Our first, conservative view asserts that AI’s impact on networks is largely limited to the data center, to connect clusters of GPU servers and the data they use when analyzing large language models. This is only “horizontal” traffic; a TikTok challenge would generate much more traffic in a large area. WAN costs won’t increase for you as a business, and if you’re an operator you won’t bring much new to the table, so you won’t see much increase in service revenue. If you don’t host AI on-premises, you can virtually ignore its impact on your network.
Compare this with the radical vision of the metaverse, our third vision. Metaverses and AR/VR are transforming AI missions and network services from transaction processing to event processing, because the real world is a set of events that push you. They also allow you to visualize how process control models (digital twins) relate to the real world, which is essential if the processes you are modeling involve human workers who rely on their visual sense.
Could it be that the reason Meta is willing to invest in AI is that the most credible and impactful application of AI for networks is the concept of the metaverse? Regardless, this AI model, by directly driving user experiences and activities, requires significant edge connectivity, so it can be expected to have a major impact on network requirements. In fact, just dipping your toes into a metaverse might require a major network upgrade up front.
Networks carry traffic. Traffic is made up of messages. More messages, more traffic, more infrastructure, more service revenue… you get the picture. Door number one, into the AI giant’s future, doesn’t lead to anything much in terms of messages. Door number three, the metaverse and AR/VR, leads to a revolution in messages, traffic and networks. I bet most companies would doubt the value of the first door and fear the initial financial impact of the third, which leaves us with the middle door, for the AI agent.
AI agents argue that AI is really just a software component, a path to a new type of automation. This does not rule out a role for giant AI hosts and utilities, nor a world of AR/VR and metaverses. It’s not really about where AI ends up, but rather how it gets there. Agents move slowly, project after project, in the same way that IT has always entered the workplace and our lives.
Meta embodies the most exciting, disruptive, investment-dependent, and ultimately profitable vision of AI. Google and Microsoft offer the most accessible and simplest version. What lies in between, the agent story, is not glamorous and surely is not instantly gratifying to a buyer or seller, but it connects the two, or it might. The success of the bridge lies in its foundation: the chips. And it’s not the super-GPUs that it depends on, but the chips in your phone.
