Here's something strange happening right now in Silicon Valley: the same people pouring hundreds of billions of dollars into artificial intelligence will openly tell you it might all come crashing down.
Sam Altman, the CEO of OpenAI, said it plainly at a press dinner in August 2025: "Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes." He compared the current moment to the dot-com bubble. The next day, the stock market dipped.
Mark Zuckerberg has talked about how AI bubbles work — infrastructure gets built out, companies take on too much debt, then "a lot of the companies end up going out of business." Google's CEO Sundar Pichai warned that "no company is going to be immune" when it bursts. Anthropic's Dario Amodei said some players are "YOLOing" their way into disaster.
And yet the spending continues. The investments keep flowing. The data centers keep getting built.
I've been trying to make sense of this for months now. How can everyone acknowledge they're inflating a bubble and keep inflating it anyway? What exactly is the AI bubble? Who's most at risk? And what happens when reality finally catches up?
Let me walk you through what I've learned.
What Is the AI Bubble?
The AI bubble refers to a theorized stock market bubble growing amid the current artificial intelligence boom — a period of rapid, possibly irrational investment that's affecting the broader economy. The concern isn't that AI itself is fake or useless. The concern is that valuations have become disconnected from reality, that companies are worth far more on paper than their actual business fundamentals justify.
Think of it like this: AI is probably a genuinely transformative technology. The internet was too. But during the dot-com boom, investors got so excited about the internet that they threw money at any company with ".com" in its name, regardless of whether that company had a viable business model. When the bubble burst in 2000, trillions of dollars evaporated. Companies that seemed unstoppable disappeared overnight.
The question now is whether we're repeating that pattern with AI.
As of late 2025, the signs are concerning. AI-related enterprises accounted for roughly 80% of gains in the American stock market over the year. The S&P 500's ten biggest stocks now represent about 40% of the entire index — a level of concentration not seen since the 1960s. Share valuations are the most stretched since the dot-com bubble, with the Case-Shiller price-to-earnings ratio for the US market exceeding 40 for the first time since that crash.
But here's where it gets complicated: unlike the dot-com era, today's big tech companies are actually profitable. They're generating real revenue. The question isn't whether AI works — it clearly does. The question is whether the spending is proportionate to the returns, and whether the current valuations can be sustained.
The Numbers That Should Make You Nervous
Let me throw some numbers at you that I find genuinely difficult to wrap my head around.
OpenAI has committed to spending more than $500 billion on AI data centers — more than 15 times what was spent on the Manhattan Project, adjusted for inflation. Sam Altman has reportedly told employees that OpenAI's moonshot goal is to build 250 gigawatts of computing capacity by 2033, roughly equaling India's total national electricity demand. Such a plan would cost more than $12 trillion by today's standards.
Combined capital expenditures from Microsoft, Alphabet, Amazon, and Meta are expected to reach roughly $440 billion over the next year, a 34% increase. Spending from US mega-caps is expected to reach $1.1 trillion between 2026 and 2029.
Meanwhile, consultants at Bain estimate that the wave of AI infrastructure spending will require $2 trillion in annual AI revenue by 2030 just to justify the investment. That's more than the combined 2024 revenue of Amazon, Apple, Alphabet, Microsoft, Meta, and Nvidia.
And here's the kicker: despite all this spending, a widely cited MIT study from July 2025 claimed that 95% of organizations that invested in generative AI were getting "zero return." The study itself was more nuanced than the headlines suggested, but it felt like the first hard data point confirming what skeptics had been muttering for months.
The two most prominent private players in AI, OpenAI and Anthropic, have yet to turn a profit. OpenAI is projected to burn through $140 billion by 2029. Anthropic is expected to burn $20 billion by 2027. For comparison, Amazon burned through $3 billion before becoming profitable. Tesla burned around $4 billion. Uber burned $30 billion. The scale of AI losses is in a different universe.
The DeepSeek Shock: A Preview of What Could Come
If you want to understand how quickly the AI bubble could deflate, look at what happened on January 27, 2025.
That morning, a Chinese startup called DeepSeek — with a valuation around $150 million and fewer than 200 employees — released an AI model that matched OpenAI's performance at roughly 2% of the cost. DeepSeek claimed it had spent just $5.6 million on computing power for its base model, compared with the hundreds of millions or billions that American companies spend.
The result was the largest single-day loss in stock market history.
Nvidia lost $588.8 billion in market value in a single day. That's more than the total value of all but 13 companies in the world. Over $1 trillion evaporated from US tech markets as investors suddenly asked a question they'd been avoiding: "What am I getting for $1 billion that I couldn't get for $6 million?"
The Nasdaq plunged 3.1%. Tech giants across the board saw their stocks drop. The entire premise of the AI boom — that you need massive infrastructure investments to compete — was suddenly in question.
