The $15.7 Trillion Bottleneck: Why AI’s Future Depends on Rewiring the Internet

For the global telecom industry gathering at ITW, that figure represents both an enormous opportunity and a growing risk. While much of the conversation around AI focuses on models, compute, and applications, a more fundamental constraint is emerging. Connectivity is emerging as the primary limitation.

The internet was designed for human interaction - email, video, cloud applications - where occasional delays or variability are acceptable. AI systems operate on entirely different terms. As enterprises move toward autonomous, agent-driven models, machines will increasingly communicate directly with other machines, executing tasks at speeds and volumes far beyond human capability. These AI agents will move massive datasets, swarm together, coordinate across multiple systems, and make real-time decisions. In this environment, the network takes on a new role as the execution layer of the digital economy.

Despite exponential growth in AI-driven traffic and investment in data centers, the way connectivity is provisioned has changed very little. Establishing high-capacity, secure connections across multiple providers can still take weeks or months, often relying on manual processes, fragmented systems, and human intervention. These delays are fundamentally misaligned with AI-driven operations. When workloads operate in real time, waiting weeks to months for bandwidth effectively halts innovation. Speed in an AI-driven economy is measured in milliseconds, far beyond the pace of conventional provisioning cycles.

At the same time, the traditional “best-effort” model of the internet is no longer sufficient. AI workloads, particularly those involving large-scale training or real-time inference, require predictable, guaranteed performance. Even minor disruptions in latency or packet delivery can undermine outcomes, waste compute resources and introduce risk into critical operations.

This is driving a shift toward deterministic connectivity, where performance is assured rather than probabilistic. For telecom operators, this represents a fundamental evolution from providing access to delivering guaranteed outcomes. Meeting these demands requires more than incremental upgrades. It calls for a transformation in how networks are consumed and controlled. AI systems cannot rely on manual provisioning or static infrastructure. They need the ability to dynamically discover, provision, and manage connectivity in real time and across the ecosystem.

This is where automation and standardization become critical. By exposing network capabilities through standardized APIs, operators can enable a model where connectivity behaves more like cloud infrastructure - elastic, programmable, and available on demand. An AI system should be

able to acquire capacity for a specific task, use it, and release it automatically, without human intervention. This shift is central to the evolution of Network-as-a-Service (NaaS) and is foundational to scaling AI.

However, no single provider can deliver this alone. AI workloads inherently span multiple networks, clouds, and geographies. To support them, the industry must move toward a federated model, where operators interconnect through common frameworks and APIs to present a unified, programmable fabric. In such an environment, AI systems do not need to navigate the complexity of individual networks. They engage with an integrated ecosystem that ensures reliable performance across different regions and providers, supporting a supply chain on demand.

As networks become more autonomous, new considerations also emerge. Security, policy enforcement, and data sovereignty must be embedded directly into the infrastructure. AI systems will always seek the most efficient path, but that path must comply with regulatory, geographic, and organizational constraints. This elevates the role of the network beyond connectivity. It becomes a control layer enforcing policy, ensuring compliance, and safeguarding operations in an increasingly automated world.

As AI networks evolve, robust cybersecurity is crucial for managing and securing diverse AI workloads. Businesses need zero trust policies, microsegmentation, defined boundaries for AI agents, and data loss prevention. While DIY solutions may suit large enterprises, NaaS cybersecurity services will be critical for most businesses' AI infrastructure.

The scale of the AI opportunity is undeniable. But its realization depends on an often-overlooked foundation. Without networks that are automated, deterministic, and globally interoperable, the projected trillions in value will remain out of reach. For the telecom industry, this is a defining moment. The future of AI will not be determined solely by advances in models or compute, but by the infrastructure that connects them. The road to a $15.7 trillion AI economy is being built now, but will the network be ready to support it.