When you send data across the internet, it traverses multiple independent networks called Autonomous Systems (AS). Each AS - whether it's your local ISP, a content provider like Netflix, or a global carrier like AT&T - must have agreements with other networks to exchange traffic. These agreements form a hierarchy that determines who pays whom, and ultimately affects your internet quality and price.

The tier system isn't an official classification - no organization assigns tier status. Instead, it emerges from economic reality: networks that need to buy connectivity from others (lower tiers) versus those that don't (higher tiers).

Understanding Network Tiers

Networks fall into three tiers based on a simple question: Can they reach the entire internet without paying anyone?

Tier 1: These networks reach the entire internet without paying for transit. They accomplish this by maintaining settlement-free peering (no-payment traffic exchange) with every other Tier 1 network. If even one Tier 1 network refuses to peer with you, you cannot be Tier 1 - you'd need to pay someone to reach that network's customers.

Tier 2: These networks peer with some networks for free but must pay at least one upstream provider for transit to reach parts of the internet they can't access through peering. Most large national ISPs are Tier 2.

Tier 3: These networks purchase all their connectivity from upstream providers. They're typically local ISPs, wireless providers, or enterprise networks. Your home ISP is likely Tier 3 unless you're with a major national provider.

The key technical difference: Tier 1 networks have no "default route" in their routers - they must know explicit paths to every destination. Lower tier networks point unknown destinations to their transit provider.

Tier 1 Networks: The Internet's Backbone

What Makes a Network Tier 1

A Tier 1 network must satisfy one critical requirement: it can deliver data to any destination on the internet using only its own infrastructure and settlement-free peering agreements. This means:

1. No default route: Their routers contain explicit paths to all ~950,000 IPv4 networks on the internet. They can't just point unknown traffic "upstream" because there is no upstream.
2. Universal peering club: All Tier 1 networks must peer with each other for free. This works because they're roughly equal in value - each brings similar global reach and customer bases. If AT&T refused to peer with NTT, then NTT would need to buy transit from someone else to reach AT&T's customers, losing Tier 1 status.
3. Balanced traffic: Settlement-free peering only works when traffic exchange is relatively balanced (typically within a 2:1 ratio). If one network sends significantly more traffic than it receives, the receiving network will demand payment.

The Current Tier 1 Networks

As of 2025, approximately 16 networks maintain true Tier 1 status. Here are the major ones with their AS numbers and why they matter:

The Cogent Exception: AS174 (Cogent) claims Tier 1 status but has a history of peering disputes. They've been de-peered by some networks over traffic ratio disputes, technically breaking the "universal peering" requirement. They maintain operations by having enough peering to avoid buying transit, but their status remains disputed.

What This Means in Practice

When data travels between Tier 1 networks:

• No money changes hands for the bandwidth used
• Routing is direct between the networks at peering points
• Outages are critical - if peering breaks, there's no automatic backup route
• Political/business disputes can partition the internet - when Cogent and Level 3 de-peered in 2005, their customers couldn't reach each other for days

For network operators, peering with a Tier 1 means:

• You're getting routes to their direct customers only
• You still need transit (from them or others) to reach the rest of the internet
• Traffic ratios matter - send too much and they'll demand payment or disconnect

Technical Implementation

Here's a imagined example of a Tier 1's BGP configuration:

# Tier 1 BGP Configuration Example
router bgp 3356  # Lumen's ASN

# Peer with another Tier 1 (settlement-free)
neighbor 192.0.2.1 remote-as 701  # Verizon
neighbor 192.0.2.1 description Tier1-Peer-Verizon
neighbor 192.0.2.1 route-map TIER1-PEER-IN in
neighbor 192.0.2.1 route-map TIER1-PEER-OUT out

# The route-maps ensure:
# IN: Accept all routes from peer
# OUT: Send all customer routes + own routes (NOT other peer routes)

# Sell transit to a customer
neighbor 10.0.0.1 remote-as 65001  # Customer AS
neighbor 10.0.0.1 route-map CUSTOMER-IN in
neighbor 10.0.0.1 route-map FULL-ROUTES-OUT out
# OUT: Send full routing table (all peers + customers + own)

The key difference: Tier 1 networks send different routes to different relationships:

• To other Tier 1 peers: Only their customers' routes
• To customers: Everything (full internet routing table)
• To lower-tier peers: Usually just their customers' routes

Tier 2 Networks: The Regional Powers

How Tier 2 Networks Operate

Tier 2 networks buy some connectivity and get some for free through peering. They're in the middle of the hierarchy - big enough to peer with many networks, but not big enough to get settlement-free peering with all Tier 1s.

