The AI Bubble

Right now, we are in an AI bubble. The tech giants, Microsoft, Amazon, and Alphabet, better known as Google, along with their AI partners, OpenAI and Anthropic, and data center operators are racing to gain first-mover advantages by being the best, fastest, and largest providers of AI and everything related to it.

The over-building of data centers is simply a byproduct of this frenzy. Data center construction across the country is temporarily driving the national economy by employing tens of thousands of construction workers and equipment manufacturing personnel, but this rapid pace is ultimately unsustainable.

In spite of a recent pull-back due to renewed hostilities in Iran or the strong jobs report, depending on who you’re listening to, the stock market is at an all-time high, and AI development is fueling it.

I won’t go into the circular money flow that is propelling the market to ever greater heights, but suffice to say, debt from one company is profit for the next, and business or consumer usage is not the source of AI profits. This will not end well.

As the cliché goes, history doesn’t repeat, but it rhymes. I’ll compare AI, data centers, and everything associated with them to innovations and bubbles from the past: spreadsheet software, the dot-com bubble of the early 2000s, and the rise and fall of Global Crossing.

When VisiCalc, then Lotus 1-2-3, and eventually Microsoft Excel entered the business landscape, there were concerns that the manual work accountants did would no longer be necessary, leading to a massive drop in employment. In reality, financial professionals were able to expand their insights; what couldn’t be done before easily became routine.

Interpreting financial data and acting on it became the required skill, and the conventional accounting task of literally putting pen to paper in a neat and orderly fashion died. Accountants who didn’t embrace Excel became unemployed accountants.

AI will similarly replace current routine tasks, but it will simultaneously open the need for whole new skillsets. At least for now AI won’t replace people directly, but people who use AI will replace those who don’t.

When a large tech company lays off software engineers “due to AI,” it really means they are freeing up cash to spend more on AI investment and not because they could replace the workers with AI.

In spite of being a people manager in supply chain management, I frequently write software code that helps me and others at my employer be more productive, and AI assistance with syntax and even full blocks of code has been invaluable. AI also helps me clarify thoughts and draw meaningful, actionable conclusions more quickly than I can on my own.

I treat AI like it’s a capable intern- smart, well-meaning, and wants to please, but in need of direction and coaching.

AI definitely makes mistakes, but it is getting better all the time.

The current AI landscape seems eerily similar to the dot-com boom and bust, which produced many flops. Anyone remember the Webvan grocery delivery business? It took a global pandemic two decades later for people to start buying groceries online for home delivery at any scale.

Yet, the dot-com frenzy also gave us Amazon. Whole new businesses were created while others declined as part of Schumpeterian creative destruction- taking investor money with them.

The big difference between now and the dot-com era is that back then, money was thrown at companies that were just ideas on paper- no structure, no people, much less profits and cash flow. Conversely, everyone can see the power and potential of AI already; there are no castles in the sky here. Big blocks of code that work and clarification of ideas concisely, even if they are in AI-ese, are proof enough for me.

Still, the overbuilding of data centers mirrors the dark side of the dot-com era. Today’s AI-adjacent infrastructure companies are similar to Global Crossing, a telecom founded during the dot-com era that went out of business partly due to “accounting irregularities,” but also because they installed much more undersea fiber optic cable capacity than demand could fill.

Investor dollars were exchanged for unused cable that sat at the bottom of the ocean under the philosophy of “if you build it, they will come.” Eventually, there will be a glut of data centers and construction will stop, until demand catches up with supply.

Currently, AI token cost, which is how AI prompts are broken down and paid for, is being heavily subsidized by big investor money flowing into the sector. As a result, the cost to end-users is very inexpensive relative to the work it can do. However, token costs must go up in the future to pay back those investors.

If there is a financial silver lining in the difficult national pushback against data center construction, it may be to reduce excess data center capacity and the AI bubble to some extent.

So the question remains of what we can do about the AI bubble.

The answer is not much, which is the nature of bubbles.

My thoughts are twofold and relate to my personal and professional life and financial situation.

While AI is basically free, I use it all the time. A year ago, I didn’t use AI at all. Now I don’t know how I’d get anything done or at least done as quickly as I do now. If you’re no longer gainfully employed or working, ask AI all the questions you can.

On the financial side, I’m not a financial advisor and definitely not your financial advisor. Having stated that, it will be difficult to predict the top and bottom of the stock market as the boom continues and the bust eventually happens so I’ll simply stay in the market. If you have money invested but won’t need it for 40 years, you’ll see many more market swings than just the AI bubble.

I lived through the dot-com era and the Great Recession, and I simply didn’t look at my retirement account statements. The market came back after both. I figured if I just kept saving, things would work out, and they did.

But now that I’m not too many years away from retirement so I’m looking at ways to principally protect some savings so that even if the stock market is down, I can still retire. The savings will earn less, but I’ll sleep better at night, and that is something money can’t buy.

Market bubbles come and go, and the current AI bubble resembles past ones, notably the dot-com era, which had the downside of heavy investor losses and bankrupt companies in the near-term but gave us businesses like Amazon and the internet we rely on every day.

AI is going to do the same thing, and it will transform how work gets done, even if there is some difficulty along the way. There are things we can do to protect ourselves- ride out the market drop and have protected savings if we need them.