By Adam Levine
Oracle's second-quarter earnings report underlined the fragility of the AI trade in the second half of 2025. It doesn't take much bad news from one company to take down the whole industry the next day.
The volatility is driven by the fact that investors are now filtering every piece of news through the bubble lens and have one foot out the door. It's an understandable reaction, but it might be self-defeating.
The question for investors shouldn't be: Are we in a speculative bubble? The questions are: Where are we in the process? And what precipitating event could pop it?
The term "bubble" always gets associated with an imminent pop. But that hasn't been the history. Quoting John Maynard Keynes, my father used to say about shorting a bubble, "The market can remain irrational longer than you can remain solvent."
Ultimately, the bubble is a journey, not an event.
The most irrational of all bubbles was the Dutch tulip craze in the 17th century. What began as a luxury good and a badge of wealth in the world's richest country turned into a speculative investment boom that sucked in even retail investors. It lasted almost three years, with a two-month frenzy at the end.
The British railway mania lasted about four years. Federal Reserve Chairman Alan Greenspan began talking about "irrational exuberance" in 1996, but the bubble didn't pop until 2000.
The dilemma for investors is that the latter stages of a bubble are usually the best time to be invested.
From Greenspan's 1996 speech until the bubble popped, the Nasdaq Composite was up 288%, including an 86% run in 1999.
Arguably, the dot-com bubble pop had less to do with the fundamentals of the internet and more to do with Japan slipping into a recession. Once the tide went out, the many unprofitable internet companies that had recently gone public were caught naked and soon had to declare bankruptcy.
Cisco Systems, which had been the poster child for the bubble in much the same way Nvidia is today, took the biggest hit, bleeding out 89% of its value over the next 2 1/2 years. On Wednesday, Cisco posted its first all-time closing high since 2000. That Cisco finally rebounded 25 years later could be good news or bad news for today's investors, depending on your time horizon.
Back to Oracle and this bubble. What caught everyone's attention in Oracle's quarterly report was its sudden cash crunch.
Once a cash-flow machine, Oracle had a free cash loss of $13 billion in the past 12 months, with $10 billion of that coming in the latest quarter. To make up for it, Oracle borrowed $18 billion in September, and there's likely much more to come.
Oracle stock fell 11% Thursday and took a bite out of AI hardware providers Nvidia, Marvell Technology, Broadcom, Micron Technology, Dell Technologies, and Hewlett Packard Enterprise, which sell to cloud providers like Oracle. They all fell at least 1% on Thursday, even as the S&P 500 closed in positive territory.
Let's not pick on Oracle, though. There are many bubble-adjacent indicators in public markets. We've seen a speculative rush into a single asset class: artificial-intelligence data centers. There have been circular financing deals with suppliers funding customers. Meta Platforms recently used a joint venture with alternative investment manager Blue Owl to keep $27 billion in debt off its balance sheet, and we are likely to see more of this type of "creative" financing in 2026.
This bubble began inflating in early 2023, centered around Nvidia and its industry-leading AI hardware. It soon mushroomed from there to other infrastructure companies, cloud providers, and energy firms. We see inflated valuations, but most aren't spectacularly high like in 2000. Looking only at public markets, we are likely halfway through a bubble inflation.
But unlike in 1999 and 2000, when we saw a record surge of initial public offerings, today's bubble may be inflating in private markets. Since 2024, $362 billion has gone into AI and energy start-ups, according to Crunchbase. In November alone, there were 14 venture capital rounds of half a billion dollars or more. Much of the data center buildout is being funded by private credit, like we see in the Meta--Blue Owl deal. Pimco owns a lot of data center debt, and remains one to watch.
Public-market investors had visibility into the 2000 bubble, but much less today -- and there's the rub. While everyone looks for signs of a bubble in the stock market, it may be lurking elsewhere, and be less apparent than a quarter-century ago.
But investors can still keep a close eye on signs of weakness in private markets. Are equity funding rounds and valuations still growing quickly every month? Are we seeing more blockbuster credit deals for large data centers, or is that slowing?
Two tests will likely come next year when the two most valuable start-ups, SpaceX and OpenAI, may go public at valuations expected to exceed one trillion dollars. It could be the beginning of the private-market bubble seeping into public markets, and that's when everyone will need to be on their toes.
I'll leave you with one optimistic thought: Whenever this bubble pops, we're sure to be left with the ingredients to build the next wave of innovation. The network infrastructure that was built during the dot-com boom became the backbone of Web 2.0. Going back to the 19th century, the railroad boom and bust left a transportation network that continues to benefit the global economy.
If you bought Cisco stock on Dec. 5, 1996, the day of Greenspan's "irrational exuberance" speech, your annualized total return would be 10.3%, edging out the S&P 500. Here's to bubbles.
Write to Adam Levine at adam.levine@barrons.com
This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
December 12, 2025 03:30 ET (08:30 GMT)
Copyright (c) 2025 Dow Jones & Company, Inc.
Comments