Is There an AI Bubble at Present or Not
ChatGPT recently had its third birthday. Hard to believe, but the application was released only in November 2022. It is even harder to imagine that there was no ChatGPT before then. In just three years, it has moved from novelty to everyday necessity for hundreds of millions of people globally.
Soaring Valuations and the Road to a Trillion-Dollar IPO
What has also changed dramatically is the valuation of the company behind ChatGPT — OpenAI. Even though the company is still private, recent funding rounds reportedly value it at around $700 billion. With plans to go public in 2026, OpenAI could become the first trillion-dollar IPO.
That is, of course, exciting: new technology, rapid adoption, and massive potential. A $1 trillion valuation may sound high, but AI clearly has transformative power. One could argue that companies like Google, Microsoft, or NVIDIA once looked expensive before they grew into their valuations.
Yet there is another side to the story.
OpenAI’s Growing Losses and What They Mean for Sustainability
OpenAI has reportedly been loss-making every year since inception. Initially those losses were modest and largely expected. More recently, widely circulated (though unofficial) estimates suggest:
-
~$1B loss in 2023
-
~$5B loss in 2024
-
~$7–10B loss in 2025
-
~$15B expected loss in 2026
These numbers are not official — OpenAI is private — but they are frequently discussed among analysts. What draws attention is not just the losses themselves, but projections that OpenAI may not reach profitability before 2030. Some estimates even suggest cumulative losses could reach $130–140B by then, implying annual losses of $20–30B for several years.
OpenAI may one day become enormously profitable or even dominant. But for now, the path to profitability appears long. That naturally raises questions: how is a near-trillion valuation justified? Who is comfortable investing at these levels? Where is the margin of safety?
This is where the AI bubble argument gains traction.
Why the “AI Bubble” Narrative Resonates
NVIDIA is rumored to be among the large investors in OpenAI, possibly at multi-billion-dollar scale. For illustration (purely hypothetical), consider the following simplified scenario:
-
OpenAI signs long-term data-center leases with Oracle worth $10B
-
Oracle finances new data centers with $40B in bank loans
-
NVIDIA supplies $20B worth of chips and hardware to Oracle
-
With ~53% margins, NVIDIA recoups its original investment through hardware profits
The argument from AI-bubble proponents is that large AI players often invest in companies that simultaneously become their customers. Even if the equity investment underperforms, the supplier side can generate enough profit to offset it.
We do not necessarily subscribe to this view, but it is worth presenting.
High Valuations, Long Timelines, and Sector-Wide Risk
What is verifiably true is that AI valuations are not cheap, and the probability of those valuations being tested in the future is non-trivial. If projections of prolonged unprofitability turn out correct, technology investors should at least remain cautious.
Yes, Amazon was unprofitable for much of its early life. But today we are not discussing one company — we are discussing an entire sector where many firms may remain loss-making for years. Not all of them can become the next Amazon. Some will succeed, some will not.
Watching the Credit Markets for Early Signals
We do not claim to know whether there is an AI bubble today. What we do know is that the possibility exists. For now, we will be watching credit markets closely, particularly for highly leveraged infrastructure players such as Oracle, and take it from there.