TIP823: From Railroads to AI: The Timeless Patterns Behind Market Bubbles w/ Kyle Grieve

Kyle Grieve argues that bubbles are not rare anomalies but consistent expressions of human psychology, specifically greed, optimism, and social proof, and that every bubble is sustained by the belief that this time is different, which causes investors to justify unreasonable prices or excessive leverage.

Grieve draws on Charles Kindleberger's five-stage framework as interpreted by Ron Insana. Displacement begins with a new technology or structural shift that pulls prices forward. Overtrading follows as retail participation drives prices higher. Monetary expansion lowers the cost of equity, enabling companies to use inflated shares for acquisitions and allowing investors to borrow further into rising assets. Revulsion is when institutions quietly exit and sell to retail investors, often invisibly to casual observers. Discredit is a complete sentiment reversal where the formerly loved asset becomes universally hated. Insana's own five-ingredient framework for bubble formation requires a eureka moment, easy money, government largesse, strong economic conditions, and an external stimulant, and Grieve notes that bubbles do not form during recessions but require GDP growth, low unemployment, and optimistic consumers.

Grieve describes a bubble as a form of time distortion in which all future growth and cash flows are crammed into the present price, leaving zero forecasted returns. RCA illustrates how a genuinely successful business can still become a bubble through multiple expansion. RCA's net income compounded at 35 percent annually during its bubble period, yet its price-to-earnings multiple expanded from 15 times in 1923 to 285 times at its 1929 peak, with the stock rising from 5 dollars to 600 dollars before taking 36 years to recover. The plank road bubble of the mid-1800s shows that these patterns predate modern markets. Between 1847 and 1857, 1,388 plank road companies incorporated across 17 states. Promoters promised annual dividends of 10 to 40 percent and claimed wooden planks lasted 10 to 15 years when the actual lifespan averaged around four years. One investor named John Taylor put 900 dollars into three local companies and received combined dividends of less than 80 dollars over 12 years, implying roughly 0.7 percent annually.

The tech bubble of the 1990s was the largest in modern history and uniquely involved widespread public participation through 401ks, mutual funds, and day trading. Bank deposits fell to 50-year lows while stocks grew to 58 percent of household financial holdings. In March 2001 the Nasdaq traded at 246 times earnings against a historical range of approximately 40 times. Wall Street invented metrics like page views and unique visitors to justify valuations when traditional measures failed. Seventy-seven percent of IPOs during the dot-com era had no profits. The underlying narrative that the internet would change the world was correct, but value creation unfolded over decades rather than quarters. Fraud flourished during the period because oversight became lax and euphoria was intense, with Enron booking imagined profits as current income, using off-balance-sheet debt, and manipulating California energy prices at a cost to the state of approximately 11 billion dollars.

On current conditions, Grieve observes that the S&P 500 traded at 31 times earnings at the end of 2025, but removing the Magnificent 7 leaves the remaining 493 companies at only 19 times earnings. US GDP growth in 2025 was 2.1 percent, the lowest since 2019 excluding the COVID recession, and unemployment is around 4 percent. The 2025 IPO market raised 38 billion dollars compared to 142 billion in 2021. Grieve does not consider the broad S&P 500 a bubble but views the assumption that its elevated price-to-earnings ratio represents a new permanent base as more dangerous than the valuation level itself. He is hesitant to call the Magnificent 7 bubble-like given their quality and growth, noting Nvidia's forward price-to-earnings falls to around 24 times on analyst estimates of nearly 60 percent EPS growth, below the S&P 500's 27 times forward multiple.

JP Morgan analysts estimate the current AI infrastructure buildout will cost around 5 trillion dollars while Microsoft, Alphabet, Amazon, Meta, and Oracle combined hold only about 350 billion dollars on their balance sheets. Thinking Machines, an AI startup with no product founded by former OpenAI executive Mira Murati, raised approximately 2 billion dollars at a 10 billion dollar valuation in its seed round and is now seeking a new round valuing it at 50 billion dollars. Grieve's general strategy is to avoid investments spending money on the AI buildout and instead own businesses already leveraging AI capabilities with recurring revenue and profitability, citing Lumine and Topicus as examples.

Grieve's practical bubble-detection tools include reverse-engineering terminal valuations to test whether implied assumptions are physically possible, monitoring merger and acquisition prices in industries he follows to judge whether transaction prices are narrative-driven, and treating a share price rising five to ten times without corresponding intrinsic value growth as a concrete warning sign. He acknowledges it is impossible to avoid narratives entirely but treats multiple expansion disconnected from compounding intrinsic value as the key red flag, and says he is willing to sell even businesses he likes if narratives rather than fundamentals appear to be supporting the price.