Enterprises must plan for the future, and the trajectory of AI matters far beyond AI itself. It influences the technology they purchase, the networks they build, and ultimately, the profits they generate. Since the future is never certain, it’s wise to explore a spectrum of possibilities—from the most conservative to the most radical. And yes, it’s also more interesting that way.

Figure 1. Networks, AI, and the Metaverse: A New Era.

The most conservative view is that the hyperscalers deploying enough GPU servers to rival power grids (and inflate residential electricity bills) are correct: hosted AI services—those you can chat and interact with—will remain the dominant model. Figure 1 shows Networks, AI, and the Metaverse: A New Era.

A more middle-ground perspective envisions AI agents—essentially advanced software components—injecting intelligence into applications, enhancing operations, and enabling real-time processes and workflows. This approach extends the automation paradigm that has existed for over half a century, modernizing it and making it far more capable.

The radical view? Metaverses and virtual or augmented reality. Imagine where we work, live, and play all existing inside a giant computer game, with AI acting as the dungeon master. Critical real-world constraints—like the fact that a wall is solid—remain, but non-critical elements can be manipulated. That wall might display an ad for a store on the left, or indicate the safest exit from a burning building if you move to the right. Everything you need to know is literally before your eyes.

Investors have probably experienced whiplash watching AI stocks fluctuate wildly over the past month, swinging between perceived risk and opportunity. While stock prices offer little guidance for the long-term, enterprise network professionals, operators, and ISPs can at least consider how each AI scenario impacts networks.

In the conservative view, AI’s network impact is mostly confined to the data center—connecting clusters of GPU servers as they crunch large language models. Traffic remains largely “horizontal”; a single viral TikTok generates far more wide-area traffic than AI. WAN costs for enterprises won’t rise significantly, and carriers won’t see much new traffic or service revenue. If you don’t host AI on-premises, its impact on your network is minimal.

By contrast, the radical metaverse vision transforms AI’s role—and network requirements—from simple transaction processing to continuous event processing. The real world is a stream of events, and visualizing processes with AR/VR digital twins becomes critical, especially when human workers rely on their senses. Perhaps this explains why companies like Meta invest heavily in AI: the metaverse may be the most credible, network-intensive application of AI. Edge connectivity becomes essential, meaning even minimal engagement with a metaverse could require substantial network upgrades.

Networks carry traffic. Traffic carries messages. More messages mean more traffic, more infrastructure, and more service revenue. In the conservative AI world, there’s little impact on messaging. In the metaverse world, AI drives a network revolution. Most enterprises may dismiss the first scenario and fear the costs of the third, leaving the middle path—AI agents—as the most practical route.

AI agents show that AI is, at its core, a software component—a new pathway for automation. This approach doesn’t rule out giant AI hosts and public services, nor does it exclude a future of AR/VR and metaverses. It’s not about where AI ends up, but how it gets there. Agents move incrementally, one project at a time, just as computing has gradually integrated into our workplaces and daily lives.

Meta represents the most ambitious, disruptive, investment-heavy, and ultimately profitable vision of AI. Google and Microsoft offer the most accessible and achievable version. The middle ground—the agent story—is less glamorous and rarely delivers instant gratification, but it serves as the bridge between these extremes. The success of this bridge depends on its foundation: the chips. Not the massive GPUs in data centers, but the chips already in our phones.

Metaverses and AR/VR require collecting and correlating enormous amounts of data, often with very low latency, close to the real-world actions being visualized. But how much AI can realistically fit into a pair of glasses or a sensor? Almost everyone already carries a smartphone. A future where AI is tightly woven into our lives and work must leverage the device we already have. Giving smartphones a central role in AI not only enables real-world integration, it promotes edge traffic and edge computing—all without breaking the bank. Many of us already have phones with AI chips, and if we add AR/VR glasses, we’ll naturally want them to link to our phones.

Even with AI agents, metaversing still raises network concerns. But the beauty of agents is that they let us ease into that future. You don’t need to equip every worker with AR/VR glasses right away. Business applications of a metaverse don’t require immediate personal visualization. You can start by placing an AI agent on each phone, letting it cooperate with your IT applications, data, and workers. This is the bridge that lets AI demonstrate real value.

Want AI networks that matter? Want to realize the true potential of AI? Focus on AI agents that empower smartphones. When they touch all aspects of our lives, they will transform both work and the network that supports it.

Source: NETWORK WORLD

Cite this article:

Priyadharshini S (2025), Exploring Networks, Artificial Intelligence, and the Metaverse, pp.252