It was an eventual week. The government shutdown was brought to a close. On Wall Street it was a week of rotation with money rotating heavily out of many tech names into areas of the market that couldn’t catch a bid for months or even years. Health care stocks, for instance, had their best week in recent memory. The rotation had less to do with improved prospects for health care but rather increasing concern that the AI euphoria was overdone. Investors sought safer havens. We explored that issue a bit last week. Today, I want to dig a little deeper.
Before talking about today’s world, I want to step back almost three decades as the Internet was becoming mainstream and Wall Street embraced it in a very similar manner to the AI craze today. Over the ensuing 30 years, few can argue that the Internet didn’t live up to its economic promise back then. Look at today’s Mag 7. Where would Amazon#, Alphabet#, Microsoft# and Meta Platforms# be today without an Internet? And don’t forget Apple, whose iPhones are used for apps across the Internet more than to make voice calls.
By 2000, however, when the euphoria of the time turned to skepticism almost overnight, we learned a lot. Pioneers rarely became the iconic leaders down the road. Of the pre-2000 era standout companies, only Amazon evolved into an Internet titan. While some of the biggest companies were late to the game, their balance sheets and ability to build an integrated ecosystem allowed them to dominate eventually. Standalone spreadsheet or word processing purveyors couldn’t compete with Microsoft that could package a whole suite of well-integrated products. The stocks of Microsoft and Amazon took big hits when the bubble burst on Wall Street, but they came back as their respective business models held together. Amazon and Microsoft had one other major factor in their favor. Both financed their growth with equity and cash flow. Younger companies simply didn’t have the resources to effectively compete.
Move forward to today. AI, like the Internet, is almost certain to change our lives over the next 30 years. It may become an even bigger economic juggernaut than the Internet. But like the Internet, AI will take some time to get traction, to reach profitable economic scale. All the buzz about humanoid robots, the eradication of call centers and humans doing coding are almost certain to be true. But that isn’t a 2026 event. When will McDonald’s replace the human taking your orders at the drive-thru window? First the company has to build or buy software. Then it will have to test it, roll it out into a few units, expand to a wider swath before finally rolling it out systemwide. That will likely take at least a few years.
When you ask a corporation to invest hundreds of millions or even billions of dollars, the first question asked is what is the expected return on my investment and when will it be achieved? When choosing a vendor, large corporations focus on the staying power of the supplier, a question that can’t be answered with confidence often when an industry is still in its pioneering phase. Using AI for coding is a far different decision that using AI-backed computers to take orders at a takeout window, where a frustrating experience can risk losing customers.
But given the size of the opportunity, it’s no wonder venture capital and Wall Street investors are chasing the dream. By now, most of us have tried apps like ChatGPT and, generally, been satisfied with the results. ChatGPT isn’t alone. A similar startup, Anthropic, aims at the enterprise more than the consumer with its app, Claude. Elon Musk offers Grok. Alphabet has Gemini, which it now integrates effectively with its search engine. Don’t ignore the advantage of integration. Others are not inventing large language models themselves but plan on integrating some or all of the models onto their platforms. Think of Microsoft and Apple in that regard. Microsoft has also chosen to incorporate artificial intelligence features within its Office suite of products using Co-Pilot as its platform. In typical Microsoft fashion, it provided the feature free for a time and now is raising prices for its Office suite, thus actually producing revenues and a return, versus ChatGPT whose user base of hundreds of millions is dominated today by enthusiasts who don’t pay OpenAI a dime.
The 30,000 foot view of the market over the last two weeks is that the tech stocks got whacked as investors fled to safer havens. While it’s true that they all went down, they didn’t sink in unison. Microsoft, Amazon, Alphabet and Apple are all down less than 10% from their highs. Apple is actually down less than 2%. But Meta Platforms is down almost 25%. Oracle is down almost a third. And most of the hot IPOs of just a couple of months ago are down 50% or more from their recent highs. Some are back near their initial offering price.
There are two overriding factors in play to explain this disparity: debt and a mismatch between expectations and reality. I’ll start with the latter. Correctly implemented, present value calculations do not ignore losses in the early years. As for projections of cash flows 3-5 years out or longer, it’s easy to argue they can be categorized as “garbage in, garbage out”. So many companies raise seed capital and early round financing based on factors like the size of their addressable markets and the prior business success of their founders. Ignoring early year losses when trying to do a present value analysis may yield a distorted valuation, one that survives in euphoria but disintegrates as reality paints a more sober picture.
But the more insidious concern is debt. OpenAI, the company that built ChatGPT wants to spend hundreds of billions of dollars with the vision that its platform will not only support the ultimate query engine of tomorrow but also an almost unimaginable dream list of applications that will result in a spectacular increase in human and corporate productivity. Both investors and corporations have been willing to pay whatever for the opportunity to be an OpenAI investor. Today, the company has an implied market cap of close to half a trillion dollars based on its most recent financing round. To put that into perspective, such a valuation would put it within the top 15 companies in the S&P 500. And that’s for a company today with a revenue run rate of about $20 billion, losing money and likely having a massive negative cash flow for several more years.
