Backblaze Publishes Q4 2025 Network Stats: Data Reveals the Rising Dominance of Neoclouds and AI Workflows

Why It Matters

The findings reveal that AI workloads are reshaping network design, forcing cloud providers to prioritize low‑latency, high‑bandwidth connections to regional compute hubs—a critical factor for competitive AI services.

Key Takeaways

• AI traffic peaks in US‑East data centers.
• High‑magnitude flows replace many‑to‑many internet traffic.
• Large migrations rose Aug‑Oct via private fiber.
• US‑West remains primary consumer traffic corridor.
• Backblaze B2 Overdrive links storage to neoclouds.

Pulse Analysis

The latest Backblaze Q4 2025 network statistics confirm that artificial‑intelligence workloads are redefining how data moves across the internet. Rather than diffuse, many‑to‑many traffic, AI models generate short, sustained bursts that gravitate toward a handful of high‑performance compute zones in the US‑East corridor—Northern Virginia, New York, and Atlanta. This concentration reflects the growing concept of data gravity, where massive training sets and inference results are stored, processed, and accessed within the same geographic footprint to minimise latency. For cloud providers, the shift signals a need to optimise peering and routing for these AI‑native flows.

Backblaze’s response, embodied in its B2 Overdrive service, creates a direct, high‑throughput path between its storage platform and the emerging “neocloud” ecosystems that host large‑scale models. The report highlights a surge in private‑fiber migrations from August through October, enabling multi‑petabit transfers that would be impossible over traditional ISP links. By measuring traffic magnitude per IP address, Backblaze demonstrates that high‑bandwidth, point‑to‑point connections are becoming the norm, prompting network architects to redesign capacity planning, adopt programmable switches, and prioritize low‑latency links for AI‑centric customers.

The broader market implication is clear: providers that cannot deliver ultra‑low latency and guaranteed bandwidth to neocloud hubs risk losing AI‑driven business to more specialised players. Enterprises are increasingly evaluating storage vendors not just on durability but on the ability to co‑locate or tightly interconnect with AI compute clusters. This trend accelerates investment in edge data centers, regional fiber dark‑loops, and software‑defined networking that can dynamically allocate resources to AI bursts. As AI workloads continue to dominate traffic, the competitive edge will belong to those who master the seamless integration of storage, compute, and network.