Dissecting the Cost of Peering: Global Evidence from Internet Exchange Points (IXPs)
An ISOC LIVE summary
Authors: Anirudh Tagat — The Internet Society; Amreesh Phokeer — The Internet Society. Published in Telecommunications Policy (Vol. 50, 2026, Article 103253), open access (CC BY-NC-ND). The work was funded by the Internet Society.
This paper examines the economics of Internet peering through Internet Exchange Points (IXPs), comparing the cost of peering with traditional IP Transit arrangements and analyzing how the difference affects broadband markets, connectivity, and telecommunications outcomes globally. Using data from 2019–2024, the authors combine publicly available IXP port pricing with proprietary IP Transit pricing data to evaluate whether peering produces measurable economic and infrastructure benefits. The authors describe it as the first paper to directly compare peering and IP Transit pricing on a global basis.
Background and Motivation
The Internet consists of more than 80,000 independently operated networks, known as Autonomous Systems (ASNs), which interconnect either through IP Transit arrangements or through peering relationships. Traditionally, IP Transit dominated Internet interconnection, but over the past two decades the Internet has “flattened” as more networks exchange traffic directly through peer-to-peer arrangements at IXPs.
The paper explains that peering allows ISPs to exchange traffic locally, improving performance, resilience, and reachability while reducing dependence on expensive international transit links. IXPs enable this by providing shared infrastructure where multiple operators interconnect. The paper reports two related counts as of March 2026: roughly 1,051 active IXPs worldwide per Internet Society Pulse (intro), and 1,051 IXPs listed in PeeringDB (data section).
The authors argue that peering decisions are shaped by two major incentives:
Lower interconnection costs relative to IP Transit
The ability to keep local traffic local through traffic localization
They emphasize that while peering is often described as “settlement-free,” operators still incur significant recurring costs, especially IXP port fees, colocation, cross-connects, transport/backhaul, and peering equipment. A non-representative survey of 20 IXPs found that recurring port fees alone represented an average reported share of about 50% of peering-related costs, with the top five cost heads accounting for roughly 84.5% of reported costs. The authors caution that this small sample cannot be treated as a proxy for the true economic cost of peering; throughout the paper, “cost of peering” refers specifically to IXP port fees.
Theoretical Framework
The paper develops a simplified model of ISP decision-making between peering and IP Transit. The authors define total interconnection cost as a function of:
The share of traffic that can be localized domestically
The cost of IP Transit
The cost of peering through an IXP
They derive a core proposition: the larger the difference between IP Transit prices and peering prices, the stronger the economic incentive for ISPs to peer. Similarly, higher levels of localizable traffic increase the attractiveness of peering.
The model also predicts “scale effects,” where larger traffic volumes amplify total savings from peering. However, the authors note that transit remains necessary for non-local traffic and resilience, meaning peering complements rather than fully replaces transit.
Data Sources and Methodology
The analysis combines multiple global datasets:
Telegeography IP Transit Pricing Suite for transit costs
peering.exposed for IXP port pricing
Packet Clearing House (PCH) data for IXP participants, traffic, and bandwidth
Internet Society Pulse data for peering intensity and resilience indicators
ITU Datahub indicators on broadband pricing, telecom revenues, and network coverage
Inclusive Internet Index data on broadband policy and market concentration
The authors define “unit savings from peering” as the difference between the per-Mbps cost of IP Transit and the equivalent per-Mbps cost of peering, measured at the same city-year-circuit level.
They also construct a “peering intensity” variable measuring the share of active ASNs participating in domestic IXPs. Econometric models (OLS with year fixed effects, and a two-stage least squares specification) then estimate how cost savings and traffic localization relate to Internet-related outcomes such as:
Broadband prices
Mobile network coverage
Telecom revenues
Average revenue per user (ARPU)
Latency
Key Findings
The study finds that peering costs are dramatically lower than IP Transit costs worldwide. Average peering cost was estimated at approximately USD 0.054 per Mbps/month, compared with USD 18.2 for IP Transit. The authors caution that the savings indicator is not the simple difference of these two averages, because the country-years for which both figures are available differ; the mean transit-minus-peering difference reported in their summary statistics is USD 1.8 per Mbps/month.
