GenAI is a battle of the titans. Huge capital investments and giant frontier models, competing to gain the trust of consumers and enterprises. The models are complex, but the business model is pretty simple: customers pay the AI model developer for a subscription. This is true for consumers and enterprises.
As specialty models and AI agents come onto the scene, it’s getting complex. Which model should you use to book your travel? To write a wedding speech? To draw an engineering diagram? To analyze the stock market? It will grow a hundred times more complex as physical AI enters the mainstream.
The early days of “let’s just sign up for the best model” are ending. We are starting the phase of: “We need help to determine the best model for this task."
Defining the “best model” is not an easy task. Do you pick the most accurate model? The fastest model? The cheapest model? Or the model with the right sovereignty? Over the next 10 years, we expect the choices to explode, so shopping around will become difficult.
Who do you trust to pick the model for you?
In the world of AI agents and physical AI, this is not an easy question. I propose that telecom operators can be the best choice to act as an “AI broker." As your connectivity provider, your telco can identify and authenticate each AI agent, to ensure that every workflow is authorized and to govern how many tokens it can use. The telco can also set guardrails for sovereignty, cost, and latency to make sure that you get results in the right way.
Mobile operators in South Korea are well underway on this journey already. SK Telecom and KT have both deployed significant computing resources as a distributed network, as well as high-level AI models in centralized data centers. Their government is protecting them from American AI competition, so they have boldly stepped forward to take a leading role in providing the AI inference itself. Their role as the natural broker to steer agents to various models will come along as well, as their sovereign protection forces outside AI players to partner with them for access to the Korean market.
We can expect to see this business play out in many other markets, from Japan to France and Germany and India. My forecast lays out a future path for sovereign AI that develops into edge AI and also a way to monetize authentication/governance.
In the United States, mobile operators don’t have sovereign protection from their government, so it’s an uphill climb. But mobile operators have weapons that nobody else can fire. First, the mobile operator has more information about the user (person or robot) than anybody else. The operator knows the location, identity, and billing tier of each customer. The operator already manages hundreds of millions of SIM cards in a very secure core network. Extending that to 100 billion AI agents is not out of their league — it’s the natural next step.
Mobile operators need to move fast if they want to take the position as AI broker. Google will certainly put together an agent to do this for you. Microsoft and Anthropic will certainly have their versions as well…and we will see them well before 6G comes to market.
The telco has natural advantages in the position that they occupy and in their ability to secure and govern usage. The ability to provide network assurance and even a guaranteed wireless link under specified conditions cannot be matched by the hyperscalers.
Come on guys, you need to get moving!
Joe Madden is principal analyst at Mobile Experts, a network of market and technology experts that analyzes wireless markets.
Opinions from industry experts, analysts or our editorial staff do not represent the opinions of Fierce Network.