Editor's Note: This article is based on comprehensive research of publicly available sources including official press releases, regulatory filings, consumer complaints, technical forum discussions, academic studies, and industry publications. I may have failed in some areas to grasp the issue entirely. The reader is advised that not everything might be correct and you should follow the sources and conduct your own research to get an adequate understanding of the subject at hand.

Before diving into the article: My parents' home has two internet connections - a Telefonica DSL line (50 Mbit/s down, 10 Mbit/s up) and a Vodafone cable connection (1 Gbit/s down, 40 Mbit/s up). Despite the Vodafone line offering 20× faster download speeds, it's been so unreliable during peak evening hours that we've had to configure the Telefonica DSL as the primary WAN on the UniFi router for the past month. The gigabit connection simply isn't usable for normal internet activities when it matters most. When I read the Vodafone announcement, I had to share my take on it.

UPDATE 08.11.2025: Since the original article I've gotten a lot of feedback where the article might fall short. I still favour IXPs over NaaS (Network as a Service) providers because of the IXP's "public" nature. However I've since then gotten to better understand why the move to inter.link might not be that bad after all. I tried to work that perspective in at the end with a section "The case for the inter.link deal"

UPDATE 09.11.2025
I edited the article a little bit to get some details fixed. Most notably: 1.) I was mentioning DT as a good and a bad example, which did not really make sense. 2.): We need a common understanding. This is why there will be a series called "How the Internet works..." that tries to explain concepts and terminology. 3.) We need data: We are working on a tool to actually collect traceroutes from the community. We will work on a mechanism (a secret provided to a requesting IP, to make sure the submitter actually has access to the IP) that will ensure data is credible.
4.) inter.link is actually quite well connected. Yes, they are a NaaS, yes they potentially charge their customers for data, but maybe this is a good thing for Vodafone customers after all. 5.) Vodafone maintains a private peering with YouTube (Heise got an update from Vodafone regarding this)
In the end I think the only real truth lays in the data. If Vodafone delivers, good for them.

There's a reason your internet feels like magic. When you download the latest Ubuntu ISO in Berlin, that data doesn't travel some convoluted path through half of Europe to reach you. It normally flows through something called an "internet exchange point"—a giant room full of routers where hundreds of networks connect directly, swapping traffic efficiently and, crucially, for free.

Vodafone Germany is about to exit that public IXP based system.

By the end of 2025, Vodafone intends to have completely withdrawn from every public internet exchange in Germany, including DE-CIX Frankfurt, the largest internet exchange on the planet. Instead, all traffic will flow through a single company called Inter.link (except for Youtube apparently -> see the Update on the Heise article), which possibly will charge content providers based on how much data they send to Vodafone customers.

Vodafone insists this move will deliver "lower latencies, more resilience, and cost savings." From my own experience Vodafone does have bandwidth problems currently, causing me to switch the primary WAN at my parents place to Telefonica, despite booked downlink capacity being 20x less on the Telefonica line. So: The question is wether this pans out. Will things get noticeably better or worse?

The open internet was supposed to work differently

To understand why this matters, you need to understand how the internet actually works. And it's not how most people think.

When you pay for internet service, you're not buying access to some centralized "internet"—you're buying access to your ISP's network. That network then connects to thousands of other networks through a patchwork of agreements and connections. The magic happens at places like DE-CIX Frankfurt, where over 1,000 networks plug into the same switching fabric and exchange traffic directly.

This system is called "settlement-free peering," and it's one of the internet's foundational principles. No money changes hands. Networks exchange traffic roughly equally, everyone saves money on long-haul transit costs, and users get faster connections because data takes the shortest possible path.

When an ISP customer wants to download a large file from a cloud server somewhere on the WWW, chances are that traffic flows directly from the DC operators network to the ISP network at an internet exchange point—maybe four or five router hops, minimal latency, no intermediaries. It's elegant. It's efficient. And it's exactly what Vodafone is abandoning.

Enter the middleman

Instead of connecting directly to content providers at neutral exchange points, Vodafone is outsourcing all its peering to Inter.link, a Berlin-based company that operates what it calls a "peering-as-a-service" platform. Inter.link operates more than 40 points of presence across 15 countries (info taken from their peering policy, while their network page claims a lot more) and claims connectivity to 300+ data centers in Europe.

The pitch is automation. Instead of managing thousands of individual peering relationships, Vodafone gets "one-click provisioning" and "single sign-on access." For Vodafone's network operations team, this is genuinely simpler. But here's the thing: operational efficiency for a telecom company and good service for customers are not the same thing.

