What Is AI and the AI Bubble?

What Is Artificial Intelligence?

Artificial intelligence is software that performs tasks normally requiring human cognition — understanding language, recognizing patterns, making predictions, and generating new content. The term was coined in 1956 at Dartmouth College, but the technology behind today's headlines is much narrower than science fiction suggests. When investors, analysts, and the media say “AI,” they are almost always referring to a specific subset called machine learning, and more specifically, a technique called deep learning powered by neural networks.

Here is the simplest way to think about it: traditional software follows explicit instructions written by humans. If you want a program to sort a list, you write step-by-step rules for sorting. Machine learning flips that model. Instead of writing rules, you feed the system enormous amounts of data and let it discover the rules itself. A spam filter trained on millions of emails learns to recognize junk mail without anyone explicitly programming the difference between “Nigerian prince” and a real message from your bank.

Deep learning pushes this further by stacking many layers of artificial neurons — mathematical functions loosely inspired by the human brain. Each layer extracts increasingly abstract features from the data. The first layer might detect edges in an image. The next might detect shapes. A deeper layer might recognize a face. This architecture, called a neural network, has existed in theory since the 1960s but was impractical until two things changed: we got enough data to train large networks, and we got enough computing power (primarily GPUs) to process it.

The breakthrough that defines the current era arrived in 2017, when a team of Google researchers published a paper introducing the Transformer architecture. Transformers solved a fundamental limitation of earlier neural networks: they could understand context across long sequences of text. This is the technology behind every modern large language model — GPT, Claude, Gemini, Llama. When ChatGPT launched in November 2022 and reached 100 million users in two months, it was the first time the general public experienced what deep learning could do in plain conversation. The response was not just enthusiasm. It was a financial event.

Narrow AI vs. AGI: A Critical Distinction

Every AI system in existence today — no matter how impressive — falls into the category of narrow AI. Narrow AI is exceptionally good at specific tasks: generating text, identifying tumors in medical images, recommending products, translating languages. It cannot, however, transfer its intelligence across domains. A model trained to write poetry cannot suddenly drive a car. It does not have general reasoning, consciousness, or the ability to set its own goals.

Artificial General Intelligence (AGI) is the theoretical threshold where a machine can match or exceed human capabilities across any cognitive task. No one has built AGI. Whether it is five years away or fifty is a matter of intense public disagreement among the people closest to the technology. OpenAI's Sam Altman has suggested it could arrive this decade. Yann LeCun, Chief AI Scientist at Meta, has said we are “nowhere near” true human-level intelligence and that current language models lack even basic understanding of the physical world.

This distinction matters enormously for investors. A significant portion of the valuation premium in AI-related stocks is implicitly pricing in a trajectory toward AGI — the idea that today's language models are early versions of systems that will eventually automate vast categories of knowledge work. If AGI is close, current valuations may be conservative. If it is decades away, or if it requires fundamentally different approaches that today's leaders may not dominate, then a large part of the market's optimism is built on an assumption, not a fact.

What Makes a Bubble?

A financial bubble is not simply a market where prices have risen. Prices rise in every healthy bull market, and sometimes those rises are justified by fundamentals. A bubble is a specific condition: asset prices detach from underlying value because investors are buying not for the cash flows the asset will produce, but for the expectation that someone else will pay more tomorrow. The economist Charles Kindleberger, in his classic work Manias, Panics, and Crashes, identified a pattern that has repeated across centuries of financial history — from the Dutch tulip mania of 1637 to the housing bubble of 2008.

The key insight is that bubbles always begin with something real. The internet did change everything. Railways did transform the economy. The technology is not the bubble. The bubble is what happens when capital — driven by fear of missing out, competitive pressure, and the simple reality that predicting the future is hard — piles in faster than the fundamentals can justify. Every bubble in history has been built on a genuine innovation. That is what makes them so dangerous: the believers are never entirely wrong about the technology. They are wrong about the price.

So, What Is the AI Bubble?

The “AI bubble” is not a single event or a single asset. It is a constellation of behaviors across public and private markets where prices and capital flows have moved ahead of what the underlying AI economics can currently justify. It does not mean AI is fake. It means the market is pricing in a future that may or may not arrive on the timeline investors are assuming.

The most visible manifestation is in semiconductor stocks, where Nvidia — the company designing the chips that power nearly every major AI model — saw its market capitalization rise from roughly $300 billion in late 2022 to over $3 trillion by 2024. That is a 10x increase in under two years, driven by unprecedented demand for its data center GPUs. Nvidia's revenue did grow dramatically, from $27 billion to $130 billion. But a $3 trillion valuation implies that this growth continues, without competitive erosion, for years to come.

Beyond the public markets, the private side is where bubble behavior is most pronounced. In 2024 alone, venture capital poured over $80 billion into AI startups. Companies founded months earlier, with small teams and no revenue, raised hundreds of millions at billion-dollar valuations. Anthropic, OpenAI's primary competitor, was valued at over $40 billion. xAI, Elon Musk's AI venture, raised $6 billion at a $24 billion valuation before having a broadly released consumer product. When capital chases pre-revenue companies at these prices, it is not investing in fundamentals. It is investing in narrative — the belief that whoever wins the AI race will capture a market so large that today's entry price barely matters.

The third pillar of the AI bubble is infrastructure spending. The four largest US hyperscalers — Microsoft, Amazon, Google, and Meta — collectively spent over $230 billion on capital expenditures in 2024, with the majority directed toward AI infrastructure. For 2025, that figure is projected to exceed $320 billion. This is the largest concentrated industrial investment since the telecom fiber build-out of the late 1990s, and it raises the same uncomfortable question: if the software revenue does not arrive fast enough, the hardware depreciates while waiting. Unlike fiber optic cable, which sat in the ground for 20 years until demand arrived, GPUs have a useful life of 3 to 5 years before next-generation chips make them economically obsolete.

The Five Signals That Matter

Evaluating whether we are in an AI bubble is not a matter of gut feeling. It requires tracking specific, measurable signals that — when read together — form a picture of how much risk the market is pricing in. This is the framework behind the AI Stock Bubble Index.

No single signal tells the whole story. Valuations can stay elevated for years if revenue grows into them. Capex can be justified if software monetization accelerates. Media hype can be a lagging indicator rather than a timing signal. The value is in reading them together — and watching how they move relative to each other. That is what the index does, updating automatically as new data and news arrive.

What Should You Actually Do?

If you take one thing from this article, let it be this: recognizing a bubble is not the same as knowing when it will pop. The market can stay irrational far longer than you can stay solvent betting against it. John Maynard Keynes said that nearly a century ago, and it remains the most expensive lesson in investing.

For most individual investors, the rational response to elevated bubble risk is not to short the market or move to cash. It is to trim exposure, avoid concentration, and resist the urge to chase returns at peak valuations. The investors who survived the dot-com crash were not the ones who timed the top. They were the ones who did not put their life savings into Pets.com at the peak of euphoria.

Use the index and its underlying signals the way a sailor uses a barometer — not as a prophecy, but as an instrument. When pressure falls, prepare for rough weather. When it rises, conditions may be safer. The score moves as new evidence arrives. So should your awareness of risk.

You cannot control the market. But you can control how much risk you take on. That is the point of tracking this. Not to call the top. To know when you are standing at the edge.