POP Goes the AI Bubble

Photo via New Scientist 

***

It feels like AI has been integrated into absolutely everything. When you shop on Amazon, when you try to listen to music on Spotify, when you go grocery shopping—even as I write this article, Google is trying to persuade me to use their new integrated AI features. I can assure you that this article was written without utilizing generative AI. 

But even as these AI models, particularly LLMs (language learning models), are implemented, the companies pushing for their use do not appear to be making a profit. OpenAI forecasted that it would burn through over $140 billion by 2029 before turning a profit. 

This raises the question: how exactly are these companies able to pay for the development of their products when they aren’t even making money? To answer this, one only has to look at past examples of goods or services whose values were artificially inflated by overexcited investors. 

In the 1600s, a botanist planted tulips from Constantinople in his home garden in Holland. After his neighbors (rudely) stole the tulip bulbs and started selling them, the wealthy began buying them up, and they came to be considered a luxury good. This phenomenon quickly grew into what is now known as Tulip Mania, and it reached a point where tulip bulbs were exchanged for increasingly valuable items, such as homes or land. This era ended in a crash when the price of the bulbs became unsustainable, making it the first example in history of an economic bubble. 

A more recent, and perhaps relevant, example is that of the dot-com bubble in the late 1990s and early 2000s. Similar to the Tulip Mania of the 1600s, the value of this new product was inflated by speculative investing rather than by its actual worth. Investors were pouring money into internet startups with the hope of turning a massive profit, until they began to panic when they had sunk all their money into these companies with no yield. Thus, by the early 2000s, many public dot-com companies failed, and the dot-com bubble burst.

Both of these events illustrate what a bubble actually is: a rapid escalation in asset prices, often caused by speculative behavior, followed by a sharp decline in value—essentially prices inflating and popping, like a bubble! 

This leads to the next question: Is there an AI bubble? While bubbles are historically hard to define while they are happening, many industry leaders have acknowledged that there is a level of overinvestment and overexcitement about AI, including Sam Altman, the CEO of OpenAI. In fact, he compared the outpour of investor support to the aforementioned dot-com bubble of the early aughts. 

Despite this acknowledgment and the brief dip in stocks that followed, AI companies and investors continue to pour hundreds of billions into data centers, infrastructure, and development. In fact, OpenAI alone has pledged half a trillion dollars to build AI data centers, and Anthropic, Google, and Perplexity are racing to follow suit. 

So these AI companies appear to be unaffected by these claims of a bubble – or at least their CEOs do. But many economists are still claiming that a potential “burst” or crash still lies ahead for this foundling industry. Perhaps identifying a bubble in real time, something that has not historically been done, is ultimately leading to its burst due to the economic anxiety it’s creating. 

It’s clear that the advent of artificial intelligence is raising more questions than it claims to answer. The one I believe we should focus on is the question that affects every company, industry, and society when a titan falls: who or what will fill the power vacuum that AI will inevitably leave?

***

This article was edited by Abigail D’Angelo.