This Is What an AI Monopoly Looks Like — And It’s Already Here

Most people think of a monopoly as one company running the show. A single villain, easy to spot, easy to point at. But the most dangerous kind of market concentration doesn’t look like that at all. It looks like five perfectly reasonable, extremely innovative companies quietly becoming the gatekeepers of an entire industry.

That’s exactly what’s happening in AI — and the mortgage and real estate worlds are sitting right in the middle of it, whether they know it or not.

From 60% to 67% in Twelve Months

In early 2024, five US hyperscalers — Google, Microsoft, Meta, Amazon, and Oracle — controlled about 60% of global AI computing power. By the end of 2025, that number had grown to 67%. That’s a meaningful jump in a very short amount of time, and every indicator suggests it will keep climbing.

The reason it’s climbing is the sheer cost of staying in the game. Building the kind of computing infrastructure needed to train and run advanced AI models costs billions of dollars. Not millions. Billions. That kind of price tag doesn’t just create a barrier to entry for startups — it creates a barrier that eliminates entire categories of would-be competitors before they even get started.

The Winner-Takes-All Machine

Economists have a term for markets that naturally gravitate toward one or a few dominant players: winner-takes-all dynamics. AI infrastructure has them in spades.

The more computing power you own, the more AI models you can train. The better your models, the more customers pay to use them. The more customers you have, the more revenue you generate to buy more computing power. Around and around it goes, with the big players getting bigger and everyone else getting squeezed.

This is not a conspiracy. It’s just math. And it’s the same math that gave us a world where two companies control most of your smartphone experience, three companies control most of your internet search, and four companies deliver most of your online shopping.

Now it’s happening to the technology that is rapidly becoming the backbone of how mortgages get processed, how leads get generated, how properties get valued, and how clients get served.

The Labs Themselves Are Dependent

Here’s one of the most remarkable — and underreported — facts in all of tech right now: even the most powerful, most celebrated AI companies in the world don’t own their own computing infrastructure.

OpenAI, the company behind ChatGPT, is almost entirely dependent on Microsoft’s computing resources. Anthropic, the company behind Claude, recently announced a major expanded partnership with Google to secure more chip capacity. These aren’t small companies scraping by — OpenAI is one of the most highly valued startups in history. And yet they rent their computing from the hyperscalers just like everyone else.

When the biggest AI labs in the world are tenants rather than landlords, that tells you something important about who actually holds power in this ecosystem.

Antitrust Is Coming — But Slowly

Regulators in the US and Europe are beginning to pay attention to this concentration. Antitrust investigations into big tech’s grip on AI infrastructure are underway in multiple countries. The concern isn’t just that these companies are big — it’s that their dominance in cloud computing gives them a structural advantage in AI that’s almost impossible for competitors to overcome.

But antitrust enforcement is notoriously slow. Cases take years. Markets move in months. By the time regulators act, the consolidation will be even deeper than it is today.

One legal analysis from Yale Law noted that antitrust enforcement operates after the fact, case by case, while AI markets are moving so fast that significant harm could be locked in well before any regulatory remedy arrives.

What Does ‘AI Concentration’ Feel Like to a Loan Officer?

Right now, it might not feel like much. AI tools for mortgage professionals are proliferating. Competition among vendors seems healthy. Prices seem reasonable.

But look one or two layers deeper. The companies selling you AI tools are themselves paying the hyperscalers to run those tools. As hyperscaler pricing shifts — and it will — those costs will work their way down to you. The question is when and by how much.

Beyond pricing, concentration creates fragility. If your AI platform runs on one hyperscaler’s infrastructure and that company decides to change its terms, raise its rates, or simply discontinue support for a particular type of workload, your vendor has limited alternatives. And so do you.

The Mortgage Industry Needs to Pay Attention Now

Fannie Mae estimated that 55% of lenders would be using AI by the end of 2025. That number is almost certainly higher today. AI is no longer an experiment in the mortgage world — it’s becoming operational infrastructure. Which means the question of who controls the infrastructure it runs on is no longer abstract. It’s a business continuity question.

The good news is that awareness is the first step. Understanding that your AI tools don’t exist in a vacuum — that they sit on top of a concentrated, powerful, increasingly expensive infrastructure layer — changes how you evaluate vendors, negotiate contracts, and think about your technology strategy.

Start asking the questions now. Because the companies at the top of this stack are going to keep getting bigger. Count on it.

Subscribe to Well That Makes Sense — because in an industry moving this fast, being informed isn’t optional. It’s your competitive edge. (And honestly, we make it a lot more fun than your CE credits.)