Internet Connectivity in 2025: More Traffic, Mobile Reach, and Uneven Reliability
Cloudflare Radar's 2025 observations show how internet growth, outages, protocol adoption, and mobile access intersect with network performance.
Data source: Cloudflare Radar Year in Review, based on Cloudflare network observations and complementary datasets
Internet growth does not remove local reliability gaps
Cloudflare Radar estimated that global Internet traffic grew 19% in 2025. Its review also counted 174 major outages worldwide, with almost half associated with government-directed regional or national shutdowns. These figures describe different dimensions: traffic volume can rise overall while access remains unreliable in particular places and periods. Cloudflare’s Year in Review provides the underlying observations and interactive views.
Mobile access is central in many markets
Cloudflare observed that mobile devices generated the majority of traffic in 117 countries or regions. That makes mobile network conditions and device capability important parts of web performance planning, especially for services used across markets with varied infrastructure.
HTTP/3 adoption depends on what is being measured
Cloudflare’s global request data showed a modest increase in the shares of HTTP/2 and HTTP/3 during 2025. Separately, HTTP Archive measured HTTP/3 on 29% of mobile HTML requests served by CDNs in its selected dataset, while origin-served mobile HTML had effectively no HTTP/3 in that sample. The two estimates are not interchangeable: one represents Cloudflare-observed request traffic, and the other focuses on selected popular websites and resources. HTTP Archive’s CDN chapter documents the sample and caveats.
What the trends mean for performance work
Measure the whole path, not just advertised bandwidth
Cloudflare’s Internet Quality data separates download and upload speed from idle and loaded latency. This is useful because a fast download result does not show how responsive a connection remains while it is busy. For calls, livestreaming, and interactive applications, upload capacity and latency under load can matter as much as peak download throughput.
Plan for mobile and regional variance
Use representative device and network profiles when testing pages. Compare server response time from multiple regions, check whether static assets are served near users, and validate important routes using real-user data. A single desktop test from one location will not expose the experience in every market.
Treat protocol support as an opportunity to measure
HTTP/3 uses QUIC and can reduce connection setup costs or limit the impact of packet loss in some conditions. Adoption alone does not prove a page is faster: content weight, server processing, cache behavior, and the user’s network still matter. Compare otherwise similar runs from the same region and device, and confirm protocol negotiation in request diagnostics.
Scope and limitations
Cloudflare Radar reflects traffic observed by Cloudflare’s network and complementary datasets, not a census of every internet connection. Its Year in Review covers January 1 through December 2, 2025; location coverage and sample sizes differ by metric. The outage, traffic, device, and protocol figures should be interpreted within those methods. They describe broad patterns and do not diagnose an individual provider or household connection.