Think of Comcast: they're massive in the US with millions of customers. Netflix wants to peer with them to reach those customers efficiently. But Comcast still needs to buy transit from Level 3 or NTT to reach networks in Asia or Africa where they have no presence or peering agreements.

The Economics That Drive Tier 2 Decisions

A Tier 2 network constantly evaluates whether to peer or pay:

When they peer for free:

• With networks of similar size sending balanced traffic
• With content providers that their customers request data from
• At Internet Exchange Points where connection costs are low

When they must pay:

• For transit to reach the global internet (usually from Tier 1s)
• For "paid peering" with larger networks (when traffic is imbalanced)
• For backup routes to ensure redundancy

Major Tier 2 Examples and Their Strategies

Traffic Flow Example

Here's an imagined example of what could happen when a Comcast customer accesses a website in Japan:

1. Customer → Comcast network (Tier 2)
2. Comcast checks routing table:
   - Is destination in our network? No
   - Is it in a peer's network? No  
   - Use transit provider (Level 3)
3. Comcast → Level 3 (Tier 1) [PAID TRANSIT]
4. Level 3 → NTT (Tier 1) [FREE PEERING]
5. NTT → Japanese ISP (Tier 2/3) [CUSTOMER]
6. Japanese ISP → Web server

Comcast in this example pays Level 3 for this traffic. To reduce costs, Comcast might:

• Establish direct peering with Asian networks at exchange points
• Cache popular content locally
• Negotiate better transit rates with multiple providers

Technical Configuration

A typical Tier 2 network juggles multiple relationships:

router bgp 7922  # Comcast's ASN

# Transit providers (pay for full routes)
neighbor 10.0.0.1 remote-as 3356  # Level3 - primary transit
neighbor 10.0.0.1 route-map TRANSIT-IN in
# Receives: Full routing table (950,000+ routes)
# Sends: Only Comcast's customer routes

neighbor 10.0.1.1 remote-as 2914  # NTT - backup transit  
neighbor 10.0.1.1 weight 90  # Lower preference
# Backup uses weight/local-pref to avoid unless primary fails

# Settlement-free peer (e.g., similar-sized network)
neighbor 192.0.2.1 remote-as 7843  # Charter
neighbor 192.0.2.1 route-map PEER-IN in
# Receives: Charter's customer routes only (~30,000 routes)
# Sends: Comcast's customer routes only

# Paid peering (when traffic is imbalanced)
neighbor 198.51.100.1 remote-as 15169  # Google
# Google sends way more traffic (YouTube) than it receives
# Comcast charges Google for this imbalanced exchange

The Peering Decision Matrix

Tier 2 networks evaluate peering based on:

Traffic Ratio:

• Balanced (1:1 to 2:1): Free peering likely
• Imbalanced (3:1 to 5:1): Paid peering or rejection
• Heavily imbalanced (>5:1): Usually becomes transit customer

Business Value:

• Peer reduces transit costs by >$5,000/month: Worth connecting
• Peer improves latency to important content: Strategic value
• Peer is a competitor: May refuse on business grounds

Imagined Cost Example (for 10 Gbps of traffic):

• Via transit: $2,000/month
• Via free peering at IXP: $500/month (port cost only)
• Savings: $1,500/month or $18,000/year per 10G

Tier 3 Networks: The Last Mile

What Defines Tier 3

Tier 3 networks are pure customers - they buy all their internet connectivity from upstream providers. This includes:

• Your local ISP serving a few thousand customers
• Wireless ISPs (WISPs) covering rural areas
• Municipal broadband networks
• Enterprise networks that need internet access
• Mobile virtual network operators (MVNOs)

These networks focus on one thing: connecting end users to the internet. They don't have the scale or traffic to negotiate peering agreements, so they simply purchase transit.