Vodafone's official press release from November 2025 emphasized that the partnership with Inter.link would "reduce time, resources, and peering costs" while delivering "lower latencies" and "more resilience." What the company didn't mention is that "peering costs" in this context doesn't mean Vodafone's costs—it means the costs have shifted to content providers, who now have to possibly pay Inter.link if they want decent connectivity to Vodafone's customers.

And if a content provider decides not to pay? That's when things get ugly.

Settlement-free peering that isn't actually free

Inter.link's public peering policy looks remarkably normal on paper. Updated September 1, 2024, it states the company is "open for settlement free peerings" and offers tiered arrangements similar to traditional internet exchanges: route server peering at IXPs with no traffic minimums, direct IXP peering above 1 Gbps, and Private Network Interconnects (PNIs) for networks exceeding 10 Gbps. The policy describes cross-connect cost-sharing but mentions no recurring charges, no per-Mbps fees, and no settlement payments between peering partners.

This mirrors how 99.5% of global peering relationships operate according to OECD data—networks pay for infrastructure access while exchanging traffic settlement-free with each other. The policy's technical requirements are actually quite sophisticated: mandatory IPv6 support, strict RPKI route validation, comprehensive BGP filtering, and modern 400GE backbone infrastructure. Nothing in the document reveals anything controversial.

Until you notice one critical sentence: "Inter.link does not peer with customers."

This single line appears to be the key to understanding Inter.link's entire business model. Here's what we can confirm versus what we must infer:

What we know for certain:

Inter.link offers settlement-free peering at public exchanges (stated in their policy)
They explicitly refuse to peer with their own "customers" (stated in their policy)
Vodafone is using Inter.link's commercial "FlexPeer" platform (confirmed in press releases)
Inter.link charges for various services including IP Transit and DDoS Protection (published on their website)
Content providers previously had direct settlement-free peering with Vodafone at DE-CIX (standard practice, confirmed by DE-CIX)

What the evidence strongly suggests: When Vodafone exits DE-CIX and routes all traffic through Inter.link, content providers face a choice. They can establish settlement-free peering with Inter.link at public exchanges—but this doesn't guarantee quality delivery to Vodafone's network, since Vodafone is no longer there. To ensure reliable access to Vodafone's 30+ million customers, they likely need to become Inter.link "customers" using the commercial FlexPeer platform.

A simplified example to illustrate the transformation:

Before (at DE-CIX Frankfurt):

Netflix connects to DE-CIX switch: €500/month port fee
Vodafone connects to DE-CIX switch: €500/month port fee
Traffic flows directly between them: €0 (settlement-free)
Total cost to deliver Netflix to Vodafone customers: €500/month

After (via Inter.link):

Netflix can peer with Inter.link at DE-CIX: €500/month port fee + €0 peering
But Vodafone isn't at DE-CIX anymore
For guaranteed delivery to Vodafone, Netflix likely needs FlexPeer: Unknown commercial rates based on traffic volume
Total cost: €500/month + Inter.link's commercial fees

Again, I might be wrong here, but this is to the best of my understanding.

The genius is in the semantic distinction. Inter.link can truthfully claim they offer "settlement-free peering" while simultaneously converting those potential peers into paying customers through ISP outsourcing arrangements. The company's peering policy contains remarkably little information about these commercial arrangements—no FlexPeer pricing, no volume-based fee structure, no explanation of the revenue model that makes their business viable.

This isn't deception—it's careful categorization. Settlement-free peering exists for those who want to exchange traffic with Inter.link itself. But reaching Vodafone customers? That's a commercial service. The traditionally binary relationship (content provider ↔ ISP) becomes triangular (content provider → Inter.link → ISP), with Inter.link extracting fees from what was previously cooperative infrastructure.

Deutsche Telekom - a blueprint for what might be coming

Deutsche Telekom, another big German ISP, has been at the center of widespread customer complaints about systematic service degradation for years. This has culminated in the project "netzbremse.de" (eng. Network Break).

Consumer organizations have documented hundreds of complaints showing that Telekom's business practices restrict customers' freedom to use the apps and sites of their choice: certain websites and services load slowly or not at all, professional cloud services are only partially functional, e-learning videos and games stutter, and video calls repeatedly break off. The coalition filing the complaint states that Telekom is the only provider in Germany that deliberately exploits these bottlenecks at the entrance to its network.

