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: Welcome back to another periodic edition of Stuff I've been thinking about. Today, I dive into the growth of local delivery and how it might be having a “Prime” moment; the shift to agentic commerce could redraw the lines in ecommerce, but it’s more likely to make the current winners even bigger; smart homes are finally getting smart thanks to AI; the rise of AI search is increasing consumer usage and cementing a familiar winner; the shift to “world building” in video games marks a big change for the industry, but the cost may be prohibitive near term; Moravec’s paradox fades as robots learn human behaviors and movements at a rapid pace; LLMs’ game theory; and a story about storytelling as the stock market shifts to LLM agents. Much more below...
Stuff:
Local Logistics’ “Prime” Moment? And, the De-Powerlawing of Restaurants
Despite mixed consumer spending and a backdrop of tariffs and hiring uncertainty, Uber, DoorDash, and Instacart all experienced accelerating growth in Q2 (vs. Q1). Let's run through some numbers: DoorDash’s revenue growth increased from 21% in Q1 to 25% in Q2 y/y and orders jumped from 18% in Q1 to 20% in Q2 y/y; the company also reported significant growth in DashPass membership. Similarly, Uber’s Delivery quarterly revenue growth increased from 22% to 23%, and the company reported 36M Uber One subscribers (in other Uber news: the average Waymo gives more daily rides than 99% of Uber’s drivers!). Meanwhile, Instacart saw quarterly revenue growth increase from 9% to 11% y/y, orders accelerate from 14% to 17% y/y, and ongoing growth in Instacart+ membership. Separately, Walmart reported 26% US ecommerce growth, led by double-digit increases in store-fulfilled pickup and delivery; they are now offering <3 hour delivery to 93% of US households. Stepping back to look at the broader economic trends, as consumer spending faced off with inflation and ongoing headline news/political uncertainty, the strength of local food/merchandise delivery suggests a shift is taking place in consumer preference for leveraging local delivery logistics. This behavioral change reminds me of when Amazon’s Prime membership hit a positive growth spiral, with the resulting network effects ultimately allowing Amazon to become a last-mile delivery powerhouse – taking on FedEx and UPS while soaking up the market share of consumer wallets. In order to be a true “Prime moment”, the trend would need to be widespread, and it’s not yet clear if today’s local delivery growth is indicative of broad adoption by consumers, or if it’s simply driven by more affluent households and the habits of younger consumers. The value analogy is not dissimilar to Prime – busy consumers willing to pay a little extra to save time in traffic and at the store, not to mention gas money (I explored many of these topics in a 2019 whitepaper).
Another interesting phenomenon is sluggish sales at some large restaurant chains, which is perhaps being driven by expansion of delivery choices. The pandemic-era digital leaders like Chipotle are perhaps now facing more competition from the plethora of options available to consumers within local delivery apps. These circumstances all bode well for advertising spend as businesses jockey to win the buy button in the mobile food/merch apps. Adding up the total marketplace revenues for Instacart, DoorDash, and Uber’s Delivery yields $54B in Q2. If I haircut that with a guess on the percent of Uber’s US delivery business, the number might be around $45B, or $180B annualized. US spend on restaurants and groceries in 2024 was $2.6T, so the big three US delivery companies (of which the majority of their business is currently food) are already devouring 7% share of spending, and when you include other services like Walmart and Amazon’s same-day grocery, I suspect the figure is well over 10% share for food alone, and certainly a rising share for general merchandise as well (we DoorDashed a laptop to our house last week!). Amazon, for their part, might be worried about this “Prime” moment in local delivery, given that they recently announced an expansion of same-day grocery delivery from 1,000 cities to 2,300 by year’s end. While this phase of expansion for local delivery might be causing some de-powerlawing of restaurants (and perhaps some marginal ecommerce share loss at Amazon), ultimately, network effects will redraw the lines and likely cement even stronger power laws. The big will get bigger, and the companies that are most savvy at leveraging advertising and meeting consumer demand/needs will drive a new phase of creative destruction in retail and restaurants. At the same time, the long tail of local small businesses can benefit from increased consumer demand for delivery if they are digitally savvy. Consolidation could become part of the process as well. Uber already has a partnership with Instacart, and the founders of Lyft just completed the 7-year planned sunsetting of their voting rights and board seats, fueling market speculation.
