The self-funding AI bubble: how artificial intelligence became a closed circuit of capital and belief
- Manelik Sfez
- Oct 16
- 5 min read
Start searching the phrase "AI bubble" and Google hesitates, then starts completing your thought: "is AI a bubble", "will AI bubble burst", "AI bubble 2025". That’s not coincidence. It’s probably collective intuition surfacing. The world has begun to suspect what markets still deny: that much of the AI boom isn’t organic demand, but a feedback loop, a self-funding circuit of capital and narrative.
“We don't know who [will lose in this AI bubble] — someone is going to lose a phenomenal amount of money.” — Sam Altman from his interview with The Verge.
1. The AI bubble and the illusion of infinite demand
AI’s current momentum looks unstoppable. Every quarter, a handful of companies — Nvidia, Microsoft, Oracle, CoreWeave, AMD — report record revenues and announce new alliances. But if you trace where the money actually circulates, you find something strange.
Nvidia sells chips to Microsoft and Oracle. Oracle builds infrastructure that hosts OpenAI. Microsoft invests in OpenAI and pays again for its cloud usage. CoreWeave rents Nvidia’s chips to both. Each dollar spent reinforces the same small circle.
This is not a market; it’s an echo chamber with a balance sheet. A self-referential system in which growth in one node strengthens the others, creating the illusion of unstoppable expansion.
At first, that loop feels efficient. Then it becomes speculative. Because it no longer measures the world’s adoption of AI, only the industry’s faith in its own reflection.
2. The AI bubble and the productivity myth
The foundation of this valuation spiral is a simple promise: that AI will multiply productivity by ten, automate work, and free humans to focus on higher value tasks.
It’s a compelling narrative. And largely untrue. The reality inside most companies is closer to what few will admit: AI saves some time, but rarely the 90% everyone imagined. The average uplift is closer to 10–20%, and often neutralized by the effort required to verify, rewrite, or integrate what AI produces.
Accuracy, reliability, and autonomy are still inconsistent. Large language models hallucinate, miss context, and require human supervision that erodes efficiency. The net gain is real but small.
Add to that a deeper bottleneck: data readiness. Most organizations lack the structured, clean, and connected data ecosystems that AI needs to operate meaningfully. Without that foundation, AI remains a clever front-end over messy back-ends. In other words, the theory doesn’t yet work in practice.
3. Amplified intelligence, amplified mediocrity
Here’s where the philosophical problem begins in my opinion. Used right, AI isn’t “artificial intelligence” at all, it’s amplified intelligence. It can accelerate thought, help materialize ideas faster, and extend human capability. It’s the most extraordinary instrument ever built for thinking at scale.
But amplification cuts both ways. AI doesn’t discriminate between brilliance and banality. It scales whatever it’s given. A sharp mind becomes sharper; a mediocre one becomes industrially mediocre.
That’s where the disappointment lies. Instead of unlocking unique human creativity, AI often smooths it out. It standardizes tone, aesthetic, and reasoning. It makes everything sound the same: clean, efficient, lifeless.
This is the paradox of automation applied to thought. When the tool designed to enhance originality begins to erase personality, something fundamental is lost.
4. The AI bubble and the faith economy
AI’s current valuation rests less on product performance than on collective belief. Every bubble has a story. The dot-com era promised a new digital economy. Crypto promised decentralized freedom. AI promises synthetic intelligence, or a civilization-scale shift in cognition itself.
The bigger the story, the harder it is to question. Investors, media, and executives form a circular narrative where everyone’s incentives align around maintaining belief. The more you say “AI changes everything,” the more you’re expected to act as if it does.
And because the main players — Nvidia, Microsoft, OpenAI, Oracle — are simultaneously suppliers, clients, and shareholders of one another, the line between revenue and reinvestment disappears. This is how faith becomes balance sheet reality.
5. When the loop breaks
Every closed system carries its own weakness. If one of the major nodes slows down, the illusion of infinite growth unravels quickly.
Imagine Nvidia’s margins tighten, or Microsoft decides that AI infrastructure isn’t producing proportional returns. That slowdown would cascade through cloud demand, chip orders, and AI valuations instantly.
Because the ecosystem is circular, not expansive, there’s little external demand to absorb the shock. The market would discover that what looked like widespread adoption was, in many cases, just internal consumption: companies building tools for themselves and counting it as global growth.
The result wouldn’t necessarily be catastrophic, but it would be disorienting. Markets would have to recalibrate to a world where AI’s commercial impact is slower, patchier, and less miraculous than forecast.
6. The AI bubble and the credibility deflation
Unlike the 2000s or 2008, this won’t be a collapse of liquidity. It’ll be a collapse of narrative. When reality fails to match expectations, valuation deflates. Confidence erodes. Hype converts into fatigue. Enterprises quietly pause projects; governments slow their digital strategies; headlines shift from “AI gold rush” to “AI disappointment.”
Beyond just economic, the danger is cultural. Disillusionment tends to swing to cynicism. What began as over-belief could end in under-trust, making future breakthroughs harder to fund and adopt. The tragedy is that AI needs to mature, not collapse.
7. After the noise
The real story of AI will start once the noise fades. When the hype layer burns off and the survivors begin building what actually works: leaner models, cleaner data architectures, real integrations that augment, and not replace, human judgment. That’s where the long-term economic and creative value lies.
But to reach that stage, the market has to go through its necessary correction: the deflation of inflated expectations, the return of philosophical discipline, the rediscovery that technology without wisdom is noise.
As I wrote recently:
“An artificial intelligence without natural intelligence at the helm will always remain sterile at best, and toxic at worst.”
AI will not destroy or save humanity; it will mirror it. The systems we build will amplify the intelligence or the carelessness of those who build them. So perhaps the question isn’t whether the AI bubble will burst. It’s whether we can rediscover intelligence before it does.
About the author
Manelik Sfez, founder of the Swiss brand consultancy Ultrabrand, brings 25 years of international business, marketing, and brand strategy experience to the table. He has worked with some of the world’s most iconic brands throughout his career. From luxury goods to global retail, financial services and technological and industry giants, he has guided companies through brand-led transformations that have enabled significant business growth.
FAQ
Is AI in a bubble right now?
Many analysts believe so. The same few companies fund and consume each other’s AI growth, creating a self-reinforcing loop that looks like external demand but isn’t.
What causes the AI bubble?
The core driver is circular investment: chip makers, cloud providers, and AI developers financing each other’s expansion while markets price infinite growth.
Will the AI bubble burst in 2025?
If enterprise adoption stalls or ROI disappoints, the market could experience a credibility deflation—less a crash than a loss of faith.
What happens when the AI bubble bursts?
Expect slowed investment, consolidation around a few giants, and renewed focus on genuine productivity gains rather than narrative-driven valuations.
How is the AI bubble different from past tech bubbles?
It’s more integrated. AI underpins chips, cloud, software, and data infrastructure simultaneously—so any correction could ripple across multiple sectors.
Can AI still create real value?
Absolutely. When used as amplified intelligence—to extend human insight rather than replace it—it can deepen creativity and efficiency.
What’s the long-term outlook for AI?
After the correction, AI will mature into essential infrastructure—quiet, stable, and far more useful than its current hype cycle suggests.
How can businesses prepare for a potential AI correction?
Focus on measurable use cases, data readiness, and human-AI collaboration rather than speculative automation promises.
Comments