNAT.

Verizon Wireless (AS22394, AS6167)

Enormous IPv4 Space Across Multiple ASNs

  • AS22394 (CELLCO):
    • 97.128.0.0/9 - 8.3M addresses
    • 75.192.0.0/10 - 4.2M addresses
  • AS6167 (CELLCO-PART):
    • 174.192.0.0/10 - 4.2M addresses

While not all subnets are active CGNAT pools, this address budget underscores why your modem can sample many different /24 blocks over time. Verizon's scale enables high IP diversity for mobile proxy users.

Scaling with Multiple Ports: Architecture & Economics

Modern mobile proxy platforms let users rent multiple ports simultaneously that all share a common prepaid bandwidth pool. This architecture provides flexible scaling for different use cases:

How the Multi-Port System Works

1
Rent Ports (Monthly Fee)

Each port provides a unique proxy endpoint with independent IP rotation. Typical pricing: $20/port/month. Users can rent 1, 10, 50, 100, or 500+ ports depending on throughput needs.

2
Buy Bandwidth (One-Time Prepaid)

Purchase GB packages (10GB, 50GB, 250GB, etc.) that all your ports share. Typical pricing: $4-8/GB. Bandwidth doesn't expire—use it anytime across any port.

3
Shared Bandwidth Pool

All ports deduct from the same GB balance. Use Port 1 for 5GB and Port 2 for 3GB—both pull from your total prepaid bandwidth. This eliminates waste from unused port-specific allocations.

Example Scaling Scenarios

Small Project: 1-2 Ports

Testing, small-scale scraping, personal automation

  • • Setup: 1-2 ports + 10-50 GB
  • • IP Diversity: 2k-8.5k unique IPs/day per port
  • • Use Case: Single-target scraping, QA testing
  • • Cost: $20-40/month ports + $40-400 bandwidth

Medium Project: 10-50 Ports

Multi-platform scraping, social media automation

  • • Setup: 10-50 ports + 100-500 GB
  • • IP Diversity: 20k-425k unique IPs/day (aggregated)
  • • Use Case: Parallel scraping, account management
  • • Cost: $200-1,000/month ports + volume pricing

Enterprise: 100+ Ports

Large-scale data collection, ad verification, market intelligence

  • • Setup: 100-500+ ports + TB-scale bandwidth
  • • IP Diversity: 200k-4.25M unique IPs/day (aggregated)
  • • Use Case: Massive parallel operations, enterprise scraping
  • • Cost: Custom pricing with volume discounts
  • • Benefits: Dedicated account manager, SLA guarantees
  • • Architecture: Load-balanced across multiple CGNAT clusters

💡 KEY ADVANTAGE: BANDWIDTH POOLING

Unlike providers that allocate bandwidth per-port (forcing you to buy 10 ports × 50GB = 500GB total even if you only need 200GB), the shared pool model lets you right-size both dimensions independently. Rent the ports you need for concurrency, buy the GB you need for volume—no waste.

Practical Guidance for Mobile Proxy Users

1. Choose Rotation Cadence by Goal

Maximum IP Diversity

For web scraping, social media automation, or any use case requiring high IP variety:

  • • Use 10–20 second rotation windows
  • • Run 12–24 hours/day for volume
  • • Expect 2,000–8,500 unique IPs/day per port
  • • Scale with multiple ports: Rent 10, 50, or 100+ ports for proportionally higher throughput (all share bandwidth pool)
  • • Monitor duplicate rate to tune cadence
  • • Note: Very aggressive rotation (<10s) may trigger bot-defense on some platforms

Reproducible QA Testing

For quality assurance, functional testing, or when you need session consistency:

  • • Use sticky sessions (no time-based rotation)
  • • Rotate only between test jobs
  • • 60+ second cadence if time-based rotation required
  • • Predictable IP behavior for debugging

2. Measure Your Effective Pool (Don't Assume City Size)

Since carriers don't publish per-city CGNAT sizes and geolocation is imprecise, the only robust approach is empirical measurement:

7-Day Logging Protocol

  1. 1
    Log every unique IP with full /32 address, /24 subnet, ASN, and geolocation (including accuracy radius) for 7 consecutive days.
  2. 2
    Test two rotation cadences simultaneously: 20-second and 60-second intervals to see how cadence affects diversity.
  3. 3
    Plot unique IPs vs. time. When the curve bends (growth slows), you're approaching local pool limits for that hour/cluster.
  4. 4
    Expect diurnal shifts. Your morning IPs may come from different CGNAT clusters than evening IPs due to carrier load balancing.

3. Expect Duplicates—Even in Large Pools

CGNAT clusters reassign popular egress IPs based on demand. Seeing duplicate IPs doesn't indicate poor quality—it's fundamental to CGNAT economics. Key points:

  • Time-based sharing is normal. The same IP might serve 100 users in the morning, different 100 users in the afternoon. You might get it twice.
  • Duplicate rate scales with n/M. Example: n=8,640 draws → ~18% duplicates at M=20k pool, ~8% at M=50k, ~2% at M=200k. Larger pools dramatically reduce duplicate rates.
  • For <5% duplicates: Target pools of 200k+ IPs with aggressive rotation, or rent multiple ports to sample from different CGNAT clusters simultaneously (requires massive carrier allocations like Verizon).
  • Duplicate /24s ≠ duplicate /32s. You may revisit the same /24 subnet with a different /32 IP, so /24 diversity can remain high even when some /32s repeat across sessions.