Software and downloads

Multiple user reports document severe throttling. GitHub downloads crawl at 50-100KB/sec on unloaded, perfectly functioning 50MBit VDSL connections—speeds that only occur from Telekom's network, while other networks deliver the same content at several MB/sec. Some Telekom customers report being unable to complete even 19MB downloads due to packet loss, describing the situation as "completely incomprehensible for 2020/2021." Python package downloads from pythonhosted.org barely reach 45 KB/s on 100 Mbps FTTH connections.

Gaming

The impact on gaming is particularly severe. Users report packet loss and "rubberbanding" occurring every 15-20 seconds, with latency jumping from 20-30 milliseconds to 200-3,300 milliseconds, making games like CS2, Valorant, Apex Legends, and Battlefield unplayable. Telekom fiber customers report 2-30% packet loss across all online games regardless of server location.

Time patterns

Services work fine in the morning when traffic is low, degrade severely during 7-11 PM peak hours, then recover late at night. Network experts have identified the root cause: Deutsche Telekom deliberately refuses to upgrade interconnection capacity when links reach congestion, creating artificial bottlenecks to pressure content providers into paid agreements.

The warning signs have been building for years

Vodafone's peering problems aren't new—they've been escalating since at least 2019. Forum posts from that period document "massive peering problems" affecting connections to hosting providers, with technical users noting routing inefficiencies where Berlin traffic was inexplicably routed through Düsseldorf. By 2020-2024, complaints had become more profound: packet loss, extreme ping spikes, and the telling symptom that VPN usage instantly restored normal speeds, many of which I can confirm from own experience. A cousin of mine, who is into roblox, complained a lot about buffer bloat and latency.

Berlin residents experienced particular degradation when Vodafone withdrew from BCIX (Berlin Internet Exchange), with forum posts noting that BCIX peering "was eliminated months ago and notably worsened network quality for Berlin residents." Major content providers including Amazon, Apple, Facebook, Google, Microsoft, Netflix, and Twitch all peered at BCIX for low-latency Berlin connectivity. Those direct connections are gone, and will now possibly replaced by longer routes through Inter.link.

Does common logic support Vodafone's claims?

Let's talk about that "lower latencies" promise, because it seems to defy basic networking principles.

Academic research measuring performance across 900+ networks found that 91% experience at least 5% latency improvement via direct peering versus transit, with median improvements of 12-15 milliseconds. The reason is simple physics: shorter paths are faster than longer paths.

This is networking 101: every additional network you traverse adds latency from routing decisions, queuing delays, and physical transmission time. Direct peering at an exchange point minimizes this. Routing through an intermediary by definition increases it.

I couldn't find any supporting measurements that would prove Vodafone's point of better latency. No traceroute comparisons. No latency benchmarks. No throughput tests. No independent technical validation. For a major telecommunications company claiming performance improvements while undertaking a massive infrastructure change, the absence of any quantitative evidence is a lack of transparency.

The satellite escape hatch

While Vodafone dismantles its open peering infrastructure, an alternative has emerged that sidesteps these politics entirely: satellite internet.

Starlink operates its own global backbone network with direct agreements with major content providers. When you access YouTube via Starlink, traffic flows through SpaceX's network to satellites and down to your dish—a fundamentally different architecture from terrestrial ISP peering arrangements.

The Starlink residential service won't match premium fiber speeds, but from our own testing we consistently reached above 1oo Mbps median downloads in Germany with 30-40ms latency. More importantly: no selective throttling. Netflix and GitHub load at the same speeds because at least from my own experience with Starlink, it doesn't play favorites.

That competitive pressure matters. Terrestrial ISPs aren't quite the monopolies they used to be. And given upcoming competition in the LEO satellite game, switching will be as easy as changing the satellite dish, no road side cable changes required.

To be fair: Starlink seems to also be customer or peer with Inter.link since they are listed on the inter.link website as of today. However, it is for sure not the only peering of Starlink.

So... Why?

Here's what Vodafone isn't saying in its press releases: I believe this migration is fundamentally about changing the internet's economic model.

Vodafone's new model moves access to Vodafone's clients behind a single gatekeeper: Inter.link. I assume that having the best available access for Content Providers means having to join Inter.link as a customer.

Professor Barbara van Schewick of Stanford Law School, whose work informed FCC net neutrality orders and EU guidelines, characterized similar practices by Deutsche Telekom as "a frontal attack on the open internet."

In August 2024 comments to European regulators, van Schewick explained that ISPs exploit their "termination monopoly" by charging "monopoly termination fees" to reach subscribers, and concluded that "paid interconnection fees violate the Open Internet Regulation's ban on discrimination and paid fast lanes."

