What Is IP Transit? A Practical Guide for ISPs and Hosting Providers
Until your address space is announced to the world and traffic can flow both ways, your network is an island. IP transit is the bridge to the mainland.
Artem Kohanevich
Co-Founder & CEO at IPbnb
Last updated
Table of Contents
item
You can rack the servers, wire the switches, and light every port - and the internet still will not know you exist. Until your address space is announced to the world and traffic can flow both ways, your network is an island. IP transit is the bridge to the mainland. It is the service nearly every ISP, hosting company, and data center buys to reach the rest of the internet, and to let the rest of the internet reach them.
It is also where a surprising share of the budget goes, and where quiet mistakes get expensive. What are you actually paying for when a quote lands at "$0.30 per Mbps"? When does peering earn its keep? And what has to be in place before you can turn transit up at all? This guide answers each of those, with particular attention to the resource that stalls more rollouts than any router ever will: IPv4 space.
The short version: transit gives your network universal reachability for a per-Mbps fee; peering supplements it later to cut costs. To buy transit you need an ASN, a BGP-capable router, an upstream agreement, and - the usual bottleneck - IPv4 address space you are allowed to announce.
What Is IP Transit?
IP transit is a commercial service in which an upstream provider gives your network access to the full internet routing table and carries your traffic across their backbone to reach every other network. In plain terms, you are buying reachability: the ability to send packets to any destination online and receive the replies.
The provider does two jobs for you. It accepts your outbound traffic and routes it onward toward its destination, and it announces your IP prefixes to its own peers and upstreams so that return traffic finds its way back to you. Miss that second half and you could send packets out while nothing could reach you.
A useful analogy: transit is a toll road. You pay for the right to travel anywhere the road network connects, and the operator maintains the on-ramps and the long-haul routes. Peering, by contrast, is a private bridge you build directly to a neighbor so the two of you can swap traffic without paying a toll. Most operators run both, but almost everyone starts with transit, because it delivers reach to the entire internet from day one.
How IP Transit Works
When you buy transit, you connect your router to your provider's router - physically over a cross connect in a shared facility, or across a longer-haul circuit - and you establish a BGP session between the two. That session is the channel through which routing information flows in both directions, as the diagram above shows.
Over that session, your provider sends you the full routing table: every reachable destination on the internet. This is not a small object. The global IPv4 table passed one million routes in 2025, and with the IPv6 table on top, a full feed now runs to well over a million entries. Your router uses it to decide where to forward outbound traffic. In the other direction, you announce the IP blocks you control, and your provider propagates those announcements outward, which is what makes your address space globally reachable.
Four things have to be in place for this to work, and we will return to each in detail:
An Autonomous System Number (ASN) that uniquely identifies your network in the global routing system.
IPv4 and/or IPv6 address blocks you are authorized to announce - a minimum of a /24 for IPv4, because most operators filter anything smaller out of the global table.
A BGP-capable router with enough memory to hold the full table.
An upstream agreement with one or more transit providers.
Most serious operators buy transit from at least two upstreams. Being multi-homed means that if one provider suffers a fiber cut or an outage, traffic fails over to the other, and it gives you leverage on both routing and price. Single-homed transit is cheaper and simpler, but it turns that one provider into a single point of failure for your entire operation.
IP Transit vs Peering
Transit and peering are often framed as rivals, but they solve different problems, and most networks run both. Transit buys reachability everywhere. Peering lets you exchange traffic directly with one specific other network, usually to cut costs and shorten the path for high-volume destinations.
IP transit | Peering | |
Cost | Paid, priced per Mbps | Often settlement-free, or minimal port cost |
Routing scope | Full internet routing table | Only the peer's own routes and its customers |
Availability | Any operator can buy it | Requires an IX presence or a direct interconnect |
Best for | Universal reachability, day-one connectivity | High-volume traffic to specific large networks |
Requirements | ASN + IP block + upstream agreement | ASN + IP block + IX membership or private interconnect |
The practical pattern goes like this. A growing hosting provider or ISP starts with transit from one or two upstreams, because it connects them to the whole internet immediately. As volume climbs, they join an internet exchange and begin peering with the networks they trade the most traffic with - large content platforms, CDNs, eyeball networks. Every bit that shifts to settlement-free peering is a bit they no longer pay transit for, so peering becomes a cost-optimization layer on top of transit, not a replacement for it. You can see who peers where, and at which exchanges, through a directory like PeeringDB.
