The Generative AI Bubble: A Prelude to Another Financial Bubble?
The generative AI bubble is drawing alarming parallels to the infamous dot-com bubble of the late 1990s, but it may induce a different kind of economic fallout. While the number of generative AI users is rapidly increasing, echoing the growth of internet users in the 90s, comparisons between the two eras are superficial at best. Some analysts and economists at the Federal Reserve Bank of St. Louis have likened the growth of generative AI to that of the internet, yet the circumstances surrounding their usage are fundamentally different.
The Cost of Entry: Then and Now
To understand the disparity, let’s consider the technological landscape of the late 1990s. To gain internet access, users needed to invest in expensive hardware and service subscriptions. For instance, a Compaq ProSignia Desktop 330 set consumers back $2,699 in 1999 (approximately $5,101 in today’s dollars), while a digital subscriber line service cost around $59.95 a month (equivalent to $113 today).
In stark contrast, today’s access to generative AI services, such as ChatGPT, is essentially free for users. Most people already own a computer and are paying for internet access. Limited-use classifications from providers like OpenAI make it even easier and cheaper for users to experiment with generative AI. As a result, the cost barrier that previously dictated user expectations and payoff is virtually nonexistent today.
Utility vs. Entertainment: A Different Kind of Challenge
The simplistic comparison of generative AI usage with the earlier internet boom does a disservice to the serious questions surrounding the economic value these technologies provide. Just because a service is widely available and free, it doesn’t automatically imply it is useful or beneficial. Much like social media that thrives on user engagement but ultimately distracts from productivity, generative AI systems can similarly prove to be an addictive form of entertainment that keeps users occupied without offering significant economic returns.
Economists speculating on generative AI’s capabilities for boosting labor productivity often examine user engagement. However, one must wonder: does intensive use of platforms like Facebook, Instagram, or TikTok actually enhance productivity? Likely not, as indicated by various studies suggesting a reduction in overall work efficiency. Consequently, the economic gains from generative AI cannot be measured on an assumption of burgeoning user base alone—they depend on tangible economic output.
High Costs Amidst Limited Payoffs
The reality of the costs associated with generative AI is stark. Algorithms that power these systems necessitate significant investment in state-of-the-art technology, training, and resources. Reports suggest that training these advanced models costs up to $100 million and relies on the latest and costliest chips manufactured by companies such as Nvidia, AMD, and Intel.
Furthermore, successful deployment and service creation remain prohibitively expensive. A recent feature by the Wall Street Journal highlighted the considerable costs involved in creating a generative AI bot. The investment in her bot project could have funded an extravagant trip to Bora Bora. This raises the crucial question: how will generative AI translate its enormous upfront costs into an economic model that justifies profit?
The Stunted Revenue Landscape
Evaluating revenue generation in the context of the dot-com bubble paints an unsettling picture. In the year 2000, the internetgenerated revenues exceeding $1.5 trillion (adjusted for inflation) by combing hardware sales, internet service revenues, and e-commerce. In comparison, generative AI currently contributes a meager slice—under $10 billion—to the global economy. This variance leads analysts like Sequoia’s David Cahn to assert that approximately $600 billion in annual revenue would be required from generative AI to validate current investment metrics, a staggering figure up to 100 times greater than the present annual earnings of major players like OpenAI and Google.
Despite OpenAI forecasting losses in the region of $5 billion against just $3.7 billion in revenue this year, it was recently able to secure a whopping $6.6 billion from investors, thereby hiking its valuation to $157 billion. This scenario draws ominous comparisons with the dot-com bubble, with economists like Goldman Sachs’ Jim Covello and Citadel’s Ken Griffin echoing warnings of potential fallout that could resemble, or even surpass, previous market downturns.
Conclusion: The Looming AI Reckoning
As we navigate through the evolving landscape of generative AI, it crucial to distinguish between user engagement and actual economic value. Since the relative low cost of engaging with generative AI might lure many, the stark reality remains that long-term profitability and productivity outcomes are as yet unproven. If this bubble bursts, the repercussions could echo through the financial markets, drawing lessons from history that have implications for investors and technologists alike.