Agentic Shopping Wars
Speaking of consumer spending habits, another hot topic lately is the shift to agentic commerce, whereby autonomous agents search for products and complete purchases on your behalf. The details are murky for how consumer adoption of AI shoppers will take place, and the big players that hold the data and control the checkout are getting ready for battle. The main contenders in the US are Amazon, Shopify, Walmart, Google, and OpenAI. Retail expert Jason Goldberg recently covered some of their opening salvos. For example, Goldberg informs us that Shopify recently updated its terms to require a human in the purchase loop; they are reportedly collaborating with OpenAI to allow humans to shop from ChatGPT (as well as working on their own agents). Amazon recently blocked Google Shopping (the standalone service separate from Google Search) from scraping its data and (at least temporarily) stopped advertising in Google Shopping (although Amazon continues to dominate advertising in Google Search, including AI search results). Amazon is focused on their Rufus AI assistant, and plans to allow you to use their shopping agents to buy from other retailers (good luck getting that permission, Amazon!). Walmart has also developed an extensive agent strategy. As Goldberg concludes, we may end up with a small number of giant walled gardens. Ultimately, AI may cement today’s big retail platforms as the gatekeepers of an increasing share of all commerce. It seems that, in all facets of retail, the big will continue to get bigger for the foreseeable future.
Closed-Captioned Security Footage
Amazon-owned camera and security system company Ring recently rolled out AI security video descriptions. The experience is great from a user perspective as you get a near-real-time description of what’s happening, such as “A person is delivering a package from a brown UPS truck” or “A white dog is walking on the driveway”, etc. The speed and accuracy of the description suggest to me that Ring may be running a compact AI model on the cameras themselves, an impressive feat that is achieved likely by constraining the universe to a small number of options such as cars, colors, animals, and people. If the processing is cloud-side, that’s equally impressive for the round-trip speed of analysis. Google is also rolling out a similar feature for their popular Nest cameras in the Google Home AI Labs program. I suspect we will see a proliferation of parameter-constrained AI models running in a distributed manner, greatly increasing the usefulness of connected devices. Fifteen years since the first Nest thermostat kicked off a resurgence in the idea of a “connected home”, it seems like we will finally see the vision playing out for consumers. In other Amazon AI news, the company has bought the always-listening wearable company Bee, which tracks your daily movements and records every conversation you have so that it can serve up a summary for you. If you could add a simple picture to that, you could effectively go back and live inside any memory, leaving me to wonder whether this intrusive nuisance is a bug or a buzz-worthy feature.
Digital Discovery
This article offers a good review of the myriad challenges businesses have faced for discovery by consumers over the course of the digital era. From the original 10 blue links of early search, to smartphones with mobile browsing, to mobile apps, to social network feeds, to voice search, to ever-changing Google algorithm demands, to short-form video, to podcasts, etc., and, now, LLMs. Effectively, AI is just another evolution in a long series of similar challenges where there has only ever been one thing that matters: if you create good, relevant content/products, you will eventually be discovered. One interesting nugget in the review article suggests that OpenAI may have shifted to Google Search from Bing for its in-house AI agents fulfilling ChatGPT conversations. That’s an interesting idea, and perhaps suggests a deeper partnership than OpenAI’s recent agreement to leverage Google’s Cloud (see the prior edition of SITALWeek for more). I might speculate that such a switch could even portend something similar to the all-important TAC deal that Google signed, which funnels tens of billions a year to Apple (and is under government scrutiny). One thing that seems to be happening is that AI is vastly improving the search experience (e.g., if you use Google Search), so users have many more reasons to reach for their phone and ask questions. Indeed, Google Lens, which lets users search by photos or images, grew 70% y/y in Q2. Google’s position of power in search combined with their leading-edge efficiency in AI could also be the reason Meta recently signed a $10B cloud deal with Google. Silicon Valley is always open for some good old co-opetition, but this remarkable use of Google by competitors is notable.
World-Building Game Changer
Former Nexon CEO and gaming industry veteran, Owen Mahoney, discusses why bigger may not be better in video games. The shift from studio to AI-based world generation requires a profound rewrite of how the industry operates. For example, Google recently introduced the impressive Genie 3, which can rapidly create entire immersive, navigable, promptable world models. Contrast this new gaming reality to enterprise software, wherein each new platform tends to create new companies without damaging existing ones. I think any game under development now, especially ones that have been in development for years (I’m looking at you GTA!) are likely to be superseded by games within AI-built worlds. For the moment, the gaming giants may be saved by the computational expense of these models delaying their commercialization until we have more breakthroughs in AI efficiency over the next few years.
