Information is everywhere, Insight is priceless

Trying to analyze the market’s irrational exuberance in real time is like trying to pinpoint exactly how drunk you are at the after-party. Beyond the hour of good and evil, there really is no telling where you are, or when it will crash into a lot of bad noise.

As it stands, US stocks are on a hoot, leaving both the West and the Rest splashing in the wake. Warren Buffet has famously said that there aren’t many good foreign stocks in which to invest, before unwinding an enormous position in very American Apple to sit himself on a Scrooge McDuck-sized pile of cash, whatever that tells us. Actually, it tells us quite a lot. Valuations are extremely high but we knew that already, and beating that drum in our current state won’t do any good. Nvidia and Tesla are trading at premiums of 30x and 125x earnings respectively. Collectively, the S&P500’s Mag 7 accounts for a third of the market value and a quarter of the profits. Most of this value is swirling around AI, which BlackRock has called a “mega force” that has buried the rest of the business cycle.

It’s also a mega-blind spot for a single enormous sector that is doubling down investment on whats being billed as a moonshot technology (and it may well be) that has yet to roll out a game-changing application to pay for itself. According to RAND, some 58% of mid to large sized businesses have deployed at least one AI model in the two years since they were unleashed. Adoption on a more regular basis, however, hovers closer to 5%. Without some quantum jump in efficiencies, application and focus before investors get spooked and customers feel cheated, we are looking out a tremendous slip in both market and the tech supporting it.

The first point, the market, is more simple than straightforward. Investors are not known for their patience – with so much instant data available, they tend to gaslight themselves into making ill-advised moves. Will the AI genie deliver? Probably, but just not in the way the developers envision. So the market will correct and it will hurt and that will be that. If you are playing with house money (read: your winnings) the correction will be tedious, but not catastrophic. If you’ve managed to leverage yourself into your positions with money you can’t afford to lose, you’ll probably be less philosophical about the reversal. Creative destruction will follow and those of us capable of critical thinking will be fine. Heading into 2025, the problem for Big Tech is nearly as phycological as technical or qualitative: It’s more about emotion and moon-dust than fundamentals. The real danger is over-leveraged investors getting spooked.

The market is widely believed to be a consensus of participating investors. That’s true enough, but it fails to account for the erratic volatility which is where you generally get your hat handed to you. So what you need to keep your eye on isn’t the consensus, but the large and very motivated seller (or buyer) bucking it – perhaps to cover losses elsewhere, or our over-leveraged investor above is getting squeezed. With valuations concentrated so asymmetrically, it will take shockingly small number of determined investors running for the door en masse to trigger a panic to drag across the entire market. As the trader and insufferable philosopher Nassem Taleb writes “The market is like a large movie theater with a small door.” And as such, “…the best way to detect a sucker is to see if his focus is on the size of the theater rather than that of the door.”

You don’t want to be the sucker.

AI: Hype & Glory

A Trillion Dollar Answer in Search of a Trillion Dollar Question

The second point is the technology itself. Nuclear ICBMs are expensive too, but at least we know what they do, how they do it and why we shouldn’t . At the moment, AI lacks the clarity of a fiery armageddon. The costs of building new generation AI models beyond most countries, let along most companies mere mortals; Microsoft plans to spend $80bn on data centers alone this year.

OpenAI launched a $200-per month subscription service last month. At that price tag, the company has plenty of takers, but they way users are deploying the technology means that the company is still loosing money on its subscription service (let’s not even get into the development and training costs). The cost of a query for a AI chatbot in computing power is 10X that of the unfashionable google search - but without the same advertising space. More subscribers won’t fix the issue, the price doesn’t cover the computational and energy costs.

For AI to pay returns it needs to produce, it needs to get to a point where it is replacing humans. To prove that day is around the corner, OpenAI recently matched an “average” human against its latest model on a set of problems designed to measure general intelligence. According to the company, the two performed roughly on par. Assuming that the tests weren’t juiced (no small assumption), the AI model cost about a $1,000 per task, where our average human cost $5. Putting aside that no one really wants to hire an average idiot at, say, $4,000 an hour, there are less qualifiable problems to consider. While obvious to anyone who has spent any time in advertising or sales, yet rarely occurs to the sort of people who run these tests, humans (average or otherwise) are rarely in controlled, context-free situations disconnected from of future consequences. The sort of situation that AI requires to require to get through a task.

At this point AI is still evolving toward a profitable usefulness, but we aren’t there yet. Remember that the first wave of the internet didn’t work as originally envisioned by its developers. It was geeky amateurs tinkering with the technology in a way that mainstream users found practical and created the all-encompassing hell-scape that it is today. For its part, Generative AI has armies of people just tinkering around. I’m no techie, but I’m fairly certain that they’re the reason why my instagram feed was suddenly a cavalcade of AI generated ladies with cartoonishly large breasts is largely due to people at home…tinkering. Although I’m not sure how they are making money off the exercise.

At any rate, the smart money is that some commercial application, or many, will arrive. Just not the way OpenAI’s Sam Altman thinks. Breakthroughs will be from people that you’ve never heard of, whose job descriptions is a lot more practical than “Disruptor in Chief.” It needs to happen on a grand scale because the R&D costs are going to be a bitch to recover.

With that in mind, if you are holding Nvidia, and the price drops, ride it out, you’re still ahead. Although the best hopes for bringing AI cost down to Earth is more competition in the right sort of chips, which will squeeze those Nvidia valuations. The future of the tech itself is likely to become focused and more nimble: Using AI for global shipping logistics to manage real-time inventories doesn’t need to plot solutions in the style of Shakespeare, or be sexy about it, either. Without a doubt a Nobel prized winning brain surgeon/best seller author could handle the pressures of being a short-order cook, but why the hell would any manger pay up for the expertise overkill? The technology only needs to perform the task required faster and more accurately than a human. Smaller, “stackable” models that are easily trained for purpose, processing well-ordered data is the sweet spot. So the clever thing is to focus on companies like Databricks which is helping companies to adapt and optimize existing data to the requirements of AI.

The returns on the second order of AI tech is already enormous; those data centers suck up a lot of power. Germany may be having a rough year, but Siemens Energy’s stock is up 300%, a better return than Nvidia, largely due to its grid-technology business. Hitachi, a Japanese conglomerate, has tripled in value in two years in the same sector. Global grid infrastructure investment will get a boost from AI, but isn’t dependent on it. That’s where the stability comes in: Europeans aren’t that environmentally conscious, but they’ve been looking for energy security since the middle of the 19th century (that race for colonies wasn’t really about religion and glory…). As emerging economies get rich enough to afford a middle class, they will buy more cars and, crucially, air conditioning because, in the immortal words of Robin Williams, it’s “hot and shitty” in the Global South.

A lack of investor patience for the AI moonshot to profitably take over our lives will likely cause an economic downturn, as well as give a high-colonic to the tech startup world. It will be a rough ride in the center, but the second order technologies will be a fair place to stash your winnings until the next wave washes out the wreckage. Although, it probably won’t happen this year. Still, the genie is out of the bottle and a lot of tinkerers are making wishes on the investor’s dime.

Some wishes will come true, so we’d better be careful what we ask for.

Note: This is for information and entertainment purposes. The 4717 isn’t endorsing any of the entities mentioned in the above insight.