Trillion Dollar Hunt on Wall Street Due to AI Gold Rush
It was the week of AI on Wall Street.
After Apple presented its AI plans on Monday, Wall Street remained quiet, but after a day’s respite, the stock market went into overdrive on Wednesday: Apple briefly overtook Microsoft as the world’s most valuable company, whereas just last week it had lost the second spot to Nvidia. What is wrong with these investors? Why the absurd swings of hundreds of billions?
Trillions, not billions of dollars
As of June 14, Apple, Microsoft and Nvidia have all passed the milestone of a $3 trillion market value, making them an exclusive “trillion-dollar” club. Due to enthusiasm among investors about the upcoming introduction of AI applications and ChatGPT in the iPhone, Apple briefly regained the title of the world’s most valuable company with a market capitalization of $3.283 trillion, just slightly higher than Microsoft’s at $3.282 trillion.
Meanwhile, Nvidia’s stock price also rose steadily, driven by investor excitement over Nvidia’s 10-for-1 stock split. Nvidia’s market capitalization experienced a meteoric rise, from $2 trillion to $3 trillion in a record-breaking 96 days, faster than Microsoft (649 days) and Apple (718 days).
Nvidia’s dominance in AI chips and strong earnings growth have fueled share price gains, with shares up more than 132% this year and 193% in the past year. But on Friday, Apple’s share price fell slightly, Microsoft rose slightly and so Microsoft ended the stock market week as it began: as the world’s most valuable company.
Everyone was buying iPods, not AAPL
Let’s face it: if the so-called investment gurus understood anything about technology, they would have been buying Nvidia shares en masse years ago. But just as there was no one 21 years ago who, instead of buying an iPod at that price, $300, bought shares of Apple (which would be worth $137, 000 today), there is virtually no professional investor who has been in Nvidia for more than, say, five years.
Professional fund investors are enthusiastic amateurs, just like you and me. Why are Apple ($3.26 trillion), Microsoft ($3.29 trillion ) and Nvidia ($3.24 trillion) now worth almost as much? Ask any analyst or investor and they’ll say in chorus: because of AI. That’s like an artist manager saying, “Doesn’t matter if I’m manager of Taylor Swift or the girl in the office who does great on karaoke nights, they’re both singers.”
The market does not seem to understand that these are totally different companies with different approaches to revenue, costs and possibly profits from applications of AI. But their perspectives are completely different.
Nvidia makes axes and shovels, Microsoft does mining
In AI, we can speak of a pure gold rush, so let’s stick to that metaphor for now. In this gold rush, Nvidia makes the axes and shovels that every miner needs. Amazon, Meta, X, Tesla, Oracle, we can go down the list of tech giants: all of them, like Microsoft, use Nvidia’s shovels and picks. Except for Apple.
There is no alternative for Nvidia that delivers the same performance per dollar invested, which is why Nvidia’s revenue growth and profit margins are already legendary. The question is how long Nvidia can maintain this position, but it looks to be leading this market at least for the next few years.
Microsoft is the biggest miner, with worldwide data centers full of Nvidia stuff. With Microsoft, unlike Nvidia, the question is whether those billion-dollar investments will lead to sufficient margin. The first noises are already being heard that Microsoft’s customers are not at all achieving the intended improvement in returns based on Microsoft’s AI applications.
This will obviously lead to price erosion and lower sales and undermine investor confidence in Microsoft’s AI plans, because the costs for Microsoft are still astronomical. These are not investments of billions, but tens of billions, and they will start to gnaw away at the profit margin.
OpenAI is goldsmith, Apple the jewelry maker
OpenAI sits just a layer above Microsoft: it uses Microsoft’s data centers and cloud services to forge gold, demanding maximum power from Nvidia chips. Demand for OpenAI’s technology, particularly its flagship ChatGPT, has been huge. Meanwhile, OpenAI is heading for annual sales of nearly $3.5 billion.
That’s why shareholder Vinod Khosla remains unabatedly optimistic. No wonder, he got in at a valuation under a billion and has already seen his investment increase hundredfold in value. Who wouldn’t smile affably at any criticism of OpenAI?
Under Sam Altman’s leadership, OpenAI likes to leak revenue figures, but we hear nothing about its burn rate, its losses. This is no wonder, because OpenAI pays enormous sums to two suppliers: Microsoft and Nvidia. Both seek maximum profit, and therefore OpenAI burns billions a year. The billions that Microsoft invests in OpenAI, it gets paid back largely by services provided.
No, then Apple, the jewelry maker of AI mining. It builds AI applications here and there into its operating system and applications, which it doesn’t call AI but Apple Intelligence, but the cost of these investments, in the big picture at Apple, is virtually marginal. And the partnership with OpenAI announced with a lot of fanfare on Monday, costs Apple nothing at all.
What Apple does is create elegant, easy-to-use products that improve everyday life, similar to turning raw gold into fine jewelry. Apple’s ambition is to offer consumer products and services that seamlessly integrate AI to improve our daily lives. The only question is: Will those AI ambitions from Apple work this time?