In December 2024, Switzerland's telecommunications regulator delivered an 11-year ruling against Swisscom for similar practices. The Swiss ComCom decision found that ISPs have a "technical monopoly of access to their end customers" and that charging content providers for traffic delivery is "not permissible" because costs are already covered by customer subscriptions. The ruling explicitly rejected the "double-dipping" model where both customers and content providers pay for internet access—precisely what Vodafone implements through Inter.link.

The Meta-Deutsche Telekom showdown sets the stage

The conflict reached a boiling point in September 2024 when Meta refused to pay Deutsche Telekom €20 million in court-ordered peering fees. Rather than comply, Meta rerouted all its traffic through third-party transit providers, completely severing its direct peering relationship with Deutsche Telekom. The German court had ruled that Deutsche Telekom could charge for peering, but Meta called this "putting the open Internet at risk and undermining net neutrality principles."

Regulators are paying attention (finally)

The April 2025 complaint against Deutsche Telekom came from a coalition including Germany's federal consumer organization (Verbraucherzentrale Bundesverband), Epicenter.works, Gesellschaft für Freiheitsrechte, and Stanford professor Barbara van Schewick. The complaint documents technical evidence of artificial bottlenecks and charges that Deutsche Telekom creates "paid fast lanes" prohibited under EU law.

BEREC (Body of European Regulators for Electronic Communications) released a report in December 2024 after a two-year investigation, identifying practices where ISPs "exploit bottlenecks at the entrance to its network to demand payments from online services" and classifying such practices as "potential violations of Europe's net neutrality law." The report listed "several examples, most of which involved Deutsche Telekom" [as cited in the complaint filing and CISPE's statement].

The Cloud Infrastructure Service Providers in Europe (CISPE) backed these findings in April 2025, noting that their members have faced "artificial network service quality degradation" in Germany since at least 2015, used as leverage for "additional unjustified payments." Technical communities have been even blunter, with forum discussions describing Deutsche Telekom's approach as a "Peering Extortion Policy" and calling it a "blatant violation of EU directives."

If the investigation finds Deutsche Telekom's model violates EU net neutrality regulations, Vodafone's identical approach faces the same legal jeopardy. Germany's Federal Network Agency has already shown willingness to act—in February 2022, it prohibited Vodafone's "Vodafone Pass" zero-rating service following European Court of Justice rulings.

The bigger picture

Vodafone's exit from public peering isn't an isolated technical decision—it's part of a broader pattern of large telecoms trying to reshape internet economics in their favor. Thomas King, CTO of DE-CIX Frankfurt, warned without naming companies directly: "We are currently observing a trend in which large market players are increasingly using their dominant position to monetize not only their Internet access business but also network interconnection."

The trend is particularly concentrated in Germany, where both major incumbent ISPs now operate restrictive paid peering models while competitors maintain open policies. IT-Administrator, a leading German IT publication, warned that Vodafone's shift "could reduce transparency, while smaller providers and content providers may face higher entry barriers" and "could impair the diversity and openness of European Internet infrastructure."

If this model proves profitable for Vodafone and Deutsche Telekom without regulatory consequences, expect other European telecoms to follow. The open internet—where networks cooperate through settlement-free peering to deliver the best possible service—gets replaced by a gatekeeper model where ISPs extract fees from both customers and content providers while under circumstances delivering worse performance.

The case for the inter.link deal

This post was posted on news.ycombinator.com and got a lot of comments there. As I mentioned in the disclaimer, there may be parts that I do not grasp entirely. I've had a few thoughts and learnings that may play in favour of the Vodafone/inter.link deal.

I get that routing and peering can be hard and cost intensive from a maintenance standpoint. Outsourcing that part for sure can be cheaper. And given that inter.link might really know what they are doing, one could say that improves quality, if Vodafone was not up to the task beforehand.

Inter.link's business model bears interesting similarities to Cloudflare's approach, except while Cloudflare mainly protects and accelerates individual websites and services, inter.link's focus is at the interconnect network infrastructure layer, sitting between ISPs and cloud providers rather than between service providers and end users. Still there are also similarities: Cloudflare even provides a specific routing service with their Argo product. This website uses Cloudflare, so I can understand the motives.

Given the name "FlexPeer" what would be the ideal hope I have, that it actually is going against bottlenecks and will actually improve the situation.

From inter.link's own material on their websites I take they are actually quite well connected to other networks. They also state to be a VPP for Google.

In the end I think time will tell how things will pan out. In the meantime I'll try to get an interview with someone who can clear up some questions that arose on my side.