It's all over but the shouting.
So, what are we to make of Jensen Huang's meteoric crash? For starters, honey traps (a term coined by John le Carré) and the cockroach theory both remain undefeated, as the black-leather-jacketed-one displayed all the judgment of a frat boy on Bourbon Street. Nvidia's board, which hadn't appeared fit to oversee a petting zoo, cut Huang loose on Thursday morning, two weeks after his entanglement with an alleged Chinese spy was revealed.
Bigger than that and salient to Barron's readers, L'Affaire Huang undermines the artificial-intelligence trade, which has been buttressing the market. Questions about the bull case for AI, which have been gnawing at the market for weeks, are now exacerbated by Nvidia's tempest. It isn't just that this bellwether stock is down nearly 25% since the news broke; it's that Huang and the company have been inextricably linked -- and with Nvidia's graphics processing units and financing, the epicenter of the AI universe.
Expect guesswork and volatility to follow. Nvidia is slated to report earnings next Tuesday, but the even-hotter potato will be Anthropic's initial public offering, which looks to be next Wednesday. The importance of the latter can't be overstated, according to analyst Cory Johnson of IPO research outfit The Quiet Period. "Investors are dying to get hold of Anthropic," he says. "It's the most important part of the ecosystem. Semiconductors, data centers, power generators, the only way all that stuff works is if the LLMs [large language models] are profitable."
At some point, the AI trade will run out of exuberance. How will it end? It's a dilemma that T.S. Eliot wrestled with a century ago in The Hollow Men, which as you remember closes with, "This is the way the world ends / Not with a bang but a whimper." In other words, the market's love affair with AI could end with bang (i.e., a crash) or a long whimper down.
Maybe this is wrong-footed, though, or what logicians call a "fallacy of bifurcation," a premise that falsely limits a decision to two alternatives. (In a 1958 interview, Eliot copped to this, saying that he wouldn't write "bang" and "whimper," again because he wasn't sure the world would end with either -- which it hasn't.)
So, instead of the AI trade ending either by crash or slow burn, the market takes an evolutionary tack. It's a theory that came from a conversation I had with AI scientist Kai-Fu Lee, author of a new book, AI Native.
I asked Lee how investors should play AI. "Clearly, I would pick applications," says Lee, who has been a senior executive at Apple, Microsoft, and Alphabet's Google. "If you look at history, from PCs to mobile to AI, semiconductors moved up first, like Intel, Qualcomm, and now Nvidia. Then, platforms like Windows, iOS, Android. But applications add up to be worth way more than platforms or semiconductors. With the PC, think Oracle and Microsoft Office. For mobile, it's Instagram and TikTok. For the AI industry, apps are just beginning."
Like which companies? Lee points to Palantir Technologies, which is using AI to help its customers in decision-making. And software-as-a-service companies? "A SaaS company trapped in one function is going to face trouble," he says. "Databricks can be an exception because they're not trapped like that." What about Salesforce? "Salesforce has leadership, history, and very strong AI teams. They probably have a shot at getting out of it OK."
Lee says that giant Chinese AI consumer applications like Pinduoduo and Doubao are far ahead of the U.S. companies. But isn't Meta Platforms well positioned? "Meta would be the last company I would invest in as a consumer software company," Lee says. "They have every reason to be a leader in AI, but it has been very mismanaged."
What about Amazon.com, Google, and Microsoft? "They're in fine shape because they're trusted by enterprises to store their data. Also, there's going to be a growth of data centers, and they can predictably know their demand." Lee wouldn't invest in companies that just invest in data centers.
Lee says the picture is less clear for Apple. "AI's really bad for Apple because they have the legacy of apps, and apps are going to eventually be accessed by the AI agents primarily, not the user." Lee says. Apple will have to migrate the iPhone to become a more AI-centric device. "You have to eat your own lunch. Steve Jobs did it, forgoing the iPod and putting it into the iPhone," he says. "Apple has the strengths to build it."
As for Anthropic and OpenAI, Lee says "being pure-play AI has advantages. They don't have the baggage of other companies. History favors pure-play companies because they can get their whole pure-play valuation out there. It's hard to say Google should have the same P/E or price-to-sales ratio as Anthropic because Google is a mishmash of many different things."
What does Lee himself invest in? "Only AI quant funds that buy and sell stock based on AI's predictions," Lee says. Spoken like a true believer.
Speaking of which, market sage Howard Marks dropped a voluminous post on Wednesday, A Layman Thinks About AI. As usual, Marks brings zingers, including this one, attributed to American economist and writer Stuart Chase: "For those who believe, no proof is necessary. For those who don't believe, no proof is possible."
For AI believers, Huang has left the building. Enter Anthropic.
Comments