How to Do Marketing After the AI Bubble Pops

Last month, Google's parent company did something it had never done in twenty-two years as a public company: it burned cash.

Alphabet's Q2 2026 numbers look spectacular at first glance. Revenue up 24% to $119.8 billion. Net income up 298% to $112.1 billion. Twelfth straight quarter of double-digit growth.

Then you read the fine print. Free cash flow was negative $5.86 billion, the first quarterly cash burn in the company's history as a listed business. Capital spending hit $44.9 billion, roughly double the year before, which means Google ran at 115% of the cash its own business generated. Executives raised their 2026 spending guidance to as much as $205 billion, the second hike this year.

And that record profit? The single largest driver was a $99 billion paper gain from marking up Alphabet's roughly 14% stake in Anthropic, after Anthropic's valuation tripled from $380 billion to $965 billion in a single quarter. (Financial Express has the full breakdown.)

Read that again. The AI profit was a markup on their own AI investment.

Google invests in Anthropic. Anthropic buys compute from Google and Oracle. Oracle buys Nvidia GPUs. Nvidia's valuation depends on the AI labs spending. The money goes in a circle, and everyone marks everyone else up along the way. Alphabet has also taken on close to $100 billion in debt and sold shares for the first time in more than twenty years. The four biggest hyperscalers are on pace to spend over $725 billion on capital investment this year.

I am not writing this to be an AI doomer. I run my business on AI. But I read that earnings report the way I'd read a client's funnel numbers, and what I see is a business model that only works while the loop keeps spinning.

Someday, that bill comes due. And when it does, the marketers who are fine will be the ones holding skills that don't depend on the loop.

Why AI Feels Indispensable to Marketers

First, credit where it's due. The reason AI has felt like magic to marketers specifically is not the chatbots. It's the workflows.

For the first time, a marketer can sit down and wire up a system that watches for a new lead, scores it, updates the CRM, triggers the right email sequence, notifies the sales rep, and logs the outcome. Programmatically. Without waiting on a developer. Tools like Zapier, Make.com, and n8n handle the plumbing, and AI handles the glue that used to require actual code.

That part is genuinely new. In 2019, that system was a six-week engineering project with a five-figure invoice. Today it's an afternoon.

I use this stuff every day. My side projects openclaw and Draftly run on AI infrastructure that did not exist three years ago. I wrote about the vendor lock-in risk already, and I stand by that too.

But here is the uncomfortable part. The AI we know and love today is not going to exist in its current form in five years. The subsidies will shrink, the prices will normalize, some of these companies will merge or die. The specific tools are rented.

Which raises an obvious question. If the tools are temporary, what should you actually get good at?

Four skills. None of them require a subscription.

1. Funnel Building

Maybe the most important one. Build the sequence from the first touch a customer has with your website all the way to the follow-up email sequence that pushes them to upgrade.

That entire system, first touch to repeat purchase, is largely what good marketing still looks like today. It was true before AI and it will be true after. The tools that run it, Zapier, Make.com, or the automation built into most CRMs, will keep working no matter what happens to the model companies. If every AI lab vanished tomorrow, the funnel you wired up in Zapier keeps firing.

The marketer who can design that sequence, know which message lands at which stage, and read the drop-off points in the data, is carrying a skill that transfers to any stack. It's also the core of the conversion optimization work I do for clients, and it's the part that stays valuable when the tooling underneath changes.

2. Design and Taste

Most AI systems are genuinely bad at design. Not bad the way a junior designer is bad, where feedback fixes it. Bad in a structural way: they produce overly templated layouts with no built-in sense for how things should look.

You can see the evidence everywhere. AI graphics with too much text to read. No hierarchy, so your eye doesn't know where to land. A pretty obvious disregard for originality, because the model is averaging a million portfolio sites and the average of a million things is nobody's taste.

Design taste is a moat. It's the difference between a site that feels trustworthy and a site that feels generated, and buyers can tell in about three seconds. The good news is taste is learnable. Look at great work constantly. Notice why it works. Steal composition, never surface. The marketers who can look at a page and know that something is off, and fix it, will out-earn the ones who can only prompt for another version.

3. The Marketing Funnel Itself

Distinct from building automated funnels: you need to be intimately familiar with how your specific customers move from problem to purchase.

How do they first become aware of the problem they have? Word of mouth, search, a competitor's ad? How do they research solutions, and who do they trust while they research? What has to be true before they'll convert on your site, and where does the journey usually stall?

Every channel changes. Search became AI answers, email became DMs, something else will change next year. The journey underneath is human and slow-moving. A marketer who deeply understands their customers' awareness-to-decision path can port that understanding to any channel that shows up. One who only knows how to operate the current channel starts from zero every time the ground shifts.

I wrote a whole piece on marketing-optimized web design that covers how this thinking shapes what goes on the page itself.

4. In-Person Marketing

If there's anything my admittedly anti-social self has learned about marketing, it's that people are always at the center of it.

Testimonials matter. Videos of real people matter. Faces, names, voices, the slight imperfection of an actual human saying what they actually think. Humans are social creatures. We buy from people we trust, and trust is built through people, which is exactly why the testimonial section converts and the stock-photo banner doesn't.

AI can draft a testimonial request. It cannot be the testimonial. As AI content floods every feed, the accounts, brands, and businesses that put real humans front and center will win attention by default, because everything around them will be increasingly indistinguishable slop. In-person is unfair advantage. Go to the event. Film the customer. Put your own face on it. My most effective marketing asset is a calendar link behind a real conversation.

What's Your Move?

Maybe the loop keeps spinning for another decade and I'm early. Being early is cheap. Being unprepared is not.

The test I'd use: pick any skill you're depending on right now and ask whether it survives if your AI subscriptions 10x'd in price tomorrow, or vanished outright. The four above pass that test for me. Funnels run without AI. Taste compounds. Customer journeys outlive channels. People trust people.

The bill comes due eventually. Build like it's tomorrow and you won't have to think about the date.

If you read this and think I'm wrong, that the subsidies hold and the loop spins forever, I genuinely want to hear that take. Find me on LinkedIn and make the case.