In a notable development for cloud infrastructure, Backblaze has reported a substantial rise in AI-driven data traffic directed towards neocloud operators, suggesting a transition from traditional internet traffic patterns to more concentrated, high-bandwidth flows typical of extensive model training and inference tasks.
Backblaze, recognized for its detailed assessments of storage hardware performance, has expanded its reporting to include a quarterly Network Stats report. The latest findings for Q4 2025 reveal a marked increase in traffic between Backblaze, neoclouds, and established hyperscalers from June through November.
Neocloud operators, specializing in GPU-as-a-service offerings, focus primarily on providing infrastructure for AI training, contrasting with traditional hyperscalers like AWS and Azure, which cater to a broader range of workloads. Backblaze meticulously categorizes network traffic, distinguishing between data from internet service providers (ISPs) and that from hyperscalers.
During Q4 2025, Backblaze observed that while content delivery network (CDN), hosting, and ISP traffic remained largely stable, both migration and neocloud traffic exhibited significant fluctuations. Migration traffic, which involves data entering the Backblaze network via fiber connections for large-scale transfers, surged starting in August before tapering off after October. In contrast, neocloud traffic saw a steady increase from July to November, peaking in October, aligning with the typical AI lifecycle of data ingestion and processing.
Despite the recent peaks in migration and neocloud traffic subsiding, Backblaze notes a persistent elevation in the overall baseline traffic levels. The most activity is concentrated in the US-West region, characterized by a higher number of unique ISP addresses, while the US-East region, particularly Northern Virginia, is a hub for neocloud activity due to its dense AI computing resources.
This trend is further illustrated by the unique IP addresses Backblaze interacts with; the US-West region features a broader array of addresses, whereas neocloud traffic is concentrated around fewer endpoints. This pattern aligns with AI pipelines that rely on stable connections between storage and compute resources.
Backblaze suggests that these developments may signal a transformation in network behavior driven by AI, as the concentration of AI traffic could lead to tighter integration of storage, compute, and network design. The company plans to continue monitoring these trends on a quarterly basis, aiming to identify whether neocloud activity becomes cyclical and to analyze additional factors such as the ratio of IPv4 to IPv6 traffic and cross-cloud connectivity.
This article was produced by NeonPulse.today using human and AI-assisted editorial processes, based on publicly available information. Content may be edited for clarity and style.
