DE-CIX launches world’s first fully-integrated AI Internet Exchange

With the implementation of Phase 1 of AI Internet Exchange (AI-IX), DE-CIX has announced that its global ecosystem of exchanges is now AI-ready. With Phase 1 completed, the company now has more than 50 AI-relevant networks – including both AI Inference as a Service and GPU as Service providers, as well as a number of cloud service providers – connected to their platform. This product evolution is available not only at India’s largest Interconnection Platform in Mumbai, but at all DE-CIX locations worldwide.

The Phase 2 of the rollout will be carried out soon and will make the DE-CIX AI Exchanges Ultra-Ethernet-ready, capable of supporting geographically distributed AI training, as model training begins to move out of centralised facilities. Rollout of both phases is being undertaken globally. As such, DE-CIX will become the first operator to offer an AI-IX that can support both training and inference.

Similar to the rollout of the DE-CIX AI-IX, AI operations in general are divided into two phases, AI inference and AI training. Both of these phases benefit from AI peering, with advantages including cost reductions, increased security, higher performance, and a reduction in complexity.

“With various predictions made for millions – even hundreds of millions – of AI agents in the coming years, the need for interconnection services to support their operations is growing massively”, explains Ivo Ivanov, CEO of DE-CIX.

AI Inference phase

AI inference – where the agent is applied to real-world situations to provide real-time insights, interaction, and support – is dependent on highly resilient, low-latency, and secure connectivity.

For Ivanov, these elements are, firstly, the millions of agents that are currently emerging, many of these being multimodal; secondly, the AI-powered devices/applications (cars, robots, processes, etc.); and thirdly, modern transmission technologies (such as fiber, 5G Advanced, and LEO satellite networks, etc.). These three elements must be interwoven in a reliable, secure, and extremely high-performance manner using direct interconnection, otherwise known as peering.

AI Training phase

In the second phase of the DE-CIX AI-IX rollout, the training of AI models is in focus. With the advent of Ultra Ethernet, a new protocol for the routing of data traffic in a quality relevant for AI, the design of infrastructure for the heavy computations of AI training is changing.

“Until now, huge, centralised data centres have been needed to quickly process AI computing loads on parallel clusters,” explains Dr. Thomas King, CTO of DE-CIX. “Ultra Ethernet is driving the trend towards disaggregated computing, enabling AI training to be carried out in a geographically distributed manner within a metropolitan area. This will revolutionise the infrastructure for AI training and offer companies new alternatives for designing resilient and more cost-effective private AI infrastructure.”