“There is a lot of talk about an AI bubble,” Nvidia CEO Jensen Huang said on the company’s latest earnings call. “We see something different.”
Of course he does. When you’re sitting on a mountain of revenue that just jumped 62% year-over-year, it probably feels like the world is perfectly sane. Nvidia pulled in $57 billion last quarter — above every analyst prediction — and Wall Street basically threw confetti in the air.
And right now, most people seem convinced that this proves there’s no bubble.
No crash coming.
Nothing to worry about.
But… I’m not buying it.
Not for a second.
What Nvidia showed us wasn’t “proof the AI boom is real.”
It was a bright neon sign saying “This tower is getting unstable.”
Let’s dig into the part everyone’s ignoring.
The Core Misunderstanding
People keep repeating the same flawed logic:
“If Nvidia is selling shovels like crazy, the gold rush must be legit.”
No. This is the same thinking that fooled people during tulip mania, the railroad bubble, the dot-com explosion, and the crypto boom of 2021.
High demand for tools doesn’t mean the underlying industry is healthy.
It just means everyone thinks it will be.
That’s how bubbles work.
And right now, Nvidia is selling the best shovels humanity has ever seen — H100s, H200s, Blackwell — the whole GPU zoo. But the buyers? That’s where the story starts to wobble.
The Fragile Cascade, Pt. 1 — The Data Arms Race
Here’s the part that freaks me out:
Most AI companies aren’t competing on product anymore.
They’re competing on compute burn.
If you’re not training a 200-billion-parameter model, you’re invisible.
If you’re not promising AGI by 2027, you’re “falling behind.”
So everyone is hoarding GPUs like they’re rare Pokémon cards.
But the data to fuel these models?
It’s… running out.
Companies are:
- scraping the internet dry
- buying old books and digitizing them
- licensing entire news archives
- paying TikTok creators for speech data
- training on synthetic data that may already be degrading model quality
If you zoom out, it looks less like innovation and more like desperation.
A bubble isn’t when people stop buying.
A bubble is when people can’t stop buying without collapsing their own hype.
Guess where we are now.
The Fragile Cascade, Pt. 2 — The Money Trail
Follow the money and things get even uglier.
A huge portion of AI spending isn’t driven by profitable usage.
It’s driven by:
- venture capital FOMO
- startups burning investor cash to appear “competitive”
- cloud providers subsidizing GPU access to lock customers in
- mega-corps racing to claim they’re “AI-first” before earnings season
Right now, the entire AI boom is propped up by the assumption that future returns will justify present-day spending.
This exact pattern has appeared before every major tech crash.
Over-investment on the belief that “this time is different.”
Jensen Huang says AI demand is real.
Sure. But demand driven by hype is still demand… until it isn’t.
The Fragile Cascade, Pt. 3 — The Output Problem
Here’s the uncomfortable truth:
Most companies using AI don’t actually know how to make money with it.
Everyone has:
- a chatbot
- a summarizer
- an “AI assistant for X”
- a slide generator
- a prototype agent system that kind of works when the planets align
But real, durable revenue?
In many industries, that part hasn’t arrived yet.
You can’t train trillion-dollar models forever on wishful thinking.
And once CFOs start asking harder questions —
like “Why is our cloud bill bigger than our revenue?” —
that’s when the cascade starts.
Nvidia’s Earnings Matter — But Not the Way You Think
Don’t misunderstand me. Nvidia is a phenomenal company.
They executed better than anyone else in the industry.
They deserve the win.
But Nvidia’s numbers don’t prove “AI is healthy.”
They prove money is flowing — aggressively — into one direction.
And that’s the whole problem.
Because if that flow slows even slightly:
- AI startups collapse
- GPU demand shrinks
- cloud GPU rentals dry up
- hyperscalers cut capex
- Nvidia’s numbers tank
- investors panic
- the market avalanches
That’s what a bubble looks like.
Not chaos.
Not noise.
A beautiful, towering structure that stands perfectly still —
right until it doesn’t.
But Is This All Doom and Gloom?
Not necessarily.
Plenty of bubbles didn’t destroy the underlying technology.
The dot-com crash wiped out thousands of companies…
but the internet survived. Thrived, actually.
Air travel had massive early bubbles.
Electricity.
Railroads.
Automobiles.
Tech evolves through cycles of excess and correction.
AI will be no different.
But pretending “Nvidia’s earnings prove AI is safe” is delusional.
They prove nothing except that we’re at the most manic stage of the hype curve.
So… Is There an AI Bubble?
Yeah.
Almost definitely.
Is it popping tomorrow?
Probably not.
But the cracks are visible.
The incentives are distorted.
The spending is unsustainable.
The expectations are impossible.
If this sounds negative, good.
Somebody has to say it.
Because right now, everyone’s busy staring at Nvidia’s mountain of gold.
But if you squint — just a little —
you can see it’s really a tower of cards with excellent lighting.
— MASHRAF AIMAN
AGS NIRAPAD Alliance
Co-founder, CTO, OneBox
Co-founder, CTO, Zuttle
Top comments (1)
There is a lot of Al