Stephen Morgan

Right now I am positioned for the market to head to new highs before the holidays and the goal is to catch as much of this inflating AI bubble as possible. As Mae West said, "Too much of a good thing can be wonderful!" In an interview on Friday, I said “Unfortunately the market is going to keep melting higher,” and I wanted to expand on that statement. I said unfortunately because the higher the AI bubble takes stock valuations, the more painful the decline is going to be ultimately for investors on the other side. Let me explain my current thoughts and why I remain positioned to hopefully benefit from valuations inflating further. It is important to be positioned for the road right in front of you while contemplating what turns might be ahead. Turning too early can put you in a ditch. Measured by the price to sales ratio of the S&P or market cap to GDP (Warren Buffett’s favorite), the stock market is more expensive today than the peak in 2000. But this does not mean they cannot go up even further in the near-term. There is a reason John Maynard Keynes said the market can stay irrational longer than you can stay solvent. So while I believe the market is suffering from “irrational exuberance” given current valuations, I believe it can extend even further fueled by easy money, strong Q3 earnings and AI optimism. As a reminder, Greenspan gave that famous “irrational exuberance” speech in December of 1996 during the internet buildout and the S&P doubled by its peak in March of 2000. Nasdaq rose 40% in 1998, then accelerated to 86% in 1999 and accelerated even further to a 24% gain in just over two months to start 2000 fueled by the inflating internet bubble. From the launch of Netscape Navigator in late 1994, the first mass market internet browser, it took over five years for the stock market to reach its bubble peak during the buildout of the internet infrastructure. Also these accelerating gains occurred despite the Fed raising rates from 4.75% in June of 1999 to 6.5% at its peak in 2000. This time, the Fed started cutting rates again from a range of 4.25-4.50% in September. The Fed is planning to spike the punchbowl before the holidays despite inflation being above target for four years. This is despite the Fed’s role of “taking away the punchbowl just as the party is getting started” to paraphrase former Fed chairman Martin. In addition, they will probably end quantitative tightening given recent credit events at the upcoming meeting at the end of October. Finally, in May of 2026, we are likely to get a new head of the Federal reserve appointed by President Trump who wants rates much lower than current forecasts. These additional rate cuts will also probably happen before the end of 2026. But this is also when my concerns over AI come in. As for AI optimism, we have not even reached three years since the launch of ChatGPT in late 2022. $NVDA quarterly revenues are up “just” 7.9x since the end of 2022 and the introduction of ChatGPT. $CSCO, which became the most valuable company in the world during the internet infrastructure buildout, saw quarterly revenues increase 15.5x from the end of 1994 to its peak. My medium-term concern remains that by late 2026 that AI revenues will not be ramping enough to validate the enormous ramp in spending in AI capex. It is a good approximation to assume that the five biggest hyperscalers ( $AMZN $MSFT $GOOGL $META $ORCL ) are about 2/3rds of the global spend on AI capex. Spending of these five is expected to reach $800B over the past three calendar years and over $400B up 60% y/y in this year alone. But the revenues generated by the AI native companies such as OpenAI, Anthropic and Perpexity is only expected to reach ~$20B this year. In an August interview, Sam Altman said how they had resisted the temptation to put a sex bot avatar on ChatGPT even though it could juice their revenues. Last week he said OpenAI would allow erotica on their platform by year-end. I think this speaks to the difficulty OpenAI is having getting customers to pay for their product relative to spending levels. When consumers have been trained over decades that they can get internet query results for free from Google, it is going to be hard to get them to pay for AI chatbot queries. There are 5 billion users querying Google. $NFLX has over 300M subscribers and is expected to generate $45B in revenues this year and $9B in free cash flow. By comparison OpenAI has over 800M users but is expected to generate just $13B in revenues this year and burn $8.5B in cash. But this is a concern for another day. Right now my goal is to be positioned to enjoy too much of a good thing.