Why Tier 3 Makes Economic Sense

For small networks, buying transit is often cheaper than peering:

Example: Small ISP with 1,000 customers using 500 Mbps average:

• Option 1: Buy 1 Gbps transit for $500/month - simple, one bill
• Option 2: Try to peer - would need presence at IXPs ($2,000/month), engineering staff, multiple connections
• Clear choice: Just buy transit

How Tier 3 Networks Connect

Most Tier 3 networks use simple dual-homing for redundancy:

# Typical Tier 3 BGP setup
router bgp 65001  # Small ISP's ASN

# Primary transit provider
neighbor 10.0.0.1 remote-as 7922  # Comcast
neighbor 10.0.0.1 route-map DEFAULT-ONLY in
neighbor 10.0.0.1 route-map LOCAL-ROUTES out

# Backup transit provider  
neighbor 10.0.1.1 remote-as 701   # Verizon
neighbor 10.0.1.1 route-map DEFAULT-ONLY in
neighbor 10.0.1.1 route-map LOCAL-ROUTES out
neighbor 10.0.1.1 weight 90  # Lower preference

# They receive just default routes, not full tables:
ip route 0.0.0.0 0.0.0.0 10.0.0.1  # All unknown traffic goes upstream

Why only default routes?

• Full routing table needs expensive routers with lots of memory
• Small ISP doesn't need to make complex routing decisions
• Simpler operations and troubleshooting

Real-World Tier 3 Examples

The Cost Reality for Tier 3

Pricing varies dramatically by location:

Good connectivity area (major US/EU city):

• 1 Gbps transit: $200-500/month
• 10 Gbps transit: $1,500-3,000/month
• Price per Mbps: $0.20-0.50

Poor connectivity area (rural/developing):

• 1 Gbps transit: $2,000-5,000/month
• 10 Gbps transit: Often unavailable
• Price per Mbps: $2-5

This pricing disparity explains why:

• Rural internet is expensive and slow
• Developing countries have limited broadband
• Some areas only have one viable ISP

Life as a Tier 3 Network

Challenges Tier 3 networks face:

1. No leverage: Must accept transit provider's terms
2. Support dependency: Rely on upstream for DDoS protection, routing issues
3. Thin margins: Compete on price with little differentiation
4. Last-mile costs: Most expensive part is reaching customers, not buying transit

Advantages they have:

1. Simplicity: No complex peering negotiations or traffic engineering
2. Low capital requirements: No need for expensive backbone infrastructure
3. Local focus: Can provide better customer service and local presence
4. Flexibility: Easy to switch transit providers if needed

Key Concepts: Peering vs Transit

The Fundamental Difference

Transit: You pay someone to carry your traffic to the entire internet. Like hiring a shipping company that delivers packages anywhere in the world.

Peering: Two networks agree to exchange traffic destined for each other's customers only. Like two neighboring businesses agreeing to hand-deliver packages to each other instead of using the postal service.

Settlement-Free Peering

When two networks exchange traffic without payment. This works when:

• Traffic is roughly balanced (neither network is doing more work)
• Both networks benefit equally (similar value exchange)
• Both have similar bargaining power

Example: AT&T and Verizon peer for free because they're similar sized, send similar traffic amounts, and both save money by directly exchanging traffic.

Paid Peering (Not Transit!)

