Backblaze has released its Q3 2025 Network Stats Report, revealing a transformative shift in global data movement driven by artificial intelligence. The report shows that AI-related traffic now accounts for nearly one-quarter of all network flows across Backblaze’s global infrastructure, marking a fundamental change in how data is generated, transferred, and consumed. Brent Nowak, Technical Lead Network Engineer at Backblaze, emphasized the significance of this shift: “The way data moves is fundamentally changing. As AI training and inferencing accelerate, data is no longer just growing—it’s flowing differently.” This evolution is reshaping the underlying architecture of digital infrastructure, prompting Backblaze to introduce a new metric: the “magnitude” of traffic. This measure captures the intensity of data flows per endpoint, highlighting that AI-driven traffic is an order of magnitude more concentrated and intense than traditional workloads. Historically, content delivery network (CDN) traffic dominated Backblaze’s network as its B2 Object Storage platform expanded to serve enterprise backup and data storage needs. However, in Q3 2025, traffic from “neoclouds”—cloud providers offering AI-specific compute, GPUs, and related services—has surged to nearly 25% of total ingress and egress. Unlike steady, distributed data flows from traditional applications, AI workloads generate short, high-volume bursts tied to dataset replication, model training, and inference pipelines. These bursts are concentrated across fewer IP addresses, reflecting the dense, high-throughput nature of AI compute clusters. Backblaze’s new magnitude metric provides a clearer picture of this shift, showing that AI traffic is not just increasing in volume but also in intensity. The data underscores a growing trend known as “AI data gravity,” where data increasingly moves in response to compute needs, performance demands, and cost efficiency rather than remaining centralized in a single cloud provider. This shift is enabling a more open, multi-cloud future. Organizations are no longer forced to lock data into a single hyperscaler to avoid egress fees. Instead, they can leverage best-of-breed services across providers, combining performance, cost, and accessibility without sacrificing control. The Q3 2025 report marks a new phase in Backblaze’s Network Stats series, which now complements its long-standing Drive Stats reports as a quarterly transparency initiative. The dataset includes detailed telemetry on bits, packets, flows, IP address volume, traffic categories, and protocol-level analysis, with metrics reported at average, 95th percentile, and maximum levels. To further support research and innovation, Backblaze will begin publishing raw network telemetry data on a quarterly basis, enabling independent analysis by developers, academics, and cloud partners. Gleb Budman, CEO of Backblaze, said, “Transparency around how data actually moves is critical for the next generation of builders. Our mission is to make customers unstoppable by solving their toughest data storage challenges. By publishing these network insights, we’re helping customers understand not only how Backblaze operates, but also how the global data ecosystem is changing.” The full Q3 2025 Network Stats Report, titled “The Magnitude of AI Workflows,” is available on the Backblaze blog.