Here is an excerpt from Owen:
In this environment, do big companies still benefit from their size? Is size a net benefit? Increasingly, the answer is No. The remaining cohort of large publishers built industrial processes around people instead of software, and systems that demonstrably don’t scale. If success means sustainable, positive returns on investment, then many of these companies have simply become too big to succeed.
Now those same legacy giants are running headlong into AI—a shift that threatens what little advantage they have left: the unique ability to produce massive volumes of specialized content. We’ve already been seeing the cracks for years, with some of the cheapest-to-produce games (Rocket League, Rimworld, Grow A Garden, R.E.P.O) garnering more dollars and/or more hours than the most expensive-to-produce games.
But it gets worse. The common view of AI is as a cost-saving tool. But like past technology revolutions, its greatest impact will likely be a redefinition of what a game is. The old genres won’t be the new genres. And the old development processes won’t apply to the new media form. As with the GPU, the Internet, and mobile, AI will usher in entirely new types of games. And once again, the incumbents will be left scrambling.
This kind of shift requires real reorganization, not just headcount cuts. It means experimenting, retooling, and redefining how games get made. And it demands hard conversations with boards and shareholders about how to invest, take risks, fail, and try again. Well-capitalized incumbents may try to acquire their way out, but those deals rarely deliver returns for shareholders. Stockpiles of cash and brand equity won’t matter if they can’t overcome their diseconomies of scale.
Resolving Moravec’s Paradox?
I enjoyed this densely packed interview with Google DeepMind CEO Demis Hassabis. In particular, I was delighted to hear that Demis is a fan of storm chasers on YouTube! (See Twister, the Livestream from #436.) Another interesting tidbit is that Veo3, Google’s wildly impressive video-generating AI, learned physics from watching YouTube videos. The intrigue is that AI is able to abstract structure in the universe from observation, which to me likely portends some interesting fundamental cosmological discoveries on the horizon. Historically, it’s generally thought that the “thinking” part of AI was going to be easy relative to the “doing” part out in the real world. This idea is captured in Moravec’s paradox – that millions of years of perfecting human motor functions, where the brain has to control a host of physical aspects for interacting with a diverse and changing environment, would be a far more challenging task (vs. logical reasoning) to accomplish with pure technology. The speed at which AI models can learn real world physics, and our ability to apply that learning to robots, is likely to turn this paradox on its head. The interview with Hassabis also left me feeling more confident in my view of the following probable trajectories for AI development: 1) query-based, conversational, agentic AI is likely being underestimated in both its magnitude and speed of impact; 2) while AI’s impact on media and gaming is likely to be far more mind blowing, its reality is also farther into the future given the costly complexity of immersive video vs. text-based agents (e.g., the current Google API charges $0.75 per second of Veo3 video with audio); and 3) at some point, we will indeed have general-purpose agentic AI operating on robotic form factors, but, there too, cost will be prohibitive on the hardware side for quite some time.
Goldilocks of Game Theory
One of the reasons human civilization has, against all odds, advanced to the current stage is our mastery of game theory. We may not always consciously recognize it, but reciprocity tends to be the value-maximizing, species-elongating strategy for interactions. We geek out on game theory at NZS Capital (NZS = non-zero sum, a game theory concept whereby each participant leaves better off than if they had not transacted with each other). It turns out that LLMs also utilize various degrees of game theory strategy. One study found Google to be more ruthless, OpenAI to be too cooperative, and Anthropic to be the best at reciprocity. If I were to pick one winning LLM based on its ability to mimic humans’ maximization of transactional outcomes, I’d have to go with Anthropic’s Claude based on this paper. It’s interesting that LLM-based robots might not only adopt human skills much faster than anticipated (see above), but that these embodied agents may quickly find value-maximizing game theory strategies to improve the longevity of their own species.
