Reminiscences of a Tech Stock Analyst
Author – Brian Piccioni
Mr. Piccioni was involved in the capital markets for over 25 years. Most recently, he was Senior Technology Strategist at BCA Research Inc., the largest provider of subscription research to the investment community. Before BCA, Mr. Piccioni was an award-winning technology equity research analyst with BMO Capital Markets. Prior to his career in capital markets, he spent 12 years as a digital electronics designer and maintains an active interest in hardware and software development. Mr. Piccioni holds the Chartered Financial Analyst designation, has completed the Institute of Corporate Directors, Directors Education Program, and received an MBA from Concordia University and a BSc from McGill University.
My career in capital markets spanned both the Dot Com Bubble of 1999/2000 and the Global Financial Crisis of 2008/2009. As a technology analyst, I admit to having been a far more informed observer of the Dot Com Bubble, and perhaps that is why the Financial Crisis frightened me as much as it did. Although I rarely have much to do with capital markets nowadays, when asked by my friends at Focus Wealth Management how I feel about the current Artificial Intelligence whirlwind, I replied that it is both similar to and different from the Dot Com Bubble, but with the potential to cause as much damage to stock markets as the Financial Crisis.
For the record, I believe AI will be as disruptive a technology as the Internet. Although some claim it will cause another industrial revolution, I do not see that as likely, nor do I see the risk of ‘artificial super intelligence’ or ‘artificial general intelligence’ as significant. The reasoning for this is complex and beyond the scope of this article, however it is important to realize that there is a big difference between appearing intelligent (as AI systems do) and actually being intelligent. AI algorithms – as useful as they are – do not function in a manner even roughly comparable to a human brain and therefore, there is little reason they will match its capabilities anytime soon. That does not preclude it from occurring in the future, as researchers develop novel approaches, but I do not believe it is possible using current methodologies, regardless of advances in related technologies.
While AI will result in job losses, there will be offsetting opportunities. I am, however, deeply concerned that unless educators do something, and fast, we will have a cohort of essentially unemployable young adults due to the use of AI at critical times during their education. Another major concern is the ease with which disinformation can be created, in particular high-quality fabricated video.
Superficially, the AI and Dot Com Bubbles have much in common, in particular with respect to company valuations – though I would argue that tech-related valuations became decoupled from reality many years ago, with cryptocurrencies and Tesla being obvious examples. The social media era, combined with lax regulation, has made it a happy time for promoters of speculations of all sorts, not just stock prices.
An important similarity between the two technologies is the need for an infrastructure roll-out requiring many billions of dollars of investment and many years of work. During the Dot Com era, companies keen to exploit the opportunity raised a great deal of capital and put it to work building the modern Internet. This led to an unprecedented increase in demand for everything from data communications equipment to excavation services to bury the fiber optics in the ground. Networking equipment providers such as Cisco Systems and Nortel Networks found themselves in the enviable position of massive revenue increases coupled with high margins, as their customers were so focused on expansion that they were largely indifferent to pricing. In turn, companies supplying network equipment, such as PMC Sierra (semiconductors) and JDS Uniphase (optical components), also saw staggering growth and expanding margins. Their suppliers, in turn, enjoyed the same windfalls. It was turtles all the way down.
Rapidly rising revenues and expanding margins are fantastic for company valuations, and many companies took advantage of this to raise capital, allowing them to spend even more freely. When a new technology emerges, there is a shortage of relevant talent – after all, how many people understood IP networking prior to the Dot Com era? Many firms were led by people with little relevant expertise, and a lot of spending was driven by capital markets considerations rather than the expectation of a favourable return on investment (ROI).
The nature of capitalism is that rising, high-margin demand leads to increased supply, and in time, lower margins. However, the nature of capital markets is that eventually investors lose patience with good stories that have no actual economic returns. Some Dot Com companies found they could no longer raise money and, since they were not structured for profit, cut spending, leading to a decline in orders to their suppliers. This led investors to further retreat from Internet-related stocks, and suddenly the era of free money was over, which led to a complete collapse in orders. By the time the dust settled, many companies were gone, and even some profitable ones were trading at net cash per share.
Of course, there is no reason to assume the same thing will happen to AI companies. Many of the leading firms, such as Google and Microsoft, are prodigious cash generators and led by highly competent management. That doesn’t mean they are making all the right choices with respect to AI-related investment, as it is impossible to predict how the technology will evolve, let alone what business models will be the most profitable. This is important given the scale of spending and how few companies are involved. The Economist recently reported that a mere handful of companies are projected to spend over $3 trillion on AI infrastructure over the next three years.
I suspect the investment in Internet infrastructure was of a similar scale, but that time frame spanned decades and involved thousands of companies. This is an important distinction: it took a lot of cancelled orders from small Internet companies before it affected the supply chain, since there were plenty of startups ready to take their place. The small number of AI players means that any slowdown would hit the supply chain hard, and this would have an impact on capital markets, further reducing capital available for AI infrastructure investment.
Whatever you think about the future of AI, investors buying AI infrastructure stocks believe that much more than $1T/year will be spent over a much longer period. For example, NVIDIA, which has a dominant position in data-centre computing technology, is trading at a Price/Earnings ratio of approximately 54, which implies high growth expectations for a company with over $165B in revenue over the past four quarters.
Infrastructure is usually funded by cash flow and/or debt, and there isn’t enough cash flow in any of the AI companies to fund such large investments. Today there may be an appetite for a few hundred billion dollars in AI infrastructure-related debt, but the question is how big that appetite is and how long it will last. Bondholders are generally risk-averse, and it wouldn’t take much to interrupt those capital flows. If that happens, infrastructure providers will report slowing orders, which will impact their share prices, as well as those of their customers, and will further impede AI infrastructure funding. As we saw during the Dot Com Bubble, virtuous cycles can become pernicious ones very quickly.
Were that to happen, the highly concentrated nature of US large-cap indexes and the popularity of index Exchange Traded Funds (ETFs) mean a lot of wealth could disappear from the economy very quickly. These losses would not be confined to AI related firms, or to the US, and likely extend to debt markets, utility capital spending plans, and so on. The Financial Crisis, as devastating as it was, saw $2 trillion erased from the global economy. And while not exactly comparable, the market capitalization of just the “Magnificent 7” stocks currently dominating the S&P 500 is about 10 times that figure.
At present, the normally broadly diversified S&P500 has reached levels of concentration that are unprecedented – the result of very strong performance over the last few years of a small number of companies. Concentration in a portfolio works very well on the way up, but equally badly on the way down. While it may be frustrating to lag the index over the near-term, long-term investors should understand the potential risks lurking in their portfolio today.