One network pays another to exchange traffic, but only for that network's customers, not the whole internet. This happens when:

• Traffic is heavily imbalanced (one network sends much more)
• One network has more market power
• The paying network still saves money vs buying transit

Example: Netflix pays Comcast for peering because:

• Netflix sends 100x more traffic than it receives (all that video streaming)
• Comcast has leverage (controls access to millions of households)
• Netflix still saves money vs sending that traffic through transit providers
• Netflix still needs other arrangements (transit or peering) to reach non-Comcast users

Here's the key example distinction:

# PAID PEERING - Netflix to Comcast
# Netflix receives: Routes to Comcast customers only (~50,000 prefixes)
# Netflix pays: $0.50 per Mbps
# Can reach: Only Comcast's network

# TRANSIT - Netflix buying from Level 3
# Netflix receives: Routes to entire internet (~950,000 prefixes)  
# Netflix pays: $1.00 per Mbps
# Can reach: Everything

Internet Exchange Points (IXPs)

Physical locations where many networks connect to exchange traffic. Think of them as network meetup points where instead of connecting cable between every pair of networks, everyone connects to a shared switch.

Major IXPs and what they mean:

Connecting at an IXP costs:

• Port fee: $500-5,000/month depending on speed
• Cross-connect: $200-500/month for the physical cable
• Benefits: Can peer with hundreds of networks through one connection

BGP and Autonomous Systems

Autonomous System (AS): A network under single administrative control. Every ISP, large company, or content provider has an AS number (ASN).

BGP (Border Gateway Protocol): The protocol that determines how data routes between ASes. Think of it as the GPS system of the internet - it figures out the best path between networks.

AS Path: The list of networks data traverses. Like a passport showing which countries you passed through:

Traceroute from you to google.com:
AS64512 (Your ISP) → AS7922 (Comcast) → AS15169 (Google)

Shorter AS paths usually mean:

• Lower latency (fewer networks to traverse)
• Better reliability (fewer potential failure points)
• Lower cost (fewer business relationships involved)

Traffic Engineering Terms

Hot-Potato Routing: Get rid of traffic as quickly as possible. Networks hand off traffic at the nearest connection point to minimize their own costs.

Example: European traffic to Asia via US carrier:

• Hot-potato: US carrier hands off in New York (closest point)
• Traffic then travels across European network to Asia

Cold-Potato Routing: Carry traffic as far as possible on your own network. Better for quality but costs more.

Example: Same European traffic:

• Cold-potato: US carrier carries traffic to Los Angeles
• Hands off to Asian carrier there (closer to destination)

95th Percentile Billing: How transit is usually billed. Measure bandwidth every 5 minutes for a month, throw away the top 5% of samples (36 hours), bill on the remaining highest usage. This allows for occasional bursts without penalty.

CDN Integration

Content Delivery Networks have changed the game by putting content directly inside ISP networks:

Traditional model:

User → ISP → Transit → Transit → Origin Server

CDN model:

User → ISP → Cache inside ISP (no transit needed!)

This is why Netflix offers free cache servers to ISPs - it saves both parties money and improves quality.

How the Internet Hierarchy Is Changing

Hyperscale Networks: The New Power Players

Google, Meta, Microsoft, and Amazon have built their own global networks that break traditional rules. They're not Tier 1 (they don't peer with ALL Tier 1s), but they don't need to be - they have something better: the content everyone wants.

Traditional model (2000s):

Content → Web hosting → Tier 3 ISP → Tier 2 → Tier 1 → Tier 1 → Tier 2 → User's ISP → User

Hyperscale model (2020s):

Content → Google's network → Direct peering with user's ISP → User

This matters because:

• Lower latency: YouTube videos start faster with fewer hops
• Better quality: No congested transit links means less buffering
• Cost savings: ISPs get expensive content traffic for free
• Power shift: Content providers now have leverage over ISPs

CDNs Have Flattened the Hierarchy

In 2010, if you wanted to reach users globally, you needed transit from Tier 1 networks. Today, CDNs like Cloudflare and Akamai have servers inside thousands of ISP networks worldwide.

