Walk into any boardroom in 2026 and you’ll smell the same thing: anxiety. It’s that specific, prickly scent of executives realizing they might have spent $50 million on a chatbot that mostly just summarizes emails poorly. Everyone is asking the same question. Is AI a bubble? Honestly, if you look at the stock charts of companies like Nvidia or the staggering energy bills from Microsoft’s data centers, it’s hard not to feel a bit of déjà vu. We’ve been here before. We saw it with the dot-com crash in 2000. We saw it with the fiber optic craze.
The money being thrown at "Generative AI" is frankly absurd. We are talking about trillions of dollars in market cap based on the hope that Large Language Models (LLMs) will eventually do more than just write mediocre poems or hallucinate fake legal citations. But here’s the thing. Bubbles aren't just about things being "fake." Usually, a bubble happens because something is very real, but everyone gets the timing wrong.
Why people think the AI bubble is about to pop
Look at the Capex. Capital expenditure is a fancy way of saying "spending money to make money later." Big Tech is spending like drunken sailors on GPUs. In early 2024, Goldman Sachs released a report titled "Gen AI: Too Much Spend, Too Little Benefit?" that sent shockwaves through the Valley. Their lead equity research analyst, Jim Covello, pointed out a glaring problem. For a technology to justify a $1 trillion investment, it has to solve $1 trillion worth of problems.
Right now? It’s mostly solving "how do I write this LinkedIn post faster" problems.
The cost of inference—the actual act of the AI thinking and giving you an answer—is still high. It takes a massive amount of electricity. We are literally reopening nuclear power plants, like Three Mile Island, just to keep the servers running. If the productivity gains don’t show up in the actual GDP numbers soon, investors are going to stop being patient. They’ll pull the plug. When that happens, the "Is AI a bubble" conversation stops being a theory and starts being a market crash.
The "S-Curve" problem and diminishing returns
There is this idea in tech called the S-curve. You start slow, you have a massive vertical explosion of progress, and then you plateau. We might be hitting the plateau faster than people want to admit.
GPT-4 was a massive leap over GPT-3. But the jump to the next generation? It's getting harder. Training these models requires "high-quality data." The problem is, we’ve already scraped the entire public internet. Now, AI models are starting to eat their own tails, training on AI-generated content, which leads to "model collapse." It’s like a digital version of inbreeding. The quality starts to drop. If the models stop getting significantly smarter, the business case for spending billions more on them starts to look pretty shaky.
The ghost of 1999 is haunting Silicon Valley
I remember the pets.com era. People think the dot-com bubble meant the internet was a scam. It wasn't. The internet changed everything. But in 1999, the infrastructure was too expensive and the use cases were too thin. You couldn't stream movies on a 56k modem.
AI feels identical.
We have the "broadband" of AI (the models), but we don't have the "Netflix" or "Uber" of AI yet. Most companies are just "wrappers." They take OpenAI’s technology, put a blue UI on it, and charge you $20 a month. That’s not a sustainable business model. That’s a feature, not a company.
Investors like David Cahn at Sequoia Capital have talked about the "AI 600 Billion Dollar Question." He calculated that the gap between the revenue needed to pay for the AI infrastructure and the actual revenue being made is massive. Like, cavernous. You can only bridge that gap with hype for so long before the gravity of math takes over.
Real-world friction you can't ignore
- Regulation: The EU AI Act is already making things expensive for companies.
- Copyright: Artists and writers are winning discovery motions in court.
- Energy: The grid literally cannot handle a 10x increase in data centers without a total overhaul.
- Accuracy: In fields like medicine or law, "90% right" is actually "100% fired."
It’s not a bubble if it actually works, right?
Wait. Let’s look at the other side.
Is AI a bubble if it’s actually saving lives? In 2024, Google DeepMind’s AlphaFold 3 basically solved the protein folding problem. That’s not a chatbot. That’s a tool that accelerates drug discovery by decades.
✨ Don't miss: Why feedback control of dynamic systems franklin Still Rules Engineering Classrooms
And then there’s coding. If you talk to any senior software engineer who isn't a total luddite, they’ll tell you that tools like Cursor or GitHub Copilot have made them 30% to 50% faster. That is a real, tangible economic gain. It’s just not as "sexy" as a robot that talks to you.
The bubble might just be in the "consumer" side of AI. The "enterprise" side—the boring stuff like optimizing logistics, writing boring SQL queries, and analyzing massive datasets—is very much real. The stock market might crash, but the technology isn't going away.
The "AI Summer" might turn into an "AI Winter"
In the history of computer science, we’ve had multiple "AI Winters." These are periods where the hype died, the funding dried up, and everyone went back to working on regular databases. We had one in the 70s and another in the late 80s.
If we are in an AI bubble, the "pop" won't be the end of AI. It will just be a massive "reset."
The companies that are just "GPT wrappers" will vanish overnight. The companies that own the actual "picks and shovels"—the chips and the energy—might take a hit, but they’ll survive. The real winners will be the ones who figure out how to use AI to do something that was literally impossible before, not just something that was "slightly annoying" to do manually.
🔗 Read more: Silver Sulfate Chemical Formula: Everything You Actually Need to Know
How to spot the "Pop" before it happens
You'll know the bubble is popping when you see these three things:
- Big Tech companies start mentioning "efficiency" more than "AI" in their earnings calls.
- The "Pro" versions of these tools start getting significantly cheaper because no one is paying for them.
- Venture Capitalists start pivoting to the "next big thing" (maybe biotech or space tech).
Honestly, we're already seeing bits of this. Microsoft and Google are under immense pressure to show that their "Copilots" are actually increasing revenue. If the next few quarters of earnings show that companies are canceling their AI subscriptions because the ROI isn't there, watch out.
What you should actually do about it
If you’re a business owner or just someone trying to keep your job, don't get blinded by the "Is AI a bubble" headlines. Even if the market crashes, the skill of using these tools is still the most valuable thing you can learn right now.
Think of it like 2001. If you learned how to build a website in 2001, you were still ahead of the game, even if the stock market was on fire.
Stop looking for "AI solutions" and start looking for problems. If you have a problem that takes a human 10 hours and an AI can do it in 10 seconds with 95% accuracy, use it. If you’re trying to force AI into a process just because it sounds "innovative," you’re part of the bubble.
Actionable Steps to Navigate the AI Hype
- Audit your Subscriptions: Look at every AI tool you pay for. If you haven't used it to actually save time or make money in the last 30 days, cancel it. The "hype tax" is real.
- Focus on Workflow, Not Tools: Don't worry about whether GPT-5 is coming. Master the tools that exist now. Learn prompt engineering, but more importantly, learn how to integrate these outputs into your actual work.
- Watch the Energy Sector: If you want to know if AI is still growing, look at energy stocks and utility companies. If they are building, the AI companies are still buying.
- Keep Your Data Clean: The biggest hurdle for AI in any business isn't the AI—it's the messy, disorganized human data. Organize your files. Clean your CRM. That makes you "AI-ready" regardless of which model wins.
The AI bubble might pop tomorrow. It might pop in three years. Or, we might just "grow into" the valuation like we did with the internet. But the underlying tech is the most transformative thing we've seen since the smartphone. Just don't bet your entire retirement fund on a company that claims its chatbot has a "soul."
It’s just math. Very expensive, very fast math.