Let me start by acknowledging that I am not an AI geek, nor a sophisticated Wall Street analyst, and I have not gone very deep into the ocean of financial machinations underlying the AI boom. But, with just a surface view, something looks rotten. I vividly remember the 2008 financial crisis, and based upon what little I understood of the underlying causes of that crisis, the AI boom looks really similar. I am also old enough to have been in the tech industry during the .com bubble in 2000, and once again, I see many similarities in the AI boom. Putting it all together, I am experiencing conflicting emotions of being frightened about the potential for financial catastrophe, while at the same time, excited for what the AI future will bring.
Let’s start with the positives. The .com era drove fantastic positive changes for everyone. Google catalogued global information and made it easily accessible to all. Amazon created the worlds largest retailer and made goods available overnight. Email became ubiquitous and made it possible to correspond with everyone at nearly no cost. Social media united the world and created communities for everyone. Against the backdrop of the .com era, cloud computing and pocket computers (AKA phones) became commonplace. Great things happened during that era, but the dark underbelly was lined with greed and some really bad ideas that eventually led to the .com bubble bursting with dramatic negative financial impact. Everyone could see the huge potential of the internet and all the innovative applications that were coming to market. There was such a rush to invest and hook your wagon to the next rising star that otherwise intelligent people made really stupid investments. At the peak, we were investing in ‘eyeballs.’ Even if a company lost money attracting every new visitor to their site, the business looked investible if it consistently grew the number of visitors. Losses did not matter, only eyeballs. The race was on to rise to the top and dominate new business segments, and capital was unlimited in the quest for scale, until it wasn’t.
In the mortgage crisis, financial engineers invented ways to monetize stacking risk on top of risk on top of risk. Low interest rates fueled a massive boom in housing, and predatory lenders created sub-prime mortgages that trapped home buyers with escalating rates they could not afford. The financial industry bundled individual mortgages into investment products and sold them globally. Credit rating agencies mistakenly gave top-tier "safe" ratings to bonds backed by risky subprime loans, and insurers created Credit Default Swaps to mimic insurance contracts to backstop the loan packages. It all collapsed when the original loans defaulted and we realized that it was a house of cards layering risk on top of risk on top of an initial unfounded proposition that borrowers could actually afford the homes they were buying.
I titled this post ‘Ouroboros,’ which is an ancient symbol depicting a snake or dragon eating its own tail. The current AI-driven data center boom looks like ouroboros to me. Great things will come from the rise of AI, and there is no doubt that it will generate real economic value. It is no wonder that like the .com boom, everyone wants to jump on board. There is also little doubt that not every entrant into the space will succeed. In the .com era, if you added ‘.com’ to your business name, your value increased and investors came knocking. Similarly, today if you add AI to your platform or your URL, your appeal goes up, or conversely, if you do not claim AI, you are not investable. AI has become the new ‘eyeballs’ for investors. Companies are being driven to incorporate AI into their applications with little care for the impact on margins, which ultimately will come back to bite investors.
The bigger ouroboros issue is the financial engineering taking place on the platform side of the AI explosion. Think of all those application vendors like the home buyers. They are driving demand for AI processing, but may not be in a position to pay the bill as the cost expands. However, the AI platform vendors see infinite growth ahead. The billions of dollars being invested is staggering. But, it seems we are double and triple counting, just as in the mortgage crisis. It starts with the AI developers —Open AI, Anthropic, Microsoft, Google, etc. They project massive future revenues, and are willing to sign long-term contracts committing to pay the data centers for the computing power to serve their projected customer demand. This is the first count, committing future projected revenues to secure computing power. Next, the companies that agree to create the data centers to provide the future computing capacity (Oracle, CoreWeave, etc.) take their signed and committed revenue streams to the financial markets as collateral to borrow the money to buy the processors for the data centers. They also commit to long-term contracts for the leases and all the infrastructure for the data centers. This is the second and third count of the future revenue the AI developers pledged.
The house of cards is dependent upon the AI developers actually generating the revenues they project, and that is dependent upon the AI-driven applications succeeding (remember the .coms), and all of the other AI consumers growing their demand dramatically. Similar to the credit default swaps, the market needs a backstop to insure the risk. Nvidia to the rescue. Nvidia makes the processors that the industry needs to do the magical computing behind AI. All of those data centers being built to service the AI demand are creating a massive demand for Nvidia processors, which means Nvidia has a massive backlog and huge future revenue projections. Nvidia needs the AI developers to succeed, and they need the data centers to keep growing and opening new centers. Here is where the snake eats its tail. Nvidia has leveraged its huge backlog and bright financial future to invest in or backstop the AI developers, and Nvidia has invested in and made debt and lease agreements with the data center builders. In other words, Nvidia is committing its financial future to support the companies and contracts upon which that future is predicated. If the AI developers do not grow as forecast, they will not fulfill their obligations to the data center builders, who will not fulfill their obligations to Nvidia, which will not be able to fulfill its obligations to support the AI developers. Hence the snake will have eaten its tail.
I have grossly over simplified this financial morass, and I very well may have it all wrong, but something smells fishy when all of these players cross-invest in each other and make giant long-term commitments all predicated on the same forecasted expansion of AI revenues. The technology landscape is changing at a record pace, and if there is a blip in the AI path, it seems like the house of cards could come tumbling down in a spectacular manner, just like the financial crisis and the .com bubble bursting. What if some bright engineers invents a better chip or an AI model that dramatically undercuts the revenue potential of the current players? While I am excited by the innovations and the potential that the AI revolution is creating, my issue is with the financial engineering and greed that seems to be creating an ugly underbelly for the industry.