Annie Tsai

AI didn’t create it, but some AI companies are capitalizing on it in a way we’ve never seen before.

For a century, monopolies were vertical. Oil. Rail. Telecom. Each industry built its own stack and controlled it end to end: supply, distribution, pricing, and labor. But even then, a far less flashy form of power was developing the entire time – the infrastructure play. Infrastructure was never glamorous, but it was the layer every other industry depended on. Western Union’s telegraph lines, AT&T’s long-distance network, Visa’s payment rails, and the U.S. interstate highway system didn’t sell a product, but rather connected everyone else’s products.

Building and owning those unsexy layers required massive upfront capital when it was hard to find and expensive to build. Yet infrastructure has always been the necessary work before innovation could happen in new categories. Control the pipes, control the flow.

Then the internet arrived and made those dependencies visible and modularized every industry. You no longer had to own the whole thing to capitalize on the infrastructure play. Instead, you just had to be the best at one layer that everyone else plugged into.

Apple then showed us with the iPhone and App Store that controlling the ecosystem meant controlling the economics. By curating distribution, gating data, and owning the interface, Apple proved that the most defensible position in a horizontal world wasn’t the product. It was the store in which every product lived.

APIs finished the job. They made those dependencies interoperable: one payment API powering thousands of stores, one cloud platform hosting millions of businesses.

And Amazon showed that connecting millions of stores could make the sum of those parts greater than the whole. They harnessed network effects, logistics, and data to become the infrastructure behind commerce itself. Power started moving across verticals rather than up and down within them.

So no, AI didn’t invent the horizontal monopoly, but it is certainly having a coming out party.

The same boring infrastructure layers such as compute, cloud, semiconductors, and data pipelines are now the profit centers of the AI era. Every model, every interface, every decision runs through those pipes.

The New Layer of Power

When Walmart embeds ChatGPT into its shopping experience, it’s about much more than convenience. Offering curious browsers the lowest friction path to convert is the dream of all dreams for every merchant, but it also gives OpenAI its own competitive advantage by centralizing ChatGPT as every human’s virtual assistant.

A single conversational layer now sits between the consumer and every product, brand and retailer. Nvidia chips? Same play. Whoever sets the cost of compute sets the interest rate on innovation. Microsoft’s Copilot? A uniform cognitive layer across the standard of business tools that much of the enterprise uses. The interface itself becomes the product.

In today’s world, this is the infrastructure play.

What This Means for Businesses

First and foremost, the infrastructure play in the AI era isn’t over, but the lanes are beginning to narrow. Chips, cloud, models, orchestration and interface each have frontrunners, but none are unassailable. Infrastructure may standardize the way TCP/IP did, and new opportunities will emerge yet where context meets capability.

The question businesses should be asking isn’t “How are we marketing in the AI era?” but “How are we training?” Differentiation depends in the consistency and quality of experience, regardless of how that experience is tapped, and how well you embed your data, tone, and logic into AI systems that consumers use to access your brand matter even more. Training, not advertising, is the vehicle for brand intelligence.

You may be quick to say that branding is dead in the AI era, but IMO branding is even more important than ever before. As businesses have less control over decentralized front-end experiences, brand is what remains consistent when distribution is ambient – it’s your recognizable voice, quality, and ethical stance that carry across every interface where you appear. More importantly, delivering a sustainable repeat purchase or relationship driven model leveraging loyalty has to be a part of the playbook at some point in the customer lifecycle.

In today’s world, brands must now compete by owning the context and hope that owning the customer relationship comes in due time. Building proprietary data loops and vertical expertise that AI models must reference in order to surface recommendations appropriately when it needs to is the key to unlocking conversions in this new sales channel.

Your Takeaway

Infrastructure has always looked boring until it’s understood to be irreplaceable. So ask yourself: What part of your business becomes more valuable the more others rely on it, and how will you train – not just market – that advantage?