SITALWeek #425
Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.
Click HERE to SIGN UP for SITALWeek’s Sunday Email.
In today’s post: the Zoltar arcade machine gives us some insight into the billions upon billions being poured into AI development; the Chinese fast-fashion loophole that is fueling the digital advertising industry; the afternoon snack war; layoffs are low as employment remains high; the widening political gender gap; and much more below.
Stuff about Innovation and Technology
“Your Wish Is Granted”
Lately, you can’t interact with a news algorithm without hearing something about another $10B+ AI spending plan. Alphabet spooked investors last week by their reported spending of $11B in capex for the fourth quarter – driven “overwhelmingly” by “technical infrastructure” – and their expectation that capex will be “notably larger” in 2024. Google started the year with $6.3B in Q1 and steadily ramped to the monster sum of $11B, giving a total of $32B in capex for 2023. Just using the Q4 run rate would imply ~$44B for 2024, or around 15% of the revenue that analysts expect the company to earn. Meanwhile, Microsoft reported a similar $10.7B in capex including finance leases for Q4, up over 20% from the prior year. We’ve also learned, in an interview with Mark Zuckerberg I referenced two weeks ago, that Meta plans to buy billions of dollars of GPUs just this year. And, OpenAI’s CEO Sam Altman is rumored to be raising billions to set up joint ventures with chip makers like TSMC to help fund chip fab. I suspect this maneuver is largely out of fear that OpenAI won't be able to secure capacity at TSMC fabs for GPUs (or, potentially, their own custom AI chips). Elon Musk, for his part, is rumored to be looking to raise $6B for X.AI, which seems almost adorably quaint given how little progress that would fund relative to the spending intentions and deep wallets of the competition. Trillion-dollar investment giant Blackstone has $15B of new data centers in development (in somewhat of a circular reference) as they buy into the capex hype. While I’ve speculated in the past that a breakthrough in efficiency, developed by AI itself, could decrease the need for leading-edge chips, that appears to not be on the immediate horizon.
All of these speculative billions upon billions of dollars in expenditures raise the obvious question: what business model is going to pay for it all? Hundreds of billions were spent to build the Internet; but, large-scale business models to justify it didn’t show up until around a decade after the dotcom crash. Currently, the leading-edge chip industry (and many trailing-edge chips) are largely paid for by two industries: advertising and cloud software. Behind the ads lie a variety of entertainment choices (e.g., video, social networks, games, etc.). I am of course generalizing here, but the vast majority of use cases for leading-edge chips are phone processors, servers, GPUs, networking, wireless connectivity, etc. Digital ads (including those funding search and chatbots) and cloud software are each multi-hundred-billion-dollar industries that (in yet another circular reference) are creating even more demand for leading-edge chips. When I look back at my post titled End of Advertising-Funded R&D? from November of 2022, it appears I was overly pessimistic about the prospects for the next wave of new markets, despite my enthusiasm at the time for AI transformer models and AR.
Let’s ignore for the moment the euphoric spending cycles early in new tech platforms that lead to bubble formation and speculate on what might sustainably pay for hundreds of billions of dollars of new chip capacity. First, I think we can safely say that the advertising (entertainment) and enterprise software industries will be heavy users of AI, and this usage will drive higher revenues for all of these segments. We can now see a clear, near-term path to holodeck-like world creation with simple text prompts. Such a feat will require a massive amount of chip investment and new leading-edge hardware; and, if it’s as fun as my VR boxing app, people and advertisers will be happy to pay for it. When you look at the progress that Google is making with projects like Lumiere, it’s not a leap to imagine how much processing power we will need. Further, AI is going to steadily replace and/or augment all white collar jobs to such an extent that companies will be more than happy to pay up for their new AI employees.
Robotics is another example of a market that could grow into the trillions of dollars in the coming decades with a variety of special purpose and general humanoid form factors powered by AI. Our new AI-enhanced assistants will be massive consumers of chips for both the hardware and cloud processing necessary for navigating our complex world.
Now, let’s go out a little further on the limb and speculate what might really be behind this euphoric FOMO GPU land grab. What’s the real pot of gold that these billionaire tech CEOs might be after? My candidate is: predicting the future. I discussed this in more detail in my longer post titled Simulacrum a few months ago. Effectively, AI is going to give us the ability to simulate countless realities in parallel, and, in doing so, will create a new type of modeling and virtual forecasting. Remember when sports books were using video games of professional teams programmed with player stats to gauge who might win real-world games? Now expand that concept until you get to a “game” that predicts the entire world on a day-to-day (or longer-term) basis (or, perhaps you might not remember this, because a quick conversation with Google Bard leads me to believe that video games were never used to calculate gambling odds!). Ultra-powerful AI holds the promise of a real Laplace’s Demon with perfect knowledge of the past, present, and future (for more on this thought experiment, check out the TV show Devs on FX/Hulu). When Nvidia announced its Earth-2 simulator in 2021, the blog post was titled “...to See Our Future”. If you had a slight edge in knowing, for example, what type of unexpected and erratic behavior could create a new type of political candidate, how much would that be worth? The benefits (financial and otherwise) of being able to predict anything, from what therapeutic drugs might work to quarterly earning stats, are obvious. Of course, all of this goes against the fundamental tenets of the biggest foundational belief we have: the world is a complex adaptive system with emerging behaviors and wildly unpredictable outcomes based on chaotic initial conditions that are constantly evolving. So, the question is, with enough chips, could you overcome the limitations of living in a complex adaptive system – with googols of variables – to virtually time travel into the future? Who hasn’t wanted their own Zoltar to grant your wishes? If you thought there was even a small chance of creating such a machine, how much would you spend to be the first one to do it?