Apple Intelligence is Siri 2.0?
Investor and former journalist MG Siegler rightly points to all of Apple’s previous (and failed) attempts to get Siri working properly. It’s the same issue that Marques Brownlee pointed out in conversation with Apple CEO Tim Cook.
More than two billion iPhone owners will get an update before the end of the year that will allow them to enter the AI era — provided their iPhone can handle it and, as a result, Apple may well get a huge sales boost from the iPhone 15 and the new iPhone 16, while for years renewing your iPhone was virtually unnecessary. Not anymore. Whoever wants to run the new Apple Intelligence, needs at least an iPhone 15.
By the way, it’s remarkable how times have changed: just last week I pointed out a podcast with legendary Wall Street Journal reporter Walt Mossberg, who used to scare the life out of the entire tech elite. Nowadays, it’s YouTuber and professional frisbee player Marques Brownlee for whom the red carpet is rolled out at the introduction of a new product.
Traditional journalism is struggling to make sense of Apple’s introduction of AI applications. This makes sense in itself, as it is all still smoke and mirrors and nothing can actually be tested yet.
It’s just sad that the Washington Post didn’t get much further than to point out the hilarious pose with which all Apple employees are portrayed these days. It has since been flatteringly christened the “Apple Power Stance,” but more often than not it looks like someone who just got off a horse.
Too much focus on LLMs and Generative AI
Those who would have followed the technology sector from some distance this week would undoubtedly get the impression that Large Language Models (LLMs) such as ChatGPT and Google Gemini, are the only and most important form of AI.
But criticism of the hundreds of billions being thrown into this branch of AI is rightly growing louder. Martin Peers of the Information has a sharp analysis:
Despite the ubiquity of AI in news coverage, one point doesn’t get enough scrutiny: Are the advances society is going to get from the new technology worth the cost?
By cost, I particularly mean the impact on the power supply. Data center demands for power threaten to suck the energy world dry, setting back efforts to shift away from carbon-emitting power sources. And to what end? Judging from how some tech companies are marketing their AI-powered services, it’s about helping consumers design a menu for a dinner party, plan a vacation or find a photo on their phone. (The idea, of course, is to get consumers to spend money on new devices or AI subscriptions.) Or it’s about helping businesses improve employee productivity, including by cutting jobs (great!). Despite the attention these uses have drawn, the real promise of AI surely is its potential to help solve existential challenges like deadly disease or dangerous drivers.
Efforts to utilize the new tech in those directions are underway. Google, for instance, has its AlphaFold project aimed at accelerating cures for diseases. Elon Musk is making AI the centerpiece of his efforts at Tesla to develop fully autonomous driving. Microsoft, meanwhile, is using AI to try to improve cybersecurity, among other things — an effort that can’t come too soon. Data breaches at major companies — including one recently at Snowflake, which affected several of its customers — have become so commonplace that they draw little attention despite the pain they inflict. No one knows the issue better than Microsoft, whose president, Brad Smith, was hauled over the coals about the company’s cybersecurity failings at a congressional hearing on Thursday.
But fighting deadly diseases, solving autonomous driving and even fixing cybersecurity aren’t cheap or quick undertakings. Yet big tech companies are spending tens of billions of dollars to develop AI, so they need a return. The danger is that the need for quick returns from the consumer and business services will skew investments, shortchanging more important needs. Unlike AI startups like Anthropic and OpenAI, big companies aren’t governed by nonprofits requiring them to put humanity’s interests first. Let’s hope AI’s biggest advances don’t end up being trivialities like saving consumers a bit of time while they play with their phones.
“LLMs suck oxygen out of any space”
Leading thinker on technology Jennifer Zhu Scott puts it this way:
- LLMs will eventually be commoditized
- LLMs are the game for only a handful of companies in the world
- LLMs aren’t the future of truly advanced AI; instead, more efficient, less data/compute/energy-hungry architecture that can more closely mimic biological brains is.
- LLMs create enormous carbon footprint and water consumption, and a large pile of the outputs is plausible bullshit, nonsense “art,” deep fake, and massive privacy invasion
- kudos to those who keep their LLMs open-source
- @OpenAI’s release ChatGPT3 in Nov 2022 sets off this mad race of LLMs actually set back the progress of advanced AI for at least 5 years
Jen Zhu Scott said previously:
It is simple: we don’t get on Mars by building taller and taller buildings on Earth — Mars being general purpose AI, or AGI.
LLMs, as clever as some applications are, remain advanced forms of the best player in a pub quiz or the scholar who always raises his finger trying to give the answer. Only the answer does not always turn out to be correct, and the facts that the diligent scholar spits out are based solely on learned knowledge.
Fundamental technology that advances society, e.g. helps eliminate disease, will not be developed on the basis of LLMs. This puts the billion-dollar investments and trillion-dollar valuations for companies engaged in the current form of AI, LLMs, in an increasingly questionable light. See you next week!
