After past technological shifts, the media industry is seeing the value move back toward distribution. For years the mantra “content is king” guided distributors to partner with creative businesses to capture IP, attention and valuation multiples. Recently, however, the emphasis is swinging the other way: the networks that deliver content to users are regaining importance.
A recent example is Comcast’s announcement this week that it will part ways with NBC and Sky News and return to being primarily a broadband provider — effectively reversing its media expansion that began in 2009. Verizon and AT&T reached a similar view after their own unsuccessful attempts to reinvent themselves as media companies.
Comparable dynamics have appeared in other industries. In banking, regulators cracked down on practices that steered customers into in-house funds, prompting a shift to “open architecture” models that largely separated distribution from product manufacturing. That change has raised questions about why banks still own asset managers. In the oil industry, drillers have also divested marketing and gas-station networks.
What does this mean for artificial intelligence? Large language models (LLMs) are, in essence, the content: the outputs, answers and text they produce. But companies that build those models — including OpenAI, Google and Anthropic — are also trying to own the pipes that bring those outputs to users, through consumer-facing apps and developer-facing APIs.
So far the AI revolution lacks a clear analogue of a phone company: a distribution network that knows it is a neutral carrier and doesn’t try to be the creative star. Several scenarios are plausible given current signals:
- A major existing software company — for example Salesforce or Microsoft — could become an open highway, available to many different AI models, acting as a neutral distribution layer.
- Alternatively, large model owners might continue to combine content and distribution, owning both the models and the channels that deliver them.
- A third path is that AI-native hardware and devices — which figures such as Sam Altman and Jony Ive are reportedly working on — create new channels that reduce the need for a pure-play distributor.
There are concrete signs of players moving to build distribution infrastructure. Anthropic announced last month that it acquired Stainless, a developer tools company, suggesting model makers are beginning to assemble parts of the pipes themselves. At the same time, the market is not yet settled: about a billion people use ChatGPT directly each month, Google is integrating Gemini into its browser and is powering Apple’s Siri intelligence, and multiple actors are vying to define how models reach end users.
Why this matters
- Competition: if one or a few platforms control the distribution channels, model diversity and market access could be constrained.
- Regulation: different regulatory approaches may be needed depending on whether content and distribution are combined or separated.
- Consumer and developer choice: openness of distribution affects which models and services are available to users and creators.
In short, the AI ecosystem has not yet resolved whether a neutral, distribution-focused player will emerge or whether dominant platforms will keep combining models and delivery. Comcast’s move, past banking shifts, and steps like Anthropic’s Stainless acquisition all show that the relationship between content and pipes will remain a central issue.