IP Transit Pricing
Transit is priced per megabit per second, per month, and the spread is wide enough to be genuinely confusing without context. In 2026, published rates run from a few cents to several dollars per Mbps, driven by four levers: how much bandwidth you commit to, where you buy it, your provider's tier, and the service guarantees attached.
Billing models. Three structures dominate:
Model | How you are billed | Best for |
Committed rate (CDR) | A fixed monthly fee for a set bandwidth commit, regardless of use | Steady, predictable traffic |
95th percentile (burstable) | Usage is sampled every 5 minutes; the top 5% of readings are discarded; you are billed on the 95th-percentile peak | Variable traffic with occasional spikes |
Flat-rate port | A fixed fee for the full capacity of the port | Simplicity, smaller networks |
What it actually costs. Volume and geography move the number more than anything else. The figures below reflect 2025 market data:
Buyer profile | Typical commit | Price per Mbps / month |
Large network, top-tier hub | 100 Gbps | as low as ~$0.05 (market floor) |
Mid-size ISP or host, major hub | 10 Gbps+ | ~$0.30 - $0.80 |
Smaller network | Sub-1 Gbps | ~$1.50 - $3.50 |
Any size, emerging or remote market | Varies | Several times the hub price |
Those floor prices - the lowest 100 GigE rates in the most competitive markets held around $0.05 per Mbps through 2025 - apply to very large commitments in mature hubs like London, Frankfurt, and Singapore. Step down in volume, or move to a market with limited subsea cable access, and the per-Mbps rate climbs quickly.
Provider tiers shape both price and reach. Tier 1 carriers own global backbones and reach the entire internet without buying transit themselves, which brings broad reach though not always the cheapest quote. Tier 2 providers blend their own network with purchased upstream capacity and often compete hard on price. Tier 3 operators are regional and buy most or all of their transit. The right tier depends on where your traffic goes and how much redundancy you need - not on the label.
The trend runs steadily downward: across major cities, 100 GigE prices fell roughly 12% per year, compounded, from 2022 to 2025, and erosion is expected to continue. One caveat worth planning around: because traffic volumes grow faster than unit prices fall, your total transit bill can rise even as your per-Mbps rate drops. Size your commitment to real, sustained peaks - typically three to six months of traffic data plus a 20-30% growth buffer - rather than over-committing to chase a lower unit rate.
How to Choose an IP Transit Provider
Once the pricing makes sense, choosing a provider comes down to matching a network to your real traffic and reliability needs. Price per Mbps is the headline; it is rarely what decides whether you are happy twelve months in. Run any candidate through this checklist:
# | Criterion | What to check |
1 | Network coverage and PoPs | Where their points of presence sit relative to your facilities and your users - fewer hops means lower latency |
2 | Route quality | The actual paths to the destinations that matter to you; test them if you can |
3 | Pricing model | Whether CDR, 95th percentile, or flat-rate fits your traffic profile |
4 | SLA | Guaranteed uptime, latency, packet loss, and what the provider pays out when it misses |
5 | BGP community support | Whether you can steer routing and control how your prefixes are advertised |
6 | IPv4 and IPv6 support | Full dual-stack, not a bolt-on - non-negotiable in 2026 |
7 | DDoS protection | Whether mitigation is included, an add-on, or entirely your problem |
Two providers can quote the same price and deliver very different service, so weigh the whole picture. You can sanity-check a candidate's real-world connectivity and peering footprint independently through PeeringDB, and inspect how their announcements look in the routing system with RIPEstat.
What You Need Before Buying IP Transit
A transit agreement is only useful once you have the pieces that let you actually announce and route traffic. Four requirements sit between you and a live connection - three are straightforward, and one is the real bottleneck.
An ASN. Your Autonomous System Number is your network's identity in the global routing system. In the RIPE region you obtain one through the RIPE NCC, usually via a Local Internet Registry (LIR) membership or a sponsoring LIR. It is a well-defined process with a modest cost and lead time.
A BGP-capable router. Hardware or software that can hold the full table and maintain stable BGP sessions with your upstreams. A solved problem at almost any budget.
An upstream agreement. The transit contract itself, ideally with two providers for redundancy.