The numbers tell the story:

• Netflix + YouTube = 35% of global internet traffic
• All CDN traffic combined = 70% of global traffic
• Traditional transit = <30% and declining

This means:

• Small ISPs can deliver Netflix without expensive transit
• Tier 1 networks carry less profitable traffic
• Peering with CDNs matters more than Tier 1 peering

Internet Exchange Growth Changes Everything

IXPs have exploded, especially outside traditional hubs:

What changed (2015 → 2025):

• Africa: 20 → 45 IXPs (enabled local content hosting)
• Global: 400 → 800+ IXPs (more direct paths)
• Average members: 50 → 150 (more peering opportunities)

Real impact example: Kenya's KIXP (2000-2020)

• Before: ISPs paid $18,000/Mbps for satellite transit
• After: Local peering at $0.35/Mbps, international at $200/Mbps
• Result: 93% cost reduction, 50% price drop for users, 10-20x usage increase

Regional Differences Matter

United States:

• Problem: Few IXPs, networks prefer private peering
• Result: Higher costs, peering disputes (Netflix vs Comcast)
• Market dominated by 4-5 large players

Europe:

• Strong IXP culture (DE-CIX, AMS-IX, LINX)
• Regulations promoting competition
• Result: Lower prices, more ISP choices

Germany's Peering Controversy: Deutsche Telekom (Tier 1 globally) refuses to peer at DE-CIX for domestic traffic, forcing German ISPs to pay for transit to reach Deutsche Telekom customers in the same city. This shows how business decisions can override technical logic.

Africa:

• Challenge: Expensive international transit ($30-100/Mbps)
• Solution: Rapidly building IXPs and keeping traffic local
• Impact: Internet becoming affordable for first time

The Real Cost of Internet Access

Transit pricing by location (1 Gbps commitment, 2025):

This explains global digital inequality - a gigabit in Lagos costs what 17 gigabits cost in Frankfurt.

How Data Actually Flows Between Networks

The Path Your Data Takes

When you access a website, your data might traverse multiple tiers. Here's a real example of accessing a European site from the US:

traceroute to www.example.de
1. 192.168.1.1 (Your router) - 1ms
2. 10.0.0.1 (ISP equipment in your neighborhood) - 5ms  
3. 72.14.0.1 (Regional ISP aggregation) - 8ms
4. 198.32.118.1 (Your ISP → Tier 2 Comcast) - 12ms
5. 66.208.228.1 (Comcast → Tier 1 Telia) - 15ms [TRANSIT PAID HERE]
6. 62.115.0.1 (Telia crossing Atlantic) - 95ms
7. 80.91.0.1 (Telia → Deutsche Telekom) - 97ms [PEERING]
8. 217.5.0.1 (Deutsche Telekom → Customer) - 99ms
9. www.example.de - 100ms

What this tells us:

• 4 different networks handled your packet
• Major latency jump (+80ms) at ocean crossing
• Transit money changed hands at step 5
• Total path: Your ISP paid Comcast, Comcast paid Telia

How Networks Decide Where to Send Traffic

BGP makes routing decisions using a priority list. Think of it like GPS with preferences:

1. Local Preference (manually set priorities)
   • "Always prefer the cheap route through peer X"
2. Shortest AS Path (fewer networks to cross)
   • Like choosing fewer connecting flights
3. Best exit point (MED values)
   • The other network's suggestion: "Please enter through Chicago, not New York"

Here's what happens when multiple paths exist:

Route to YouTube (AS15169) from your ISP:
Option 1: ISP → Comcast → Google (AS path: 2 hops) ✓ CHOSEN
Option 2: ISP → Level3 → NTT → Google (AS path: 3 hops)
Option 3: ISP → Local IXP → Google (AS path: 1 hop, but no IXP available)

Why Your Internet Sometimes Sucks

Common bottlenecks in the tier system:

Congested Peering Points:

• Two networks peer but don't upgrade capacity
• Like a highway narrowing from 6 lanes to 2
• Symptoms: Buffering during peak hours (7-11 PM)
• Real example: Netflix vs ISPs (2014) - streams buffered until peering was upgraded

Hot-Potato Routing Gone Wrong:

• Networks dump traffic ASAP to save money
• Your data takes inefficient paths
• Example: Dallas to Houston traffic going through Chicago

Single Transit Provider:

• Tier 3 ISP has only one upstream
• That link fails = internet dies
• Common in rural areas with one fiber path

Real Performance Impact

Actual measurements across tiers:

Packet loss by scenario:

