Look, I'm going to be blunt about this because somebody has to be. I thought about this all night and woke up at 4am to write, because I'm clearly in need of therapy.

Every CMO in America right now is high-fiving themselves because they fired their agency, replaced three copywriters with Claude, and watched their "efficiency metrics" go up. They're saving money! They're innovating! They read a McKinsey report and now they're AI-first!

Great. Congratulations. You know who else is celebrating? Google. Because while you were busy saving $8,000 a month on your agency retainer, Google's ad revenue jumped 13% year over year to $74.18 billion in a single quarter. Your cost per click went up 10%.

Your impressions went down 15%. And the AI tools you're paying for? They're training every other advertiser in your category to bid against you harder, faster, and with better creative.

You didn't save money. You funded your own competition.

According to Gartner's 2025 CMO Spend Survey, 39% of CMOs plan to cut agency budgets and 22% say generative AI has already let them reduce their reliance on external agencies. Meanwhile, 60% of senior US marketing leaders say they're spending less on agencies in 2025 as a direct result of AI, according to a Typeface survey.

Sounds like a win, right? Now look at the other side of the ledger.

US paid search ad spend is projected to grow over 12% in 2025, reaching $144 billion. But impressions are down 15% year over year, which means you're paying more to reach fewer people. Average Google Ads CPC rose from $4.01 in 2022 to $4.66 in 2024, a 10% average annual increase, with sectors like real estate seeing jumps over 25%.

You cut your agency. That saved you maybe $60,000 to $180,000 a year. Your CPCs went up 10%. If you're spending $500,000 a year on Google Ads, that's $50,000 gone. Add the $24,000 to $120,000 you're now spending on AI tool subscriptions that didn't exist three years ago. Add the internal time your team spends prompt-engineering, QA-ing AI output, and managing the new stack.

Net savings? Somewhere between zero and "you actually spent more."

Here's what's actually happening, and it's so obvious it's almost embarrassing that more people aren't talking about it.

When everyone in an auction gets a better bidding algorithm, the auction gets more expensive. That's it. That's the whole thing.

AI-powered bidding drives aggressive spending because the algorithms are designed to win conversions, not to save you money. Google's smart bidding, your fancy third-party optimization tool, your competitor's fancy third-party optimization tool: they're all fighting each other in real time, ratcheting bids upward in a computational arms race that no human media buyer would ever wage.

CPC inflation between 2024 and 2026 has accelerated across nearly every industry. This is not a blip. It's a fundamental transformation driven by AI, privacy changes, and aggressive automation.

Mark Read, the former CEO of WPP, the largest advertising company on Earth, was refreshingly honest about this at London Tech Week. He told CNBC that AI disruption is "unnerving investors in every industry" and is "totally disrupting our business." He didn't sugarcoat it. He said AI would "make all the world's expertise available to everybody at extremely low cost." Which sounds great until you realize that "everybody" includes your competitors, and "low cost" means the competitive floor just got raised for everyone simultaneously.

Maurice Levy, CEO of Publicis Groupe, echoed the point. He told CNBC that AI-driven tools are enabling "personalization at scale like never before." But personalization at scale for everyone is just... the new baseline. It's table stakes. You're not special because your AI personalizes ads. Everyone's AI personalizes ads.

Let me explain the mechanism, because it's almost elegant in how efficiently it transfers money from advertisers to platforms.

Step 1: Google launches AI Overviews. These AI-generated summaries appear above search results and answer user queries directly. Google itself acknowledged this "Great Decoupling" in June 2025: impressions remain high, but click-through rates to websites are falling because users get answers from the AI summary.

Step 2: Your organic traffic craters. Businesses that relied on SEO for free visibility are now forced into paid search.

Step 3: More advertisers competing for fewer clicks drives CPCs up.

Step 4: Google's ad revenue goes up. Your costs go up. You buy AI tools to "optimize" your way out of it. Those tools make your bidding more aggressive. Which drives CPCs up more.

It's a beautiful flywheel. For Google.

Leela Nair, Ebiquity's Managing Director for Asia Pacific, put it plainly: "This isn't a typical investment cycle, it's an AI arms race where infrastructure doesn't automatically translate to revenue." She said her clients now demand tangible commercial returns from AI pilots, ending the era of exploratory tests. Translation: the honeymoon is over.

Here's something the AI companies themselves are barely keeping a lid on. This stuff is expensive to run. Not expensive like "oh we need a bigger marketing budget" expensive. Expensive like "we might go bankrupt" expensive.

Dario Amodei, CEO of Anthropic (the company behind Claude), said the quiet part out loud at the New York Times DealBook Summit. He warned that if revenue timing is off by even a year, an AI company faces bankruptcy. "If my revenue is not $1 trillion, if it's even $800 billion, there's no force on Earth, there's no hedge on Earth that could stop me from going bankrupt if I buy that much compute."

He went further, criticizing unnamed competitors (cough, OpenAI, cough) for reckless spending. "I think there are some players who are 'YOLO-ing,' who pull the risk dial too far, and I'm very concerned."