Algorithmic Fashion
The US textile industry is trying to fight back against the small-package loophole exploited by the Chinese apps Shein and Temu to take over the US fashion industry. The rise of these prolific peddlers of disposable clothing (and other trash, trinkets, and whatever the heck those weird clickbait objects are in all of their ads!) have disrupted industries and brought a windfall of billions of advertising dollars to companies like Meta. Temu is owned by the Chinese public company PDD, while Shein is looking to IPO in the US. Shein reportedly uses algorithms to automatically identify and create hot clothing items based on TikTok trends and the like with little to no human involvement. The online discount apparel trend is impacting the fast-fashion giants and the discount apparel retailers in the US, many of which lack a viable online strategy and certainly aren’t using AI algorithms to instantly produce knockoff goods. Even brands like Gap and others are blaming Shein for cutting into profits. Curiously, the Chinese fashion apps are more popular with Gen X and Boomers. The trade loophole being exploited by Shein and Temu, according to the NYT, is called de minimis and it allows anything under $800 to slip by without tariffs. More typically, a retailer would place a bulk order from overseas (large enough to require paying tariffs), and then break it up to ship to stores or ecommerce customers. Congress identified the potential issues with the law in a 2023 report titled “Fast Fashion and the Uyghur Genocide” (page 2 of this PDF covers the key points). I don’t have a view as to whether Shein and Temu are sustainable business models. I will say that it’s hard to get too excited about companies that need to spend billions of dollars on social networking ads to draw customers and still haven’t created a lawful, legitimate business model (Chinese companies, no doubt led by Shein and Temu, accounted for 10% of Meta’s revenues and 5% of their 16% y/y growth, according to their last quarterly conference call). But, the concept of algorithmic driven fashion and, more broadly, AI-driven retail purchasing is more interesting. TikTok last week announced that any video will be shoppable using AI to identify anything and everything for sale. The speed at which fashion moves from smartphone screen to closet makes your head spin.
Starbucks Snacking: GLP-1 Casualty?
Starbucks is claiming to be the victim of an afternoon snack war as they see fewer visitors at that time of day. Although it appeared to not surface on the company’s earnings call last week, I can’t help but wonder if their real enemy in the afternoon snack war is the GLP-1 class of weight loss drugs? In related news, STAT reports that researchers are looking into whether GLP-1s can also treat depression and anxiety: “Early data and anecdotes suggest that this class of GLP-1 drugs could help patients feel less depressed and anxious. The treatment may also fight the decline in cognitive and executive function that many people with mental health disorders experience, like worsening memory and losing the ability to focus and plan.” Of course, the reason GLP-1's were created was because of the afternoon snack wars…
Miscellaneous Stuff
Fermenting Plastic
Researchers have engineered a bacterial strain to upcycle plastic into a biodegradable “spider silk”, which could have a host of commercial applications. The report is only a preliminary proof-of-principle, but it offers an interesting option for untangling ourselves from our ever-accumulating polyethylene trash.
Stuff About Demographics, the Economy, and Investing
Surprisingly Strong Employment Data
Despite fears of AI-driven layoffs and a seemingly never-ending string of press releases about large corporate downsizings, the macro data imply that we are below trendline, with around 1% discharges in 2023 vs. 1.25-1.5% for the previous decade, according to a BI story. The tech sector is closer to its long-term average rather than above it. The big issue may not be layoffs but a lack of new job creation, which has been anemic for the IT sector (see AI Layoffs). One generation that’s working a lot harder than before is Gen Z. The WaPo reports that teens with jobs climbed from a low of 34% in 2014 to 37% last year. I recall that I and my fellow Gen Xers were all labeled slackers, but that’s just because we were a very small birth cohort – there weren’t enough of us to fill all the afterschool, weekend, and summer jobs to serve the Boomers! It turns out, depending on the year, Gen X had 40-50% working teenagers (similar to Boomers) making the real slackers Millennials and Gen Z, despite the recent uptick. The increased youth workforce is largely functioning as a pressure-relief valve for understaffed businesses like restaurants.
Political Gender Gap
I tend to read reports that classify generalized trends across populations with a great deal of skepticism, but it’s hard not to take notice of recent data purporting a liberal/conservative gap widening between men and women, as, anecdotally, the trend feels somewhat real. According to data from Gallup (as reported by BI), over the last decade, the political gender gap widened from around 5% to over 15%, with women going from 30% to 42% liberal and men staying around 20% liberal. Kurt Vonnegut once said: “Thanks to TV and for the convenience of TV, you can only be one of two kinds of human beings, either a liberal or a conservative”. Now, presumably thanks to social media’s digital tribalism, those polarizing terms are being subject to genderfication.
How to Have Meetings That Don't Suck
In case you missed it, check out our new whitepaper on how we approach meetings at NZS Capital in a way that we actually get value out of them.
✌️-Brad
Disclaimers:
The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC. This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry.
I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.
Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results.
Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.