• Within single network: <0.01% (well-managed)
• Across peering link: 0.01-0.1% (normal)
• Congested peering: 1-5% (video buffering starts)
• Overloaded transit: 0.5-2% (noticeable degradation)

Traffic Engineering in Practice

Networks manipulate routing to achieve business goals:

Making traffic take a specific path:

# Force traffic through expensive transit provider last
route-map PREFER-CHEAP-PEER permit 10
  set local-preference 150  # Higher = more preferred

route-map USE-EXPENSIVE-TRANSIT permit 10  
  set local-preference 50   # Lower = use only if necessary

Telling others to avoid your network (during maintenance):

# AS Path prepending - make path look longer
route-map MAINTENANCE permit 10
  set as-path prepend 64512 64512 64512
# Other networks see: "64512 64512 64512 64512" and choose alternate paths

Emergency traffic blocking (DDoS response):

# Blackhole community - tell upstream to drop traffic
ip route 192.0.2.1/32 null0
router bgp 64512
  network 192.0.2.1/32 route-map BLACKHOLE
  
route-map BLACKHOLE permit 10
  set community 65535:666  # Universal "please drop this" signal

The Economics of Internet Connectivity

How Networks Make Money

The tier system is about who pays whom:

Tier 1 Business Model:

• Revenue: Selling transit to Tier 2/3 networks ($millions/month from large customers)
• Costs: Infrastructure, submarine cables, data centers
• Margins: 20-40% EBITDA typical
• Key: Must maintain Tier 1 status or lose negotiating power

Tier 2 Business Model:

• Revenue: Selling transit to smaller ISPs and enterprises
• Costs: Transit from Tier 1s, infrastructure, peering
• Strategy: Minimize transit costs through peering
• Example: Comcast saves ~$50M/year through aggressive peering

Tier 3 Business Model:

• Revenue: Monthly subscriber fees
• Costs: Transit (often 30-50% of operating costs)
• Margins: Often <10% due to transit costs
• Survival: Compete on customer service, not infrastructure

Real Transit Pricing (2025)

What networks actually pay for bandwidth:

Tier 2 buying from Tier 1 (estimated, please verify on your own):

Volume          Price per Mbps   Monthly cost for commitment
100 Mbps        $2.00           $200
1 Gbps          $0.50           $500
10 Gbps         $0.30           $3,000
100 Gbps        $0.15           $15,000
1 Tbps          $0.08           $80,000

Tier 3 buying from Tier 2 (marked up 2-5x, estimated, please verify on your own):

100 Mbps        $5.00           $500
1 Gbps          $2.00           $2,000
10 Gbps         $1.00           $10,000

The 95th Percentile Billing Game

Networks don't pay for peak usage - they pay for sustained usage:

How it works:

1. Measure bandwidth every 5 minutes (8,640 samples/month)
2. Sort all samples from lowest to highest
3. Discard top 5% (432 samples = 36 hours)
4. Bill on the highest remaining sample

What this means:

• Can burst to 10 Gbps for up to 36 hours/month
• Only pay for 1 Gbps commitment
• Smart networks "bank" their burst hours for peak events

Gaming the system:

# Traffic shaping to stay under commitment
if current_hour_burst_usage < 36_hours:
    allow_full_speed()  # Use burst allowance
else:
    rate_limit_to_commitment()  # Avoid overage charges

Peering Economics: When Free Exchange Makes Sense

Networks constantly evaluate: Is peering cheaper than transit?

Example calculation - Tier 2 network considering joining an IXP:

Costs (estimated, please verify on your own):

• IXP port (10G): $2,000/month
• Cross-connects: $500/month
• Engineering: $1,000/month (amortized)
• Total: $3,500/month

Benefits (estimated, please verify on your own):

• Traffic offloaded to peering: 5 Gbps average
• Transit cost avoided: 5 Gbps × $0.50 = $2,500/month
• Net loss: $1,000/month

But wait - other benefits:

• Lower latency to peers (better customer experience)
• Redundancy (not dependent on single transit)
• Access to content networks (Netflix, Google)
• Marketing: "We peer at major IXPs"

Decision: Join the IXP despite direct loss.