Another Hindenburg Readying for Flight
You can't inflate a really great bubble until the frivolous, risk-seeking lending comes to the table. From the FT: “JLL estimates $170bn of assets will require construction lending or permanent financing this year. Between now and 2029, however, global spending on data centres will hit almost $3tn, according to Morgan Stanley analysts. Of that, just $1.4tn is forecast to come from capital expenditure by Big Tech groups, leaving a mammoth $1.5tn of financing required from investors and developers.” Indeed, debt usage is already on the rise for data centers, doubling this year to $60B and tallying up to over a trillion dollars in the next five years. For example, the California utility PG&E doubled its pipeline for data center electricity demand in the last five months to 10 gigawatts, and Pennsylvania's PPL has plans to power as much as 14 GW worth of data centers (up 32% from three months ago) over the next decade. Here’s to hoping that revenue flows follow all those investments. This type of capex cycle may end up reaching farther into the economy than we might currently fathom – its labyrinthine tendrils only to be discovered upon catastrophic failure. Another factor often present in brewing bubbles: special-purpose vehicles. These opaque ponzi-like schemes are on the rise as investors of all types are looking for access to private AI companies. Buyer beware.
Storytime
Investing is a form of storytelling. CEOs spin tales about their companies and try to rally the workforce to manifest them over a long time horizon. Investors decide if they too believe the stories or not. Most of the time, the stories are fiction, fantasy, or even fairy tales. Occasionally, visionary entrepreneurs pen a nonfiction, or even a compelling fiction that turns out to be so predictive of the future that it serves as prior art for reshaping reality (think of the Steve Jobs Reality Distortion Field!). There are also stories about economics, politics, and the world at large that influence the stories about companies and investments. Investors create their own stories about businesses as well, and the resulting investment ideas can end up in either a canonized history book or a throwaway dime novel. Even trying to unravel the truth of past stories can be fraught, as hindsight is only as good as the incomplete and unreliable human narratives on which history is based.
Our job as active investors is to attempt to identify the objective pieces of a story and then determine to what degree an investment might be pricing in those truths. At NZS Capital, we developed a framework based on complex adaptive systems a little over a decade ago to try to inoculate ourselves from falling for too many believable fantasies. To briefly summarize part of the framework, we build two portfolios in one – one that’s firmly anchored in reality, and one that lets us dream a bit more, lest we miss out on the next big reality-shaping narratives. Our “resilient” portfolio is populated by stories that we have a high degree of confidence are (or will be) nonfiction. And, our “optionality” portfolio is populated with stories we are less certain (but still optimistic) of being rooted in reality. Over time (and plot twists), some of these optionality stories graduate to nonfictional, resilient positions, some remain hopeful long shots, and some turn out to be pure fantasy. Our job is to match position sizes to the underlying, evolving truths of each story. The taller the tale, the smaller the position. Now, there is a point to this, er, story that I am telling you, which I will arrive at momentarily. The daily volume in the stock market is dominated by forces beyond us lowly active investors. Indeed, most of the volume is reflexive signals acting on signals, perpetrated by high-frequency trading, quantitative strategies, retail investor memes, flows in and out of passive funds and ETFs, stranger-than-fiction news headlines, etc. In contrast, when I started in this profession 27 years ago, the stories told by CEOs and investors about companies and their stock prospects constituted the vast majority of signal in the market. Today, it’s not clear how much, if any, impact investors’ stories have on the daily prices of stocks. And, in some cases, it appears to me companies are losing complete control of their own narratives as well. To be clear, I am not complaining about the current state of affairs; indeed, I am grateful for the opportunities provided by the enhanced volatility and transactional volume, even if it means that our own opinions are no longer the “signal” in the market. With the rising noise level in general (see the last issue), the vast cloud of stories virally zipping around the world seem to have impacted the narratives of most investors, perhaps to a point of absurd reflexivity. And, now, we have something very different happening: all of that volume in the market, previously programmed in some form or another by humans guiding machine learning algorithms (or retail investor brains programmed by social media news cycles, etc.), is slowly being taken over by LLMs and agentic AI. I suspect autonomous AI trader bots are writing their own signal algorithms and creating their own stories. They are telling those stories to each other and executing trades. We can see clues that this shift is happening in a recent study that found meaningful drops in trading activity during ChatGPT outages. I think that tidbit of information gives us, well, the rest of the story as to what will soon define the stock market on a day-to-day basis (if it’s not already the dominant force, which I suspect it is). This agentic investing evolution will create even more noise and less signal in the daily price of any given stock. Again, this turn of events spells good news for us active investors who still think we can find stories that, with any luck, will turn out to be superior nonfictional investments.
NZS Capital published our Q2 2025 update letter in July.
✌️-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.