4. Remember: Mobile Is CGNAT by Design

Both T-Mobile's IPv6-only + 464XLAT architecture and Verizon's multi-million address IPv4 space point to the same reality: aggressive address sharing via CGNAT is the foundation of mobile networks.

The Bottom Line

Model your expected outcomes using the math above, measure empirically with 7-day logging, and don't promise fixed "NYC pool sizes" that carriers don't publish and that change dynamically. Focus on what you can control: rotation cadence, hours active, and statistical modeling.

Frequently Asked Questions

Can I get city-specific mobile proxy pools (e.g., only NYC IPs)?

No reliable way exists to guarantee city-level precision with mobile proxies. CGNAT egress is regional (multi-state), and IP geolocation has 50-100+ km accuracy radii. You'll get IPs that geolocate broadly to the Northeast US, but not pure NYC. If you need city precision, consider datacenter proxies with known locations.

Why do some mobile proxy providers claim 'millions of IPs'?

They're citing the carrier's total ASN allocation (e.g., Verizon's 8.3M in AS22394), not the active CGNAT pool each port can reach. One port can practically access 2k-8.5k unique IPs/day with fast rotation—impressive, but not millions. To scale throughput, users rent multiple ports (each port generates independent rotations while sharing bandwidth). Marketing often conflates theoretical ASN size with per-port reachability.

How do I verify my mobile proxy provider's IP diversity claims?

Run the 7-day logging protocol: log every unique /32 and /24 with timestamps. Plot cumulative uniques over time. If they claim &quot;10k IPs/day&quot; but you plateau at 3k, you've measured the truth. Request your provider share their methodology or provide test credentials for verification.

Is 20% duplicate rate bad for mobile proxies?

Not necessarily. With 4k-8k draws/day per port against a 20k-50k pool, 15-25% duplicates are mathematically normal. If you need <5% duplicates, you need either: (1) larger pools (200k+), (2) fewer draws (slower cadence), or (3) multiple ports to sample from different CGNAT clusters simultaneously. Users can rent 10, 50, or 100+ ports to scale both throughput and IP diversity.

Do mobile proxies work better than residential proxies for bot detection?

Mobile IPs (4G/5G CGNAT) often have higher success rates because they're inherently shared (one IP = hundreds of users) and constantly rotating. This mimics natural mobile user behavior. Residential proxies can work well too, but mobile's built-in IP churn provides natural anti-fingerprinting. Success rates depend on target platform's sophistication.

Conclusion: Model It, Measure It, Don't Guess It

The question "How many unique IPs can each mobile proxy port generate?" doesn't have a single answer—it depends on:

  • Carrier and ASN (T-Mobile, Verizon, AT&T have different pool sizes)
  • Rotation cadence (10s vs 60s makes a 6-12x difference in draws)
  • Hours active (12h vs 24h doubles your opportunities)
  • Reachable CGNAT pool size (varies by region, time of day, load)
  • Natural duplicate collisions (birthday paradox kicks in around n²/2M)

Our occupancy model shows you can expect 2,000–8,500 unique IPv4s per day per port with aggressive 10-20s rotations, assuming large carrier pools (50k-200k+ IPs). This is why mobile proxies can deliver impressive IP diversity without you "owning" a city-specific allocation. For higher throughput needs, users can rent multiple ports (10, 50, 100+) that all share a common prepaid bandwidth pool—each port operates independently with its own IP rotation cycle.

Best Practices for Mobile Proxy Users

  1. 1.Don't rely on city-level pool claims. CGNAT is regional, and geolocation is imprecise. Focus on /24 subnet diversity instead.
  2. 2.Measure empirically with 7-day logging. Track unique /32s, /24s, ASNs, and geo-radii to understand your actual reachable pool.
  3. 3.Tune rotation cadence to your use case. 10-20s for max diversity, 60s+ for stability, sticky sessions for QA.
  4. 4.Accept duplicates as normal. 15-25% duplicate rate is mathematically expected in CGNAT environments. It's not a quality issue.

By understanding the mathematics and infrastructure realities of mobile CGNAT, you can make informed decisions about rotation strategies, set realistic expectations, and properly evaluate mobile proxy providers.

The mobile proxy advantage isn't about "owning millions of IPs"—it's about efficiently sampling from massive, shared carrier pools that naturally rotate and provide authentic mobile device behavior. Users can scale throughput by renting multiple ports that all share a common prepaid bandwidth pool, with each port generating independent IP rotation cycles.

Ready to Experience High-Diversity Mobile Proxies?

Test our mobile proxy network with real 4G/5G devices on T-Mobile, Verizon, and AT&T. Aggressive rotation cadences. Thousands of unique IPs per port.

View PricingExplore 17 Countries

1GB free trial • No credit card required • Test on your platforms

Related Reading

Trustware 101: AI Agents & Blockchains

Understanding next-gen trust infrastructure for autonomous systems.

Free Proxy Testing Tool

Test your mobile proxies: DNS leaks, anonymity, performance metrics.

USA Mobile Proxies

T-Mobile, Verizon, AT&T networks across all 50 states.

Mobile Proxy Pricing

GB-based plans with volume discounts. No expiration.