The Submarine Cable Game

Who owns the pipes matters:

Traditional model (2000s):

• Consortiums of telcos share costs
• Each gets capacity proportion
• Sell excess as commodity

New model (2020s):

• Google/Meta build private cables
• Keep 80%+ capacity for themselves
• changes power dynamics

Cable costs and control:

Result: Content providers becoming infrastructure owners, disrupting traditional tiers.

Future Technical Developments

What's Actually Changing

400G/800G Ethernet Adoption: Current reality vs. hype:

• 100G ports: Still 80% of backbone (mature, cheap)
• 400G ports: 15% adoption, mainly Tier 1 core routes
• 800G: Still testing, won't be mainstream until 2027+

Why it matters:

• Cost per bit dropping 40% with each generation
• Single 400G port replaces four 100G ports
• Power usage only doubles (not quadruples)

Segment Routing - Simplifying Traffic Engineering: Instead of complex MPLS tunnels, networks can specify exact paths:

Traditional: Configure tunnels at every router
Segment Routing: "Go via router A, then B, then C"

Adoption: 35% of Tier 1s, but most Tier 2/3 don't need it.

Satellite Internet's Impact on Tiers

LEO constellations are bypassing traditional infrastructure:

Starlink's model:

• User terminal → Satellite → Ground station → SpaceX network → Peering/Transit
• Completely bypasses local Tier 3 ISP
• Ground stations need Tier 1/2 connectivity

Impact on traditional ISPs:

• Rural Tier 3 ISPs losing customers to Starlink
• Urban ISPs unaffected (fiber still faster/cheaper)
• New peering relationships at ground station locations

Real performance (2025):

Edge Computing Changes Peering

5G and edge computing put servers at cell towers:

Traditional cloud path:

Phone → Tower → Backhaul → Core Network → Internet → Cloud (50-100ms)

Edge computing path:

Phone → Tower → Edge Server at tower (5-10ms)

This means:

• Content cached at thousands of edge locations
• Less traffic traversing tier networks
• ISPs becoming compute providers, not just connectivity

Regulation and Data Sovereignty

Governments are fragmenting the global internet:

Data Localization Requirements:

• Russia: Data must stay in country
• China: Great Firewall controls all international traffic
• EU: GDPR requires data handling compliance
• India: Payment data cannot leave country

Technical implementation:

def route_traffic(source, destination, data_type):
    if data_type == "payment" and source.country == "IN":
        # Must stay within India
        return find_domestic_path(destination)
    elif data_type == "personal" and source.region == "EU":
        # Must comply with GDPR
        return find_gdpr_compliant_path(destination)
    else:
        return find_optimal_path(destination)

Result: Networks building regional infrastructure instead of relying on global Tier 1s.

Carbon-Aware Routing (Experimental)

Networks starting to consider power sources:

def carbon_aware_route_selection(paths):
    for path in paths:
        path.carbon_score = calculate_carbon(
            path.distance,
            path.network.power_source,
            path.network.pue  # Power Usage Effectiveness
        )
    
    # Balance carbon impact with performance
    return min(paths, key=lambda p: p.carbon_score * p.latency)

Real implementation challenges:

• Renewable energy varies by time of day
• Longer paths might use cleaner energy
• Customers won't accept degraded performance

Current adoption: <1% of networks experimenting

What Actually Matters for the Future

Consolidation continuing:

• Expect 10-12 Tier 1 networks by 2030 (down from 16)
• Regional Tier 2s being acquired by larger players
• Small Tier 3s going out of business or merging

Hyperscale dominance:

• Google/Meta/Amazon/Microsoft control majority of new submarine cables
• Direct relationships with eyeball networks, bypassing traditional transit
• Building their own last-mile infrastructure in some markets

IXP growth in developing markets:

• Every African country will have an IXP by 2030
• Local content hosting becoming viable
• Transit prices dropping 90% when local peering established

Technical skills shortage:

• BGP/networking expertise concentrated in few companies
• Automation reducing need for network engineers
• Tier 3 ISPs struggling to find qualified staff