Carrier Ethernet Upgrades Have Accelerated for AI
Telecom operators are scaling networks to 800G standards to support the massive data flow required by AI workloads.
Updated on Sept. 29, 2026 in Data Centers
Live Poll
Do you trust that current telecommunications infrastructure is capable of handling the growth of AI?
Telecommunications providers have increased investments in Carrier Ethernet infrastructure to meet the high bandwidth and low-latency demands of AI models. By upgrading backbone networks to 400G and 800G standards, operators aim to ensure stable data streams that prevent costly GPU idling.
Why it matters
AI training workloads generate massive, continuous data streams that require perfect synchronization between edge sites and data centers. Reliable connectivity is a strategic asset for enterprises, as even minor packet loss can lead to significant reductions in processing efficiency.
A 0.1% packet loss rate results in a 13% drop in GPU utilization, highlighting the need for precise traffic management. Operators are currently leveraging VLANs and SDN automation to handle these complex AI traffic patterns.
The players
Vodafone Idea
This telecommunications company recently achieved a 1.6Tbps transmission milestone on a mesh Data Center Interconnect network.
Verizon
The major telecommunications provider recently finalized a $1 billion agreement with Google to bolster data center connectivity.
Lumen Technologies
This infrastructure firm is a key player in the sector, reporting $9 billion in new Private Connectivity Fabric deals.
The details
Infrastructure upgrades have become a priority as 57% of enterprises continue to train AI models on-premises, requiring seamless integration between local hardware and cloud systems. Major companies are backing these shifts, with Verizon signing a $1 billion deal with Google and Lumen Technologies reporting $9 billion in Private Connectivity Fabric deals.
Timeline
Operators plan to direct 41-80% of network spending toward AI infrastructure over the next three years.
The Tech Race
The transition to 400GE and 800GE standards follows a pattern set by the 400GE and 800GE backbone network standards to accommodate modern AI infrastructure needs. This shift marks a fundamental departure from legacy network architectures that were not designed for the extreme data throughput of AI training.
Users may experience faster model performance and reduced downtime as businesses successfully integrate high-speed networking into their AI workflows. These network enhancements reduce the risk of stalled AI projects that currently suffer from inefficient, bottlenecked data transmission.
The takeaway
Reliable high-speed networking has moved from a technical luxury to a critical operational requirement for modern AI development. Companies that fail to modernize their data transfer capabilities face significant performance drops and diminished returns on their hardware investments.
Further reading
Learn more about the infrastructure behind modern computing in our Data Centers section.
Live Poll
Do you trust that current telecommunications infrastructure is capable of handling the growth of AI?
