The AI bubble

The tech world can't stop debating whether AI is a bubble. Trillions of dollars are flowing into infrastructure, valuations are stretched to dot-com-era levels, and most companies investing in generative AI are seeing zero measurable returns. It's a fair question. But asking "is AI a bubble?" misses the more interesting one: what kind of bubble is it, and what gets left behind when it pops? The short answer: yes, there's inflated demand. Yes, there's an enormous amount of capital chasing speculative returns. But no, AI is not going away. The better comparison isn't NFTs. It's the internet.

The numbers are hard to ignore

Worldwide AI spending is forecast to hit $2.52 trillion in 2026, a 44% year-over-year increase according to Gartner. Morgan Stanley estimates nearly $3 trillion of AI-related infrastructure investment will flow through the global economy by 2028, with more than 80% of that spending still ahead. Goldman Sachs expects hyperscalers alone to invest over $500 billion in 2026. At the same time, the returns aren't matching the spend. A widely cited 2025 MIT Media Lab report found that despite $30 to $40 billion in enterprise investment into generative AI, 95% of organizations were getting zero measurable return on their P&L. OpenAI, the poster child of the boom, projected $12 billion in revenue for 2025 alongside an $8 billion operating loss, with losses expected to double to $17 billion in 2026 and again to $35 billion in 2027. A January 2026 Forbes analysis estimated total AI revenue at less than $50 billion against over a trillion dollars in investment. That gap between spending and returns is exactly what makes people reach for the word "bubble."

The dot-com parallel is the right one

Every time someone calls AI a bubble, the instinct is to compare it to the most recent one: crypto and NFTs. But that comparison falls apart quickly. NFTs were built almost entirely on speculation about future value. The underlying utility was thin, the use cases were circular (buy an NFT to sell it to someone else at a higher price), and when sentiment shifted, the entire market collapsed. There was no foundational technology beneath NFTs that the broader economy depended on. The bubble popped and left very little behind. AI is fundamentally different. Companies aren't just speculating on token prices. They're building data centers, training models that can write code and diagnose medical conditions, and deploying tools that are already changing how millions of people work. The technology is real, even if the current valuations aren't fully justified. The dot-com bubble is a much better lens. In the late 1990s, billions poured into internet companies with no revenue, no business model, and no path to profitability. Pets.com became the punchline. When the bubble burst in 2000, the Nasdaq dropped nearly 80% and trillions in market value evaporated. But here's the thing: the internet didn't go away. The infrastructure that got overbuilt during the bubble, the fiber optic cables, the data centers, the broadband networks, became the backbone of the modern economy. Amazon, Google, and countless other companies that survived the crash went on to become some of the most valuable businesses in history. The bubble was real. The technology was also real.

What's different this time

There are important differences between the AI boom and the dot-com era that are worth noting. First, the companies driving AI spending are profitable. Microsoft, Google, Meta, and Amazon are funding their AI infrastructure out of massive existing cash flows, not through speculative IPOs or fraudulent accounting. During the dot-com bubble, companies like WorldCom were inflating demand through over $11 billion in fraudulent accounting. The Sarbanes-Oxley Act of 2002 was a direct legislative response to that kind of behavior. Second, AI is already generating real revenue, even if it's not enough to justify current valuations. Microsoft's Azure cloud service grew 33% year-over-year in Q3 FY25, with AI contributing 16 percentage points of that growth. Google Cloud revenue grew 48% year-over-year to $17.7 billion in Q4 2025. These aren't Pets.com numbers. Third, the productivity gains are measurable at the individual level. Research from MIT Sloan found that generative AI can improve a highly skilled worker's performance by nearly 40% compared to workers who don't use it. The Dallas Fed has noted that most studies find AI significantly boosts productivity, particularly for less experienced workers. Stanford research tracking 200,000 U.S. households found that ChatGPT users complete digital tasks meaningfully faster. The challenge is that these individual productivity gains haven't yet translated into enterprise-level returns at scale. That's the gap the market is betting will close.

The bubble within the bubble

Capital Economics' John Higgins made an interesting distinction in early 2026: the AI stock bubble, meaning the run-up in share prices of AI-adjacent companies, had already burst. Nvidia's stock and others had corrected significantly from their peaks. But a deeper, more structural bubble was still building in the form of massive infrastructure commitments that may take years to pay off. This is the more concerning dynamic. When big tech free cash flow could drop up to 90% in 2026 as capital expenditure outpaces revenue growth, the question isn't whether the technology works. It's whether the timing of returns matches the timing of investment. Companies are building for a future that may arrive, but may arrive later than their balance sheets need it to. There's also the problem of circular financing. Some analysts have flagged that AI companies are essentially funding their own future revenue: cloud providers invest in AI infrastructure, startups use those cloud services, and the resulting "AI revenue" flows back to the same hyperscalers who made the investment. That's not fraud, but it does mean the headline revenue numbers can look more impressive than the underlying organic demand.

Why AI isn't going anywhere

Despite all of this, AI is not the kind of technology that disappears when the hype fades. It's a general-purpose technology, more like electricity or the internet than like NFTs or 3D TVs. The International Monetary Fund estimates AI will affect nearly 40% of jobs globally. Goldman Sachs projects it could increase global GDP by 7% over a decade. McKinsey puts the annual economic potential at $17.1 to $25.6 trillion. Even if these numbers are optimistic by half, the scale of impact is enormous. More practically, AI is already embedded in products that hundreds of millions of people use every day. It powers search results, code completion, email drafting, image generation, medical imaging analysis, and fraud detection. These aren't speculative use cases. They're shipping products. The Congressional Budget Office has noted that businesses implementing AI can be expected to become more productive than those that don't, and that widespread AI use would boost economic growth. The question isn't whether AI is useful. It's whether the current level of investment is proportionate to the near-term returns.

What happens when it corrects

If history is any guide, here's what a correction looks like: valuations come down, some overextended companies fail, capital becomes more disciplined, and the surviving players build durable businesses on top of the infrastructure that got overbuilt. The dot-com crash wiped out hundreds of companies but left behind the infrastructure for cloud computing, e-commerce, social media, and the modern internet. The fiber optic cables that were overbuilt in the late 1990s became the highways of the digital economy. Something similar is likely to happen with AI. The data centers being built today, the chips being designed, the models being trained, these won't disappear when market sentiment shifts. They'll become the foundation for whatever AI looks like in 2030 and beyond. The investors who got burned in the dot-com crash weren't wrong about the internet. They were wrong about the timing and the specific companies. The same will likely be true for AI. The technology is a necessity for humanity's next chapter. The current pricing might not be.

The bottom line

Is AI a bubble? In the financial sense, almost certainly. Valuations are stretched, spending is outpacing returns, and the market is pricing in a future that hasn't arrived yet. But is AI a fad that will disappear like NFTs? Not even close. The underlying technology is too useful, too widely adopted, and too deeply integrated into the global economy to go away. The bubble will correct. The technology will remain. The internet survived the dot-com bust. AI will survive this one too.