DE-CIX: Ultra Ethernet & AI Driving Disaggregated Networks

DE-CIX, the world's largest internet exchange point operator by traffic volume, is preparing for a connectivity transformation in how AI workloads are processed across networks.

The firm is expecting AI to drive the adoption of disaggregated computing, where memory, compute and storage components operate separately within data centre infrastructure rather than as integrated units.

The transformation comes as the wider technology industry prepares for the roll-out of 5G Advanced in 2025, marking the next evolution in mobile connectivity standards.

DE-CIX CTO Dr Thomas King says: "Standards, open architectures and APIs will make tomorrow's network interoperable and intelligent. 6G will not only be up to 100 times faster than 5G, but it will also integrate AI to secure, manage and control networks. This level of intelligence will be needed in all connectivity technologies to manage future AI-driven data flows."

Ultra Ethernet could lead a shift away from InfiniBand

DE-CIX suggests that Ultra Ethernet technology is starting to enable a transition to replace InfiniBand, the traditional high-throughput communications architecture used in data centres.

"If computers and storage were connected via InfiniBand in an AI data centre, they had to be located only a few metres apart. Ultra Ethernet can bridge larger distances," says Thomas. “The standard is less complex, easier to use, is based on established and widely used Ethernet technology and allows large language models to be trained even in a metropolitan area.”

AI is inevitably driving transformation across the mobile connectivity industry, acting as a welcome solution for companies looking to boost their optics.

The technology can offer telcos a transformative opportunity to drive innovation, enhance customer experiences and unlock new revenue streams. However, the path to becoming an AI-native organisation is not without its challenges, given increasing regulatory landscapes and ethical concerns.

Analysis by McKinsey has indicated that telcos implementing the most advanced RAI practices could deploy use cases that collectively capture up to US$250bn in value worldwide by 2040 - a figure representing 44% of the full industry-wide value created by AI during that period.

DE-CIX CEO Ivo Ivanov says: "Smart value creation demands equally smart technologies. AI not only offers support for more intelligent network management, but for operational excellence in all areas of the telecommunications industry. 

“From network optimisation to energy efficiency, to fraud detection and customer service, smart solutions can enhance every single business process."

New approaches to AI in 2025

The integration of AI technologies is reshaping workforce requirements across the telecommunications sector.

"AI will create new roles and free up workers to undertake different challenges," says Ivo. "Human specialisations will focus more on areas that AI cannot solve and AI can also support companies to enable their staff to develop new skills and change their roles."

This transformation in workforce capabilities aligns with the increasing demands placed on network infrastructure by AI applications. Likewise, the traditional approach of training large language models (LLMs) in centralised data centres is evolving towards more distributed architectures, driven by space constraints in existing facilities.

"Whereas huge, centralised data centres were previously needed to be able to quickly process computing loads on parallel clusters, in the future it will be necessary to be able to train AI models in a more decentralised way - if only for the reason that space for data centers is limited everywhere," says Thomas.

This architectural shift requires new approaches to interconnection between distributed facilities.

"The solution is provided by AI Exchanges, which can interconnect disaggregated computing workloads and AI services via high-speed connections, and which are already prepared for the coming technological leaps in the AI market," says Ivo.

This shift enables operators to distribute AI workloads across multiple sites where power and space remain available, addressing both capacity and energy efficiency requirements.

"Connectivity is the essence of every AI application," says Thomas. "The history of DE-CIX over the past 30 years has shown how important a neutral and interconnected Internet is for the economy and prosperity in the world."

Explore the latest edition of Mobile Magazine and be part of the conversation at our global conference series, Tech & AI LIVE and Cloud & 5G LIVE.

Discover all our upcoming events and secure your tickets today.

Mobile Magazine is a BizClik brand