A Handful of Companies Control the Majority of the Internet's Core Infrastructure
Presenting our research on DNS and web hosting consolidation to the IETF DINRG community.
Presented at IETF DINRG, November 2025
The Internet feels vast and distributed, but the reality is far more concentrated than most people realize. In our recent presentation to the IETF’s Decentralized Internet Infrastructure Research Group (DINRG), we shared findings that reveal just how dependent the modern internet has become on a handful of organizations—and why that matters for everyone who relies on the web.
The Main Finding: Extreme Consolidation
Our research paper, “Measuring the Consolidation of DNS and Web Hosting Providers,” analyzed the Tranco top 1 million domains plus the top 10,000 sites from six global vantage points. What we found was striking:
Two organizations—Cloudflare and Amazon—each host over 30% of popular domains for both DNS and web hosting services. When you expand to the top five providers (adding Akamai, Fastly, and Google), these organizations control 60% of index pages in the top 10,000 websites and host the majority of external page resources.
Perhaps even more concerning is the fact that over 70% of top domains rely on a single organization for DNS resolution. This means that when one of these providers experiences an outage, the impact ripples across a massive portion of the internet.
A Global Phenomenon
One might assume that Internet infrastructure would vary significantly by region, with different providers dominating in different parts of the world. Our measurements from six globally distributed vantage points revealed otherwise: the consolidation trend is remarkably consistent worldwide. The same major players dominate regardless of where you’re accessing the internet from.
This wasn’t always the case. The internet was designed to be resilient through distribution and redundancy. Yet economic forces, economies of scale, and the complexity of modern infrastructure have driven websites and services toward a small number of hyperscale providers.
Our Measurement Pipeline
To understand consolidation, we built a comprehensive measurement pipeline that:
Identified authoritative nameservers for DNS hosting across all domains
Detected web content hosting providers for both index pages and external resources
Mapped infrastructure to organizations rather than just individual autonomous systems (AS)
Measured from multiple vantage points to capture global patterns
The distinction between measuring by organization versus by AS is critical. Amazon, for example, operates multiple autonomous systems (Amazon-02, AMAZON-AES, etc.), but consolidating these under the organizational umbrella reveals the true extent of concentration.
DNS and Web Hosting is Consolidated
Looking at our nameserver analysis, Amazon and Cloudflare dwarf all other providers. In the top 10,000 domains:
Amazon hosts over 2,500 domains for DNS (with most being reachable)
Cloudflare serves approximately 1,300 domains
The next tier (Akamai, Google) each serve only a few hundred domains
For web hosting, the picture is even more dramatic. Among the top 100 websites, Cloudflare alone hosts around 70% of domains. As you move down to the top 10,000 sites, the major five providers (Cloudflare, Amazon, Akamai, Fastly, Google) maintain consistent dominance at around 40-50% market share each.
Three Risks of Consolidation
1. Reduced Resilience to Outages
The consolidation of infrastructure means that single points of failure now affect enormous swaths of the Internet. We’ve already seen this play out:
The 2016 Dyn DDoS attack took down major websites across the internet
The 2021 Facebook BGP incident made Facebook and its properties unreachable worldwide
Routine AWS degradations affect thousands of dependent services
As one participant in at the DINRG discussion noted, the October 2025 AWS outage was caused by a race condition in automated DNS updates—a seemingly small technical issue that cascaded across services like DynamoDB, EC2, and network load balancers.
When infrastructure is this consolidated, these “unusual diseases” in one provider can become internet-wide epidemics.
2. Increased Potential for Censorship and Content Control
Beyond technical failures, consolidation creates concerning dynamics for online speech and content moderation. Large hosting providers exercise significant discretion over what content gets hosted and what doesn’t.
As Andrew Campling astutely pointed out in the Q&A, this consolidation has profound implications for free expression. We need only look at cases like Kiwi Farms—where decisions by a handful of infrastructure providers about whether to host particular content effectively determined whether that content could exist online at all—to understand the stakes.
When the majority of websites depend on the same five organizations for hosting and DNS, those organizations become de facto gatekeepers of online speech. Whether one agrees or disagrees with specific content moderation decisions, the concentration of this power in so few hands represents a fundamental shift from the internet’s original distributed architecture.
3. Privacy and Surveillance Concerns
With so much traffic flowing through a small number of providers, these organizations gain unprecedented visibility into internet usage patterns. The consolidation creates what amounts to surveillance chokepoints, where a handful of companies can observe—and potentially be compelled to share—information about vast portions of internet activity.
Extending the Research
The discussion following our presentation highlighted important directions for future work.
Availability Zones and Backend Dependencies
Brian Trammell raised a crucial limitation of our current study: we measured front-end behavior and didn’t provide visibility into back-end dependencies within cloud providers. A website might be “hosted by Amazon,” but which availability zones? What happens when there’s an outage in US-East-1 specifically?
Future research should incorporate IP geolocation to analyze dependencies within cloud providers’ infrastructure. Are sites truly distributed across availability zones, or do the consolidation patterns go even deeper than our organization-level analysis suggests?
Building on ICANN’s Work
Christian Huitema pointed us to ongoing consolidation studies at ICANN, which are examining similar questions about DNS infrastructure. Rather than duplicate effort, we’re eager to collaborate. Our measurement pipelines and data are publicly available, and we’d like to work with ICANN and others to create an ongoing “consolidation monitor”—a dashboard that tracks these trends over time and makes the data accessible to researchers, policymakers, and the public.
New Challenges: ECH and Encrypted DNS
As the internet evolves, new protocols like Encrypted Client Hello (ECH) will obscure some of the signals we used in our measurements. Future studies will need to adapt methodologies to maintain visibility into consolidation trends even as more infrastructure becomes encrypted and opaque.
What Can Be Done?
The concentration of internet infrastructure didn’t happen by accident, and it won’t be reversed overnight. But there are concrete steps the community can take:
Establish ongoing monitoring of consolidation trends with public dashboards and regular reporting
Develop resilience techniques specifically designed for data center and cloud-wide outages, not just individual server failures
Create mechanisms for multi-homing that make it practical for organizations to distribute across providers
Examine policy implications of infrastructure consolidation for competition, speech, and internet governance
Build guardrails for automation as AI and automated systems increasingly manage critical infrastructure
An Invitation for Collaboration
All of our code, data, and measurement metrics are publicly available. We’re committed to transparency in this research because we believe consolidation is a question that affects everyone who uses the internet. It’s not just an academic curiosity—it shapes the resilience, openness, and character of the global network we all depend on.
If you’re working on related questions, whether at ICANN, within a major provider, at a research institution, or as an independent researcher, we’d welcome the opportunity to collaborate. The internet’s future depends on our collective ability to understand and address these structural shifts.
Thank You to DINRG for Re-surfacing this Work
I want to extend my sincere gratitude to the hosts and participants of DINRG for the opportunity to present this work and for the thoughtful questions and suggestions that emerged from our discussion.
The IETF community’s commitment to understanding and improving internet infrastructure is exactly the kind of collaborative effort needed to tackle these challenges.
Special thanks to Brian Trammell for pushing us to think about availability zones and backend dependencies, to Christian Huitema for connecting us with ICANN’s parallel research efforts, and to Andrew Campling for highlighting the critical implications for free expression and online speech.
Read the full paper: Measuring the Consolidation of DNS and Web Hosting Providers
Authors: Synthia Wang, Kyle MacMillan, Brennan Schaffner, Nick Feamster, Marshini Chetty
Access our data and code: (forthcoming, stay tuned)
Questions, comments, or interested in collaborating? Please reach out! Subscribe to this newsletter, or reach out to me directly.