OpenAI Is About to Disrupt a $1T Industry

For thirty years the advertising business has been built around a single blind spot: what you actually want. It bought your browsing history, your loyalty-card receipts, your location trail, and your age and postcode, then paid data brokers to stitch the pieces into a profile it hoped was close enough to sell against. Acxiom, one of the biggest of those brokers, has boasted about holding as many as eleven thousand attributes on a single person.

Then people started telling ChatGPT what they were trying to do, where to travel, which software to buy, how to handle a diagnosis, what to cook for someone who cannot eat gluten. They are also uploading all kinds of documents, talking about web sites and writing posts for social media. No cookie ever captured that kind of customer context, and no data broker ever sold it. That is the real reason OpenAI walked into Cannes Lions this June and told a room full of marketers, in its own words, that it is “clearly in the advertising business now.”

The clearest sign that OpenAI means it was the person delivering the message. Dave Dugan spent twelve years at Meta, most of them selling ads to the world’s largest brands, before OpenAI hired him in April to build the same machine on top of ChatGPT. When one of the two companies that defined modern digital advertising loses its head of global clients to a chatbot maker, the chatbot maker is not dabbling.

Most people heard the $100 billion forecast and laughed, because the first ads inside ChatGPT will look like sponsored cards, a slightly smarter banner. That reaction is older than the internet. Every new medium arrives looking similar to the one it replaces, so the first television commercials were just radio scripts read aloud on camera, and the earliest web ads were magazine pages pasted onto a screen. The thing that actually rewrites the game never shows up in the first demo, because nobody has built it yet. Now we have a system that hears a customer’s intent in plain language and that can be sold to the highest bidder.

Start with what happened to coding

To see where this goes, consider what AI recently did to writing software. Before AI, building software meant developers using editors, reading documentation, hunting through Stack Overflow, wiring up dependencies, running tests, and fixing whatever the last change broke.

Anthropic’s Claude Code collapsed that entire loop. Now you hand an agent a repository and a task, and it reads the code, makes the edits, runs the tests, and returns work that is mostly finished. Wired reported that Claude Code reached a billion dollars in annualized revenue within about six months of launch and was running at $2.5 billion by early 2026, while OpenAI’s Codex crossed a billion in the same window. All of that happened inside software development, a market far smaller than advertising and far harder to bill for.

That is the pattern we’re going to see in advertising. A slow, human, low-context job got handed to an agent that finally had enough context to do it, and a multi-billion-dollar business appeared almost overnight.

Advertising has always been guessing

Every part of the old ad stack exists to answer one question the industry could never quite reach: who is ready to buy this, right now. Brokers determine your identity, search and social determine your interest, exchanges auction your attention, brands buy the impression, and attribution tries afterward to work out what any of it accomplished. It is an enormous and expensive apparatus built to predict commercial intent, a fact that ChatGPT now receives for free during many of your conversations with AI.

Dugan gave Cannes the number that makes this concrete. Roughly one in five of the conversations people have with ChatGPT carry direct commercial intent, he told the room, across a base that Marketing Dive reported at more than 900 million weekly users. He called those people “super intentional.” Set that against a display ad, which lands on someone in the middle of doing something else, where maybe one impression in a thousand reaches anyone near a decision.

Memory turns intent into a profile

Live intent by itself would already be the most valuable ad surface built since Google Search. Memory is what carries it somewhere the old industry cannot follow. Google owned your keywords and Meta owned your social graph, yet each of them only ever saw a sliver of you. An AI assistant you talk to every day can hold the whole picture, the project you are stuck on, the budget you mentioned last week, the brand that let you down, the trip you take every spring, and so much more.

OpenAI says today’s ads read only the current conversation, that organic answers stay independent, that chats stay private from advertisers, and that it does not sell your data. That last promise is the easy one to keep, because the new stack does not need your data to leave the building. OpenAI never has to hand an advertiser a file on you. With AI memory it can assemble a detailed profile of who you are and what you are about to buy in the moment you ask, match that profile to whichever advertiser fits, and charge for the match while the profile itself never moves. That is what the new ad stack looks like, it runs entirely on memory, and it is why memory is about to become the most contested ground in the business.

Why $100 billion is not a joke

Follow the money OpenAI has already put on the board. It launched the ad product in February, crossed $100 million in annualized revenue within about six weeks with more than 600 advertisers signed on, and has told investors to expect roughly $2.5 billion this year, $11 billion in 2027, $25 billion in 2028, and $100 billion by 2030.

Coding is the reason to take that curve seriously. In little more than a year, AI rebuilt software development from the ground up. The whole workflow changed, an industry that had run the same way for decades was rewired almost overnight, and the revenue followed immediately. Anthropic rode that shift from about $9 billion in annualized revenue at the end of 2025 to a $30 billion run rate by April, and Sacra put it near $47 billion by May, with coding making up close to half.

Advertising is next, and at the same scale. What is coming is not a better-targeted banner or a leaner funnel, it is a new stack built on a signal the old industry never had, carrying buying intent at levels off any chart it has ever measured.

The real asset is context

The people questioning ChatGPT ad numbers are picturing the old system with some new features, which completely misses the scale of what is happening. Advertising exists to answer one question, who wants this thing right now, and the entire apparatus of brokers, exchanges, and tracking pixels was a workaround for not being able to ask the customer directly.

Now the customer opens ChatGPT and says it outright. Acxiom’s eleven thousand data points, the pride of the old world, are a mountain of guesses about who you might be, and they look primitive next to a system that knows exactly what you want, who you are and how you buy.

The advertising industry is about to be completely disrupted and customer context is what is going to power it.