Nvidia's shares recovered somewhat in the following days. By the end of 2025, the stock had regained much of its lost ground. But the DeepSeek shock demonstrated something important: the AI bubble isn't invulnerable. A single unexpected development can wipe out hundreds of billions of dollars in perceived value.
And DeepSeek wasn't a fluke. It was a demonstration that the "compute moat" — the assumption that whoever spends the most wins — might not be as solid as investors believed. If algorithmic efficiency can substitute for raw spending power, then a lot of expensive infrastructure might become unnecessary.
What's Inflating the Bubble?
Three main forces are pumping air into the AI bubble right now.
The first is genuine demand for AI capabilities. This isn't imaginary. Companies do want AI. ChatGPT reached 100 million users faster than any application in history. OpenAI's revenue is running around $13 billion a year, and Anthropic is targeting $9 billion in 2025 run-rate. The technology works, and people are willing to pay for it.
The second force is infrastructure buildout. The leaders of top AI companies all stress that they're bottlenecked by limited access to computing power. Startups can't get the GPU allocations they need. Hyperscalers are rationing compute. If the market really is as supply-constrained as tech leaders claim, aggressive infrastructure buildouts might be warranted.
But the third force is where things get dicey: circular financing.
Here's how it works. Nvidia invests $100 billion in OpenAI. OpenAI uses that money to buy Nvidia chips. This makes Nvidia's sales numbers look strong, which props up Nvidia's stock price, which gives Nvidia more money to invest in AI companies, which buy more chips. The money flows in a circle, and everyone's valuations go up.
It's not just Nvidia and OpenAI. Microsoft owns a major stake in OpenAI. OpenAI has a $300 billion deal with Oracle. CoreWeave is financed partly by Nvidia and rents computing capacity to OpenAI. OpenAI took a 10% stake in AMD. The web of interconnected investments is dense and confusing — and that's partly the point.
During the dot-com bubble, this kind of circular financing helped hide weak end-demand until it couldn't be hidden anymore. Equipment vendors would finance customers who would then buy their equipment, making sales numbers look healthy while actual demand lagged. When the mismatch became undeniable, the bubble burst.
Anthropic's Dario Amodei has flagged this concern publicly. "If you start stacking these where they get to huge amounts of money," he said, "and you're saying, 'By 2027 or 2028 I need to make $200 billion a year,' then yeah, you can overextend yourself."
Who's Exposed — And Who Thinks They're Safe?
One of the stranger dynamics of this bubble is how tech leaders use bubble talk to position themselves as winners while implying their competitors will lose.
Altman says the excess is in small startups with "three people and an idea" getting funded at crazy valuations. "That's not rational behavior," he said. "Someone's gonna get burned there, I think." He positions OpenAI — which has raised over $10 billion and is valued at $500 billion — as fundamentally different from those risky startups.
Amodei implies that OpenAI is the one "YOLOing" by making commitments too aggressive to sustain. He positions Anthropic as more prudent, despite the fact that Anthropic is also expected to burn billions before profitability.
Zuckerberg argues that Meta can absorb expensive miscalculations because of its strong cash flow. OpenAI and Anthropic, he suggests, risk bankruptcy if they misjudge timing — but not Meta.
Google's Pichai warns that no company is immune, which sounds like humility but also serves to spread the blame if things go wrong.
Everyone agrees there's a bubble. No one thinks they'll be the ones left holding the bag.
The reality is probably messier. Large tech companies with diversified businesses and strong cash flows — Meta, Google, Microsoft, Amazon — can probably survive a correction. They have other revenue streams to fall back on. The companies most at risk are the ones that are highly leveraged, unprofitable, or entirely dependent on AI hype for their valuations.
That includes not just small startups, but potentially OpenAI and Anthropic themselves. It includes infrastructure companies that have bet everything on continued AI spending. It includes the data center builders, the energy companies hoping to supply AI's electricity needs, and the semiconductor companies whose valuations assume ever-increasing demand.
How Could the Bubble Burst?
Based on conversations tech executives and investors have had publicly, the bubble is most likely to pop if overfunded companies can't turn a profit or grow into their valuations.
This could happen in several ways.
The most straightforward scenario: AI revenue growth slows or stalls. Companies that made massive infrastructure commitments based on projections of exponential growth suddenly find themselves overbuilt and underwater. The headline deals that propped up stock prices come into question. Investors start asking hard questions. Stock prices fall. The circular financing stops flowing.
A second scenario involves technological disruption — something like DeepSeek but bigger. What if a breakthrough in chip design or quantum computing makes current infrastructure obsolete? What if someone figures out how to train AI models at 1% of current costs? Hundreds of billions in data center investments could become stranded assets overnight.