The authors identify several major downstream effects associated with peering cost savings:
Lower Broadband Prices
A one-percentage-point increase in savings from peering was associated with an average USD 4.5 decline in monthly broadband prices on the ITU price measure, with a smaller corresponding decline on the Inclusive Internet Index measure. The authors note that retail prices depend on many factors beyond interconnection costs (spectrum, energy, taxation, deployment), and frame the result as a change in the underlying cost structure rather than evidence of automatic pass-through to consumers.
Expanded Mobile Network Coverage
Greater savings from peering were associated with approximately a 4.8 percentage-point increase in mobile network coverage. The heterogeneity analysis indicates this effect is driven mainly by countries with high peering readiness and by LMICs, rather than holding uniformly across all markets.
Increased Telecom Revenues
The paper reports that a one-percentage-point increase in peering-related savings was associated with roughly a 2.3 percentage-point increase in telecom revenues (from the predicted-savings regression). Interestingly, these revenue gains did not significantly increase ARPU, suggesting growth came through expanded adoption and broader market reach rather than charging existing customers more.
Traffic Localization Effects
In a separate, smaller-sample specification where peering intensity was instrumented by local traffic exchange rather than cost alone, the paper found a large, statistically significant increase in ARPU (an 83-percentage-point increase per unit increase in peering intensity) and indications of higher telecom revenues and mobile coverage, though several of those estimates lacked statistical significance. These results come from different model specifications and samples than the 2.3-point revenue figure above.
The authors did not find strong evidence that increased peering significantly reduced mobile latency in the short run.
Regional and Policy Differences
The paper finds important heterogeneity across countries and regulatory environments.
The positive effects of peering savings were strongest in:
Moderately regulated economies
Low- and lower-middle-income countries
Countries with growing but not fully mature peering ecosystems
The authors interpret this as evidence that developing economies may derive especially large benefits from IXP development and traffic localization.
Countries with stronger peering readiness tended to see larger reductions in broadband prices and stronger improvements in mobile coverage. Meanwhile, countries with lower peering readiness experienced more pronounced telecom revenue gains from peering savings.
Limitations and Caveats
The authors repeatedly stress that their results are associative and structural rather than causal. They acknowledge several important limitations:
Limited overlap in available datasets across years and regions
Difficulty identifying truly exogenous variation in peering intensity
Potential reverse causality between traffic growth and peering
The peering.exposed data captures essentially only port fees, excluding backhaul, transport, redundancy, colocation, cross-connect, and operational costs
Retail broadband prices being influenced by many additional factors, including spectrum, energy, taxation, and infrastructure deployment costs
Potential oversampling of countries where IXPs already exist, since peering pricing relies on crowdsourced data
They also note that peering decisions are affected by strategic behavior, bargaining dynamics, market concentration, and regulatory environments beyond simple cost considerations.
Policy Implications
The paper concludes that IXPs and local peering arrangements are likely important contributors to more affordable, resilient, and accessible Internet ecosystems. The findings suggest governments and regulators should:
Support IXP development and integrate it into national broadband strategies
Encourage transparent interconnection markets
Promote neutral, open, voluntary, and commercially sustainable interconnection
Improve local traffic localization
Reduce barriers to interconnection infrastructure
Importantly, the authors caution against an “all-out focus on IXPs” at the expense of IP Transit, private interconnects, CDNs, caches, domestic backbone, and access-network investment — IXPs are one building block within a holistic policy approach. They argue these measures could be especially impactful in developing economies, where savings from peering appear most strongly associated with improved telecommunications outcomes.
RESOURCES
Dissecting the cost of peering: Global evidence from internet exchange points (IXPs) — the paper itself, open access in Telecommunications Policy (Jun 2026)
Anirudh Tagat — co-author; economist and former In-Residence Economist at the Internet Society (ISOC)
Amreesh Phokeer — co-author; Internet Measurement and Data Expert at the Internet Society (ISOC)
The Internet Society (ISOC) — author affiliation and funder of the study
peering.exposed — crowdsourced IXP port-pricing dataset used for peering costs
PeeringDB — community-maintained interconnection database; source of the paper’s 1,051-IXP count
Packet Clearing House — source for IXP participant, traffic, and bandwidth data
ISOC Pulse IXP Tracker — peering-intensity and active-IXP data underpinning the analysis
Sustainable Peering Infrastructure Grant Program — ISOC Foundation funding for IXPs in LDCs/SIDS, the policy programme the findings speak to (Mar 2026)
Local Infrastructure, Lower Costs — companion ISOC analysis on peering and affordability (May 2026)