When Intelligence Becomes Cheap, What Will Marketing Still Be Worth?

IBM’s $240 million agreement with Together AI may look like another infrastructure announcement about GPUs, cloud capacity, and faster AI. For marketers, the more interesting question lies elsewhere: What happens when intelligence itself becomes cheap enough to be used almost everywhere, all the time? The answer could change where marketing creates value—and what remains distinctly human.

IBM’s multi-year agreement with Together AI will bring a large cluster of NVIDIA HGX B300 systems to IBM Cloud, with the infrastructure expected to become available in early 2027. Together AI plans to use that capacity to run large-scale inference for open models.

On the surface, this is a technology story, but marketers should pay attention.

For the past few years, the AI debate has focused largely on what models can do. Which one writes better? Which reasons better? Which generates better images?

The next phase may be shaped by a less glamorous question: “How cheaply can intelligence be produced?” And that could prove far more disruptive.

From using AI to operating through AI

Training creates an AI model. Inference is what happens every time someone actually uses it. Every generated ad, product recommendation, customer-service response, campaign analysis, or agent decision consumes inference capacity. Asking AI to create ten campaign concepts costs relatively little....