Everyone’s throwing money at AI like it’s the new gold rush.
Over $100 billion flowing into AI startups annually. OpenAI raised $40 billion at a $300 billion valuation while hemorrhaging cash. Anthropic, Databricks, xAI—each pulling in billions. We’ve got 150+ AI unicorns now, with new ones minted every month.
Everywhere you look: another mega-round, another pitch deck with “AI-powered” slapped on it, another founder promising to revolutionize an industry they learned about last Tuesday.
I’ve spent 15 years building things—real products, for real companies, that actually have to work. I’ve seen hype cycles come and go. And I’ll tell you this much: history doesn’t reward the people who follow hype. It rewards the ones who understand cycles.
So here’s the question keeping everyone up at night: Are we in an AI boom that will define the next decade, or a bubble waiting to pop?
The Money Is Absolutely Insane
AI companies are capturing over 40% of all venture capital in the US despite being less than 20% of funded companies.
This isn’t normal. This is frenzy.
Generative AI alone pulled in nearly $50 billion. Average contracts are hitting $530,000. One in three companies is investing over $25 million in AI infrastructure.
But here’s what nobody wants to say out loud: most of these companies aren’t making money. They’re not even close.
I’ve Seen This Movie Before
Major financial institutions are issuing warnings about a “sharp market correction.” The Fed Chair is calling valuations “fairly highly valued”—central banker speak for “this looks bad.” Some analysts say this bubble is larger than dot-com.
The symptoms are textbook.
Circular deals everywhere. Nvidia invests in OpenAI. OpenAI buys Nvidia chips. Nvidia invests in xAI. xAI buys more Nvidia chips. Oracle builds data centers using—surprise—Nvidia chips. It’s vendor financing with extra steps. We saw this in the late ’90s telecom bubble.
Fundamentals don’t matter. OpenAI is valued at $300 billion while losing billions annually. They’d need to 10x revenue just to break even. Yet investors keep writing checks like profitability is optional.
Everything is “AI-powered” now. Companies that were struggling to raise six months ago added “AI” to their deck and tripled their valuation.
FOMO is driving decisions. Investors are funding companies they don’t understand, in markets they haven’t researched, because they’re terrified of missing the next OpenAI.
If you’ve been in tech for more than five minutes, you know how this story ends.
But Here’s Why It Might Not Be Crazy
Look, I’m naturally skeptical. But the bull case isn’t just wishful thinking.
AI isn’t a product. It’s infrastructure. Like electricity. Like the internet. And unlike the dot-com era where adoption was speculation, this is happening right now.
78% of companies are already using AI in at least one business function. That number was 55% two years ago. Generative AI adoption jumped from 33% to 71% in the same period.
The economics are improving fast. The cost to run high-performance AI models has dropped by over 280x in two years. Computing power per dollar has fallen 75%.
Companies are actually paying for this. Average enterprise AI contracts exceed half a million dollars. Real operational budgets are being reallocated—not innovation theater money.
The sectors seeing ROI aren’t theoretical. Customer support automation is saving millions. Predictive analytics catches problems before they become disasters. Document processing that took days now takes minutes.
The difference from past bubbles? The infrastructure is being used before it’s fully built. The internet took a decade to get mainstream adoption. AI got there in two years.
The Part Nobody Talks About: Most of This Is Garbage
Here’s what I see every day building automation systems for B2B companies.
There are two types of AI startups, and the difference is everything.
First, the buzzword crew. They took ChatGPT’s API, wrapped it in a UI, wrote “AI-powered” forty times in their pitch deck, and called it revolutionary. Zero proprietary tech. No defensible moat. Entirely dependent on OpenAI not eating their lunch.
I sit through demos thinking, “I could build this in a weekend.” And if I can build it in a weekend, it’s not a business—it’s a feature.
Then you’ve got the builders. They’re solving real problems that cost real money. Cutting qualification time from three hours to five minutes. Scaling outreach from 50 to 500 contacts without adding headcount. Reducing response times from six hours to five minutes and watching conversion rates climb.
The framework is dead simple: painkiller vs. vitamin.
A tool that automates contract review and cuts legal processing from 5 days to 5 hours? Painkiller—it saves $50K in labor costs. A chatbot that helps you “brainstorm better”? Vitamin—gets cancelled when budgets tighten.
Painkillers solve urgent, expensive problems. When budgets get tight, you protect painkillers and cut vitamins.
Most “AI startups” are vitamins. The ones that survive are painkillers.
What Happens Next
The correction is coming. It always does.
Smart money is already shifting. The days of “grow at all costs” are over. Investors want unit economics. Retention rates. Proof that customers would riot if you disappeared.
Over the next 6-18 months, we’ll see consolidation. A few platforms will own entire categories. Most will shut down. The survivors won’t be the ones with the best demos—they’ll be the ones with the best economics and the deepest integrations.
This isn’t speculation. This is pattern recognition.
My Take: It’s Both, and That’s the Point
Yes, we’re in a bubble. Probably 70% of current AI valuations are nonsense.
But we’re also in a legitimate boom. Real infrastructure is being built. Real productivity gains are happening. Real businesses are emerging.
The internet bubble was real—and it also gave us Google, Amazon, and the foundation of modern commerce. Most dot-coms died. The ones that survived changed everything.
Same thing is happening now. The bubble will pop. Valuations will crater. Most of today’s AI darlings will vanish. But beneath the hype, something durable is being built.
The question isn’t whether there’s a bubble—there obviously is. The question is whether you’re building something that survives the pop.
If you’re building in AI, here’s my advice:
Build something that saves time or makes money. Ideally both.
Solve expensive problems, not interesting ones.
Embed deeply into workflows. Make yourself painful to remove.
Focus on retention over acquisition.
And for the love of everything holy, stop adding “AI-powered” to your pitch deck unless the AI is actually the product.
The bubble will pop. The boom will continue.
Make sure you’re building for the boom.
