OpenAI Wants Google's Clicks

ChatGPT's Ad Machine

If you own the front page of the internet, your business model usually relies on a very specific type of math. Google handles roughly 14 billion searches every day, but the vast majority of those are economically worthless. If you search for "what time is it in Tokyo" or "how to boil an egg," Google provides the answer, but no advertiser is particularly excited to pay for that interaction.

The trillion-dollar valuation of Alphabet is built on its ability to distinguish between two fundamentally different types of user intent:

• Informational Query: This is the vast majority of search volume. Think "how to fix a broken zipper" or "what is the capital of Spain." These queries generate little to no direct revenue. They are a public utility designed to keep you from using Bing.
• Commercial Query: This is where a user is effectively holding their credit card up to the screen. Searches like "best budget laptops" or "direct flights to Miami" are auctions where advertisers bid ferociously for the final click. This is the tiny, highly profitable engine that funds the whole operation.

The business model of search is to subsidize the first category with the profits from the second.

For the last three years, OpenAI has been the intellectual leader of the AI boom while remaining an also-ran in the search business. Estimates suggest ChatGPT handles about 66 million search-like queries a day, which is roughly 0.5% of Google's volume. If you are Google, you can live with losing the "how to boil an egg" market.

But the market for boiling eggs doesn't pay OpenAI's $1.4 trillion utility bill. Last Friday, the company announced it is finally coming for the money, confirming it will begin testing advertisements at the bottom of ChatGPT responses, starting with its free and cheapest paid tiers.

There is a mechanical irony here. For years, OpenAI resisted ads, citing the need to preserve trustworthiness and objectivity. But OpenAI has also signed up for roughly $1.4 trillion in infrastructure commitments over the next decade. It turns out that objectivity is a luxury that becomes increasingly difficult to maintain when you have to pay for 30 gigawatts of computing power. You eventually reach a point where you might not need a model that is fundamentally aligned with human values; but you definitely need a model that is fundamentally aligned with a high conversion rate.

The Funnel Shift

The real threat to Google isn't that people will stop using a search bar; it's that the "job to be done" is shifting.

Google's structural advantage is owning the discovery and intent capture layer: it routes users from broad needs to options, and it monetizes that routing heavily via ads — especially for commercial intent queries

ChatGPT, however, is designed to collapse the funnel. OpenAI's own data suggests that about 40% of its usage is "doing" rather than "asking"—drafting emails, writing code, or planning itineraries. When a user asks an AI to "plan a 3-day trip to London and find me a hotel with a gym," they are skipping the ten blue links and the five minutes of browsing where Google typically shows them several dozen ads.

By embedding ads directly into that synthesized answer, OpenAI is attempting to intercept the most valuable commercial intent before it ever reaches a traditional search engine. If you can provide the specific hotel recommendation and the ad for it in the same sentence, you have effectively captured the "expensive click" without needing the majority of "zombie" queries that Google has to manage.

The Memory Machine

To make this work, OpenAI is leaning into a feature it calls "memories." Users have been encouraged to let ChatGPT store personal preferences to make conversations more efficient. If you tell the bot you have a gluten allergy or that you prefer flying Delta, it remembers.

From a user experience perspective, this is a convenience. From a balance sheet perspective, it is the world's most precise advertising database. Google has to guess what you want based on your recent search history; ChatGPT knows what you want because you have spent hours explaining your life to it in natural language.

The company expects to make low billions from this in 2026. This is a modest start compared to Google's $200 billion search engine, but the trajectory is what matters. The "code red" that OpenAI triggered last year wasn't just about making the models smarter; it was about the realization that building a magical software brain requires an industrial-scale revenue machine.

For some time now, the bear case for Google was that AI would make search obsolete. The real risk is more mundane: AI might just make search ads more efficient, and OpenAI has realized that it owns the most valuable data set for targeting them. Google's defense has been to launch "AI Mode" with its own bespoke discounts and ads, implicitly admitting that the future of the internet is no longer about finding a website—it is about talking to a merchant who remembers your shoe size.

More on Google Search:

• Google search ad clicks hit five-year high as Q4 spend rises 13% (Search Engine Land)
• Google files to appeal search monopoly case (CNBC)

On Our Radar

Our Intelligence Desk connects the dots across functions—from GTM to Operations—and delivers intelligence tailored for specific roles. Learn more about our bespoke streams.

Hyperscalers Verticalize Energy Supply

• The Story: Alphabet has acquired renewable developer Intersect Power for $4.75B, while Amazon and Meta directly fund SMR nuclear projects, signaling a shift from buying power via PPAs to owning the energy generation infrastructure itself. (WSJ)
• The Operations Implication: This verticalization transforms energy from an operating expense (OpEx) into a massive capital expenditure (CapEx). By bringing energy development in-house, hyperscalers are taking on construction, permitting, and regulatory risks previously borne by utilities. This swells their balance sheets with non-core infrastructure assets, potentially dragging down returns if these energy projects face delays or cost overruns typical of the nuclear/power sector.

P.S. We're building a handful of bespoke intelligence streams for members. If you're facing a specific intelligence challenge, drop us a line here.

You received this message because you are subscribed to ARPU newsletter. If a friend forwarded you this message, sign up here to get it in your inbox.