As AI reshapes how data moves through networks and data centers, GPU-centric neocloud providers are emerging as something more than alternatives to hyperscale cloud platforms.
That’s the key takeaway from cloud storage provider Backblaze’s Q4 2025 Network Stats report. Published in late January, the analysis shows sustained growth in traffic between Backblaze’s storage layer and neocloud compute destinations. The pattern mirrors broader market and industry assessments from Synergy Research Group and McKinsey & Co.
Following its first network transparency report last summer, Backblaze said its expanded fourth-quarter dataset indicates neocloud demand remained high as AI workloads continued to scale. From July through November, the company observed a sharp rise in traffic between Backblaze and neocloud providers. That activity peaked in October before settling into a higher baseline by year’s end.
“Few forces are reshaping network behavior faster than AI,” said Brent Nowak, technical lead network engineer at Backblaze.
According to the report, neocloud traffic stood apart from the more predictable, steady patterns seen across content delivery networks and traditional internet service providers. Instead, Backblaze identified concentrated, high-bandwidth flows consistent with AI workloads.
“This reflects a familiar AI lifecycle,” Nowak wrote on Jan. 29. “Large datasets — including images, videos, and metadata — are ingested and consolidated, then exported for training and experimentation.”
Geographically, much of the fourth-quarter neocloud activity clustered in the US East regions with dense AI compute availability. That concentration reflects the importance of proximity and latency in high-throughput AI data transfers, Nowak said.
“It’s important to keep latency … lower to achieve consistent high bandwidth rates for AI data transfers,” he wrote. “For now, that gravity is pulling activity towards the East Coast.”
Backblaze cautioned that it’s too early to determine whether neoclouds are permanently reshaping traffic concentration patterns.
“Since this is our first quarter with a full dataset, it’s a bit early to draw long-term conclusions,” Nowak said.
Even so, industry analysts are pointing to similar dynamics. In recent research, McKinsey & Co. described neoclouds — providers that specialize in GPU-as-a-service — as beneficiaries of surging AI demand due to flexible contracts, faster provisioning and specialized infrastructure. McKinsey noted that GPU pricing from neocloud providers can be significantly lower than hyperscale alternatives, making them especially attractive to AI startups and emerging workloads
“Neoclouds present lower barriers to entry than traditional cloud providers,” McKinsey analysts wrote in November. “Standing up a compute cluster does not require building a full tech stack, as a hyperscale platform does — so new entrants can move quickly to capture unmet demand.”
Backblaze’s network observations align with cloud market data from Synergy Research Group. In November, Synergy reported that while Amazon, Microsoft and Google collectively held about 63% of global cloud infrastructure spending in the third quarter of 2025, neoclouds ranked among the fastest-growing segments outside that group. Providers such as CoreWeave, Crusoe, Nebius and Lambda continue to expand their footprint, including through the channel, even as they remain smaller in aggregate than the hyperscale providers.
The implication, analysts said, is not that hyperscalers are losing dominance. Rather, the Big Three aren’t capturing all incremental growth, particularly in workloads that demand specialized hardware and rapid provisioning. Backblaze’s visibility into sustained neocloud traffic suggests those environments are landing active AI workloads.
The company’s latest network data provides early evidence of a significant structural shift in web traffic. AI workloads are creating network behavior that looks fundamentally different from legacy patterns, and a growing share of that activity is flowing through neocloud infrastructure.
For partners, the bottom line is that neoclouds aren’t niche anymore because they sit outside the hyperscaler orbit. Instead, they represent distinct compute environments where AI data movement already is occurring at scale — and where ecosystem support will become more critical as AI matures.
“As we gather more quarters of data, we’ll be watching closely to see how cyclical neocloud activity becomes, how regional concentrations shift, and how the growing ecosystem of AI-focused ISVs continues to change the shape of the network,” Nowak said.