An IPv4 address block - the piece that stalls rollouts. To announce routes you need address space you are authorized to originate, and for IPv4 the practical minimum is a /24, since most networks filter longer prefixes. The catch is supply. The RIPE NCC exhausted its freely allocatable IPv4 pool in November 2019, so you can no longer simply request a fresh block. That leaves three routes to the space you need, and they are not equivalent:
Route to IPv4 | Speed | Cost structure | Commitment |
RIPE waiting list | 12 - 24 months for a single /24 | Recurring LIR fees | One /24 maximum; 24-month transfer lock |
Buy on the transfer market | Weeks | Large upfront capital | You own it, but cannot re-transfer for 24 months |
Lease | Days | Predictable monthly cost | Flexible; scale up or down as needed |
Two constraints tighten this further. The RIPE waiting list gives a new LIR just one /24, and current waits run past a year. And under RIPE policy (RIPE-807), IPv4 addresses cannot be transferred for 24 months from the date they are received - whether by allocation, transfer, or a change in business structure. That lock ties up capital and rules out quick reallocation, which is a large part of why a growing number of ISPs and hosting providers lease the IPv4 blocks they announce rather than buy them.
If you need space ready to announce, you can lease IPv4 blocks through IPbnb, with clean routing history and RPKI ROAs already in place. For sizing, see our guide to IPv4 block sizes; for the supply picture, IPv4 address exhaustion; and if you are weighing membership against leasing, RIPE NCC membership costs, benefits, and pitfalls breaks down the trade-off.
IP Transit for Hosting Providers
Hosting providers feel the IPv4 squeeze more sharply than almost anyone, because address space is not just infrastructure for them - it is inventory. Dedicated IPs for customers, shared-hosting pools, mail servers that need clean reputation, CDN nodes, per-server addressing: all of it draws on the same finite supply. A host scaling from a rack to a floor can go from needing a /24 to needing a /20 or larger within a few quarters.
That creates a specific risk that pure transit buyers rarely think about: reputation. When you announce IPv4 space for hosting, the routing and blocklist history of those addresses feeds directly into deliverability and customer experience. A block with a history of abuse can land your customers on spam blocklists before they send their first email. This is where leasing from a vetted source beats scraping together whatever cheap space is on offer - a reputable marketplace checks the blocks it lists and provides RPKI ROAs, so your announcements can be cryptographically validated. That cuts both hijack risk and the odds of inheriting someone else's bad reputation. (Our explainer on RPKI and BGP hijacking prevention covers the mechanics.)
The pattern that works for most growing hosts combines three things: transit for universal reachability, peering at an exchange once volume justifies it, and leased IPv4 that scales with customer demand instead of forcing a large purchase every time the business grows. Reachability, cost, and inventory stay aligned as you expand.
Frequently Asked Questions
How does IP transit work in simple terms?
Your provider announces your IP address blocks to the rest of the internet over BGP and carries your traffic across their backbone in both directions. You send them everything destined for the wider internet; they hand you the full routing table so you know how to reach any destination.
What is the difference between IP transit and bandwidth?
"Bandwidth" describes capacity - how many bits per second a link can carry. IP transit is the service that turns that capacity into global reachability by routing your traffic to and from every other network. You are paying for reach, measured and billed per Mbps.
What is an IP transit agreement?
It is the commercial contract with your upstream provider. It typically specifies the committed bandwidth, the billing model (committed rate, 95th percentile, or flat-rate), the SLA for uptime and performance, and the terms for announcing your prefixes.
Do I need my own IP addresses to buy transit?
Yes. To announce routes and be reachable, you need address space you are authorized to originate - a /24 or larger for IPv4. You can lease it, buy it on the transfer market, or wait for a RIPE allocation.
How much does IP transit cost?
Anywhere from about $0.05 per Mbps for very large commitments in competitive hubs to $1.50 - $3.50 or more per Mbps for smaller commits and remote markets. Volume, location, provider tier, and SLA are the main drivers.
Building or Growing a Transit Network?
If you are standing up or expanding a transit network, the IPv4 block is usually the piece that takes the longest to solve. It does not have to be.
Need IPv4 for your transit network? Lease clean, ready-to-announce IPv4 blocks through IPbnb - with routing history and ROA records handled, so you can bring up BGP quickly instead of waiting on a purchase or a place in the queue.
Sitting on IPv4 you are not using? If your allocation includes idle blocks, hosting providers and ISPs building transit networks need exactly that space. Monetize your idle IPv4 through IPbnb and turn dormant addresses into recurring revenue.