A third scenario involves interest rates. The Federal Reserve's rate cuts in 2024 and 2025 created a supportive environment for growth stocks. Rising rates would make speculative investments less attractive and could trigger a correction in overvalued tech stocks.
A fourth scenario involves regulation or geopolitical disruption. Export controls on AI chips, restrictions on data usage, antitrust actions against dominant players — any of these could change the economic calculations that current valuations depend on.
The honest answer is that nobody knows exactly how or when the bubble will burst. Goldman Sachs compared the current AI boom to where tech stocks were in 1997 — several years before the dot-com bubble actually burst. We might have more room to run. Or the correction could start tomorrow.
What This Means for You
If you're an investor, the AI bubble should make you cautious but not necessarily paralyzed. AI is probably a genuinely transformative technology, and some companies will build enormous, sustainable businesses around it. The trick is figuring out which ones.
History suggests that timing matters enormously. If you were an Amazon shareholder from its 1997 IPO to now, you've done phenomenally well — even though the stock dropped 90% during the dot-com crash. If you were a Pets.com shareholder, you lost everything.
Diversification is probably your friend. The concentration of market value in a handful of tech stocks is historically unusual and potentially dangerous. If AI stumbles, the effects will ripple far beyond the companies directly involved.
If you work in tech, the bubble should inform how you think about career decisions. Companies flush with bubble money can offer generous salaries and exciting-sounding projects. But if the funding dries up, layoffs follow. The tech industry shed hundreds of thousands of jobs in 2022-2023 when the pandemic-era bubble deflated. An AI correction could trigger similar upheaval.
If you're a consumer, the bubble probably doesn't affect you much in the short term — except that you might notice AI features being aggressively pushed into products that don't obviously need them. Companies desperate to show AI traction will find ways to add AI to everything, whether or not it improves the user experience.
The Strangest Part: Everyone Knows
What makes this moment surreal is the honesty.
Previous bubbles caught people by surprise. The dot-com crash, the housing crisis — in each case, warnings existed but most participants believed the good times would last forever. This time is different. The people inflating the bubble are the same ones publicly warning about it.
OpenAI chairman Bret Taylor put it well: "I think it is both true that AI will transform the economy, and I think we're also in a bubble, and a lot of people will lose a lot of money. I think both are absolutely true at the same time."
He compared it to the internet. Webvan failed, but Instacart succeeded years later with essentially the same idea. The technology was real even though the bubble was real. When the dust settles, society benefits from the inventions that survive — even as many investors lose fortunes.
Jeff Bezos made a similar point in October 2025: "This is real. The benefit to society from AI is going to be gigantic." But individual companies and investors can still get crushed along the way.
Maybe AI will deliver returns spectacular enough to justify current valuations. Maybe efficiency breakthroughs will make infrastructure spending look wise in retrospect. Maybe the bubble deflates slowly rather than bursting catastrophically.
Or maybe not. As Altman said at that dinner: "Someone is going to lose a phenomenal amount of money. We don't know who."
FAQ
What is the AI bubble?
The AI bubble refers to a theorized stock market bubble growing amid the current artificial intelligence boom. The concern is that valuations of AI companies and related stocks have become disconnected from fundamental business realities, driven by speculation and hype rather than actual profits and sustainable growth. The bubble involves massive capital expenditures on AI infrastructure, circular financing arrangements between tech companies, and stock prices that may not be justified by current revenues.
Are we currently in an AI bubble?
Most tech leaders — including Sam Altman (OpenAI), Mark Zuckerberg (Meta), Sundar Pichai (Google), and Dario Amodei (Anthropic) — publicly acknowledge signs of a bubble. A July 2025 MIT study found 95% of organizations investing in generative AI were getting "zero return." Stock market concentration in AI-related companies has reached levels not seen since the dot-com era. However, unlike dot-com, today's major AI companies are generating real revenue, which complicates the picture.
How is the AI bubble different from the dot-com bubble?
Key differences include: (1) Today's major tech companies are profitable and generating substantial revenue, unlike many dot-com era companies; (2) AI technology is already deployed at scale with proven use cases; (3) Valuations, while stretched, are lower relative to earnings than dot-com peaks — the top AI-related stocks trade at roughly 26x forward earnings versus 70x+ during dot-com; (4) The bubble is being discussed openly by participants rather than ignored. However, concerns about circular financing and overinvestment mirror dot-com patterns.
What is circular financing in the AI bubble?
Circular financing refers to arrangements where companies invest in each other in ways that can artificially inflate valuations. For example, Nvidia invests $100 billion in OpenAI, which uses that money to buy Nvidia chips, boosting Nvidia's sales figures. Microsoft owns a stake in OpenAI while being a major customer. CoreWeave receives financing from Nvidia while renting computing capacity to OpenAI. Critics argue these arrangements can mask weak end-demand and create a house of cards.