The money being raised is a combination of debt and equity. Much of that equity comes from companies that expect to gain significant business from OpenAI. In several cases, OpenAI has signed huge contracts with platform providers like Microsoft and Oracle# stretching out many years. However, as noted earlier, when it comes to 5-year forecasts, it’s largely garbage in, garbage out. No one knows the actual size of the market in five years, nor does one know which competitors will be the leaders and which will lag behind. History repeatedly tells us that there will be many more failures than winners. ChatGPT is a name consumers know well but it isn’t nearly as dominant in the corporate enterprise world.
Companies without cash flows have a voracious appetite for capital. Some comes from equity investments, some from present and future suppliers hoping to reap revenues in return, and some comes from debt. Oracle may be the poster child when it comes to debt. It has recently raised a massive amount of debt to support its own infrastructure growth to meet future promised contractual demand. I haven’t seen those contracts and I doubt other investors have seen all the details. But if the promises of enterprises like Stargate and OpenAI, for whatever reason don’t materialize or do so over an elongated timeline, then one has to question whether Oracle can achieve its growth targets of 2028-2030. When Oracle, at its last earnings call, extolled all these opportunities and promised revenues, its stock rapidly rose 50% almost overnight. But as more sober eyes noted that there were risks, that Oracle could end up building more capacity than what would be needed, the stock gave back virtually all its post-earnings gains. Nothing I just said should be construed as a negative outlook for Oracle. If Stargate and OpenAI achieve their goals, Oracle will be a big winner. But (1) it’s too early to separate winners from losers, (2) debt magnifies both the future risks and rewards, and (3) contracts for future capacity aren’t worth much if the contracted demand fails to materialize. Oracle is making a big bet. It might win and it might not. Two months ago, investors only saw the upside. Last week, they only saw the downside. It might take a few years to determine the true value of its enormous opportunities. The risk/reward balance reflected in its stock certainly makes more sense today than a month ago.
This week, Nvidia# reports earnings. That will be the market’s major focus. Current demand appears strong and its next generation of chips has everyone excited. But its China business is now effectively zero, a combination of political constraints imposed upon it both by China and the U.S. Which country will win the ultimate AI race? The answer almost certainly isn’t one country or the other but rather there will be winners residing here and there. We certainly had the early lead but Chinese companies are catching up. China is growing its electricity production in grid support much faster than us. Its base of STEM graduates is growing at a much faster pace than us. Looking down the road several years, does anyone believe there will be two totally segregated AI ecosystems, one Chinese-based and one American-based? China has already demonstrated that it can work around the absence of Nvidia chips and make several excellent large language models. Political decisions can create short-term roadblocks that will only serve to slow overall progress. Necessity is the mother of invention. Political barriers will be overcome over time. As for Nvidia’s earnings report itself, demand remains robust and will continue that way as the world continues to build out capacity for future needs. When might capacity exceed supply? That’s a big question. With both demand and supply growing rapidly, it is likely that for some interim period there will be temporary imbalances creating some excesses. Any short-term excess capacity will almost certainly lead to a meaningful correction in stock prices should it occur. But just as the burst Internet bubble gave way to the emergence of applications like Facebook and Google, any market purge will seed a new generation of leaders. Whether it is Nvidia or others, the best semiconductors will be the heartbeat of what lies ahead. Until we get there, new winners and losers will emerge. Companies with huge ecosystems and wide moats like Microsoft and Apple should remain major players. But as the market has shown us over the last few weeks, skepticism over the fortunes of big names like Oracle and Meta Platforms remind us that we are traveling on a highway with a fair number of potholes. Few thought in 1999 that Yahoo and AOL would be afterthoughts a decade later. As we have learned through the experiences of companies like Apple and Microsoft, the ultimate winners will build ecosystems that attract and retain customers. There are two differences between 1999 and 2025. First, the sizes of the investments are enormous. Startups can win within vertical markets but won’t dominate overall. Second, in 1999 capacity was being built in the expectation of future demand. Today, the demand for capacity is outstripping supply. Not only do we lack computing capacity in some cases, but we lack electrical capacity as well. Building new chips and computers that use dramatically less power will be key to fulfilling future promise.
The future looks exciting. But getting there won’t be easy. Many will try and fail. Some of today’s largest companies will fail or at least come up well short of current expectations. That reality hit Wall Street last week. It isn’t a time to panic. But it highlighted the realization that there will be more losers than winners over time. Few pioneers climb the mountain to the summit. But those that do yield big rewards for their supporters. If you are riding the wrong horse in the race, get off and remount. If you are on the winning horse, enjoy the ride.
Today RuPaul is 65. Danny DeVito turns 81 while Martin Scorsese celebrates his 83rd.
James M. Meyer, CFA 610-260-2220