Mark Cuban has been even more direct. He said the major AI players are "spending everything, consuming every resource that they can just in case it's winner take all." He compared it to the 1990s search engine wars and warned that the bubble could burst fast. "They may be overspending, and if they overspend or get too caught up, the bubble is in the competition between all those models because that could pop just like that with any new technology."

And the cost of actually running AI agents? Cuban flagged this too, responding to a clip from the All-In Podcast where investors Jason Calacanis and Chamath Palihapitiya revealed the real-world expense of deploying AI agents: in some cases over $300 per day, adding up to over $100,000 annually. Cuban called it the "smartest counter" to predictions that AI will replace large numbers of workers.

Cuban's take on AI agents was characteristically blunt: "Agents are still like college interns that come in hungover, make mistakes, and don't take responsibility for them."

Someone is paying for all of this compute. And if it's not the AI companies' investors (who are getting increasingly nervous), it's going to be you, the advertiser, through higher API costs, higher subscription fees, and the relentless upward pressure on every auction you participate in.

Now here's where it gets really uncomfortable.

AI in advertising doesn't democratize. It amplifies. If you're a small business spending $2,000 a month on Google Ads, a 15% CPC increase means 15% fewer clicks. You can't absorb that. You cut back. You lose visibility.

If you're a Fortune 500 company spending $10 million a month, that same 15% CPC increase barely registers, because your AI systems are simultaneously running thousands of creative variants across dozens of channels, executing predictive models on customer lifetime value, and finding efficiencies that simply don't exist at the small-business scale.

The wealthy enterprise can run AI agents around the clock, doing sustained complex work at scale. Legal review, financial modeling, competitive intelligence, full campaign orchestration. These aren't $20-a-month chatbot tasks. These are heavy compute workloads that cost real money and generate real competitive advantage.

The small business gets a chatbot that helps draft Instagram captions. The enterprise gets a digital marketing department that never sleeps. Both are "using AI." The outcomes are not remotely comparable.

Cuban nailed this dynamic when he was asked about AI replacing workers. The real question, he said, isn't whether AI is capable. It's whether the economics work. And right now, the economics work really well if you can afford to spend $100,000 a year on AI agents. They don't work so well if your entire marketing budget is $100,000.

There's one more dimension to this that deserves attention, because it's coming fast and it's going to change everything.

OpenAI has started putting ads in ChatGPT. Anthropic ran Super Bowl commercials explicitly promising Claude would stay ad-free. The battle lines are being drawn.

Cuban has been vocal about this too. He warned about putting ads inside AI model responses: "If we just want to put up little display ads, JPEG ads, in the sidebar, great. But if it's in the response that comes back from the model, boy, you're going to have lots of problems."

He's right. And the implications for advertisers are massive. When AI models start recommending products, services, or solutions as part of their "neutral" responses, and those recommendations are influenced by advertising dollars, the entire concept of trusted AI advice collapses. Every recommendation becomes suspect. Every "objective" analysis could be bought.

Cuban didn't mince words: "The last thing we need is to have algorithms designed to maximize revenue driving LLM output and interactions. They are already recommending brands and we don't know if they are getting paid for it."

Follow the money. It's not complicated.

Google and Meta win. Higher CPCs, more advertisers forced into paid search, AI Overviews cannibalizing organic traffic. Their revenue goes up regardless of whether your ROI improves.

AI tool companies win. Every marketing team now needs AI subscriptions. New mandatory cost layer. Congratulations, you're paying rent to a landlord that didn't exist four years ago.

Large enterprises win. They deploy AI at transformative scale, finding real efficiencies that generate massive competitive advantages. The ROI is genuine. But you have to be able to afford the compute to get there.

Small and mid-size businesses lose. Higher platform costs, new tool subscriptions, reduced organic reach, competition from AI-powered incumbents. You were promised democratization. You got a more expensive treadmill.

Daniela Amodei, co-founder and president of Anthropic, summed up the broader dynamic with a single sentence that should keep every CFO up at night: "The exponential continues until it doesn't."

The advertising industry is living through a massive wealth transfer, and most of the people participating in it don't realize they're on the wrong side.

AI tools are real. They work. They generate real efficiencies for the people who use them. But in a competitive auction system, individual efficiency gains get socialized into higher prices for everyone. The savings don't stay with you. They migrate to the platforms that run the auctions and the companies that sell the tools.

This is not a new pattern. It's what happened with programmatic advertising. It's what happened with social media marketing. It's what happened with SEO. Every time a new tool promises to make advertising cheaper and easier, the competitive equilibrium adjusts, the floor rises, and the platforms capture the surplus.

The difference this time is the scale. We're talking about hundreds of billions in infrastructure spending, trillion-dollar valuations, and energy consumption that rivals entire countries. The stakes have never been higher, and the gap between the winners and losers has never been wider.

So the next time someone tells you AI is going to cut your advertising costs, ask them a simple question: Cut them compared to what?

Because if the answer is "compared to last year before everyone else had the same tools," then sure. For about five minutes.

After that, you're just running faster to stay in the same place. And Google is counting every step.