What was the DeepSeek shock?
On January 27, 2025, Chinese startup DeepSeek released an AI model matching OpenAI's performance at roughly 2% of the cost — claiming just $5.6 million in training expenses versus billions spent by American companies. This triggered the largest single-day loss in stock market history, with Nvidia losing $588.8 billion in market value. The event demonstrated that the AI bubble could burst rapidly if the assumption that massive spending is necessary for AI success proves wrong.
How much money is being spent on AI infrastructure?
The numbers are staggering. OpenAI has committed to spending over $500 billion on data centers. Combined capital expenditures from Microsoft, Alphabet, Amazon, and Meta are expected to reach $440 billion in the next year. US mega-cap spending on AI infrastructure is projected at $1.1 trillion between 2026-2029. Consultants estimate this spending would require $2 trillion in annual AI revenue by 2030 just to justify the investment — more than the combined revenue of the six largest tech companies.
Which companies are most at risk if the AI bubble bursts?
The highest-risk companies are those that are unprofitable, highly leveraged, or dependent entirely on AI hype. This potentially includes OpenAI (projected to burn $140 billion by 2029) and Anthropic ($20 billion by 2027), despite their prominence. Infrastructure companies, data center builders, and energy companies serving AI demand are also exposed. Larger diversified companies like Meta, Google, Microsoft, and Amazon have other revenue streams and are considered more resilient.
What could trigger an AI bubble burst?
Potential triggers include: (1) AI revenue growth slowing below projections, making infrastructure investments look overbuilt; (2) Technological breakthroughs that reduce the cost of AI dramatically, like DeepSeek demonstrated; (3) Rising interest rates making speculative investments less attractive; (4) Regulatory actions or geopolitical disruptions; (5) A major AI company failing or being exposed as less capable than claimed. Goldman Sachs compares current conditions to 1997, suggesting the bubble could continue for years — or burst sooner.
Is AI itself a bubble, or just AI stocks?
AI technology is real and delivering genuine value. The bubble concerns center on valuations, spending levels, and expectations — not the technology itself. Even when bubbles burst, the underlying technology often continues advancing. The internet survived the dot-com crash; e-commerce eventually fulfilled its promise through survivors like Amazon. Similarly, AI will likely transform the economy regardless of what happens to current stock prices.
Should I sell my AI stocks?
This isn't financial advice, but the situation warrants caution. Diversification can protect against sector-specific crashes. Historical precedent shows that during bubbles, even fundamentally sound companies can see their stock prices drop dramatically — Amazon fell 90% during dot-com before its eventual recovery. Consider your risk tolerance, time horizon, and whether you can stomach potentially years of losses before any recovery. Professional financial advice is recommended.
What happens to the economy if the AI bubble bursts?
AI-related investment now accounts for a significant portion of GDP growth and stock market gains. A sharp correction could trigger broader economic effects: layoffs in tech and related sectors, reduced capital expenditures, declining stock portfolios affecting consumer spending, and potential credit market stress from debt taken on for AI infrastructure. However, the overall economy is more diversified than it was during dot-com, which could limit contagion.
How long could the AI bubble last?
Nobody knows for certain. Goldman Sachs suggests current conditions resemble 1997 — three years before the dot-com peak. Private markets move more slowly than public markets, potentially extending the timeline. However, unexpected events like DeepSeek demonstrate that corrections can happen suddenly. Tech executives suggest watching for: declining corporate profits, rising debt levels, Fed rate cuts (which actually support bubbles), and widening credit spreads.
Final Thoughts
The AI bubble is real. The people creating it know it's real. And they're inflating it anyway because the potential upside — if AI delivers on even half its promises — is too large to ignore.
Bret Taylor's framing stuck with me: two truths existing at once. AI will probably transform the economy. And a lot of people will probably lose a lot of money. Both things can be true simultaneously.
For investors, workers, and observers, the challenge is navigating this uncertainty without either panicking or becoming complacent. The technology is real. The hype is also real. Separating them is harder than it looks.
What I keep coming back to is something Jeff Bezos said about the original internet boom: "When the dust settles and you see who the winners are, society benefits from those inventions." Pets.com died. Amazon survived. The internet changed everything regardless.
The same will probably be true for AI. Some companies will become the Amazons. Many more will become the Pets.coms. And the rest of us will try to figure out the difference before it's too late.
The most honest thing anyone has said about this moment came from Sam Altman himself: "Someone is going to lose a phenomenal amount of money. We don't know who."
For now, we're all just hoping it isn't us.
Related Articles
Mark
Dani
