Хайтек→ original

Global backbone bandwidth grows 23% to 2259 Tbit/s

Global backbone network capacity grew 23% over the year to a record 2259 Tbit/s. Hi-Tech reported the figure, citing TeleGeography. Simple arithmetic puts the previous level at about 1837 Tbit/s, an increase of roughly 422 Tbit/s in twelve months. The number measures channel capacity, not the data actually transmitted. The original report does not name the causes of the growth. But with data centers and clouds expanding, the figure serves as an indirect indicator of the load on global infrastructure. • 2259 Tbit/s: a new record • +23% year over year, about +422 Tbit/s in absolute terms • At this pace, capacity doubles in roughly 3.3 years

AI-processed from Хайтек; edited by Hamidun News
Global backbone bandwidth grows 23% to 2259 Tbit/s
Source: Хайтек. Collage: Hamidun News.
◐ Listen to article

Global backbone network capacity grew 23% over the year to a record 2259 Tbit/s, or about 2.26 Pbit/s. Hi-Tech reports the figure, citing the analytics firm TeleGeography. Counting back, the level a year ago was about 1837 Tbit/s, which means the network added around 422 Tbit/s. That is more than the planet's entire backbone capacity would have been in the previous decade, judging by the orders of magnitude the industry typically works with.

What exactly was measured

The headline of the original publication refers to "internet traffic," but the body text talks about bandwidth. These are different things. Bandwidth shows how much data the channels can carry, that is, how much capacity has been built and connected. Traffic shows how much data has actually passed through them. Operators always keep headroom, so utilization is usually well below capacity, and the two metrics can move in different directions. The Hi-Tech report does not disclose which methodology was used or how the backbone was defined for the count, so compare this figure with other reports cautiously.

The growth rate and what it means in arithmetic

23% annual growth looks modest next to the revenue charts of AI companies, but compounding is relentless. At this pace, capacity doubles in roughly 3.3 years. Sustaining that trajectory means continuously laying new cables, lighting additional fiber pairs and upgrading equipment at the ends of the lines. These are capital expenditures that pay back over years, so decisions are made in advance, with an eye on expected demand.

The original material does not name the causes of the current growth. It is logical to assume that video services, clouds and the expansion of data centers, including those built for AI workloads, all contribute. But without TeleGeography data on the structure of demand, this remains a hypothesis, and attributing the growth specifically to neural networks would be an exaggeration. Training large models usually happens inside campuses and regional clusters with their own high-speed links, while traffic reaching the backbone is more often tied to delivering results, replicating data and serving user requests.

Key figures

  • Global backbone network capacity: 2259 Tbit/s (a record).
  • Annual growth: 23%.
  • Estimated level a year earlier: about 1837 Tbit/s.
  • Estimated absolute increase: roughly 422 Tbit/s.
  • Doubling time if the pace holds: about 3.3 years.
  • Data source: TeleGeography, as retold by Hi-Tech.

What teams building on this infrastructure should check

For those who host services or AI workloads across several regions, record capacity does not guarantee low latency or stable pricing. Backbone capacity is growing globally, but bottlenecks appear on specific routes and at specific traffic exchange points. So measure latency and packet loss between your own sites instead of relying on market-wide averages.

Second: when planning data replication and distributed inference, build in channel headroom based on your own volume growth, not on the global 23% figure. Third: when comparing internet reports, always check what was measured, capacity or actual traffic, and by what methodology. Without that, figures from different sources are easy to confuse, and conclusions end up resting on a faulty basis.

ZK
Hamidun News
AI news without noise. Daily editorial selection from 50+ sources. A product by Zhemal Khamidun, Head of AI at Alpina Digital.
✦ Unified AI Workspace

Access Frontier LLMs & Multimodal AI in One Workspace

Discover NomadNet AI (nomadnet.ai) — unified access to frontier language models, research synthesis, and multimodal workflows.

Try NomadNet AI →Launch NomadNet Agent
What do you think?
Loading comments…