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 recently.
Click HERE to SIGN UP for SITALWeek.
In today’s post: Welcome back to another periodic edition of Stuff I've been thinking about. In this edition, I speculate the weighing machine is still alive and functioning as higher rates and widening ranges of outcomes impact many parts of the growth stock world, in particular the topic du jour: software. But first: of that a brief look at the synergy of autonomous and human drivers; simulated world model for AVs; dwindling computer science majors; and a look at how LLM-driven search is unlikely to disintermediate most digital platforms and is more likely to be a win-win.
Mini Stuffs:
Autonomous Efficiency
Uber shared some interesting self-driving rideshare stats (PDF slides 5-10) on their latest earnings call. First, in cities that launch autonomous, the overall use of rideshare experienced growth acceleration, particularly for first-time riders. Autonomous vehicles conduct 30% more rides per day and arrive 25% faster than human drivers, on average. Uber also noted an interesting trend in the variability in rideshare demand, with peak demand during commuting hours and weekend nights surging to many multiples of the base rate. With that highly uneven demand, AVs would need to sit largely unutilized to meet all demand for ridesharing, implying a long runway for synergy between human drivers and AVs.
Waymo’s World!
In other AV news, Waymo has released a new world model based on Google’s Genie 3 foundation. Utilizing data from over 200 million fully autonomous miles driven, the Waymo World Model has run billions of miles of additional simulated complex situations for Waymo Drivers (that’s what Waymo calls their AVs). The Waymo World Model is built upon Genie 3—Google DeepMind's most advanced general-purpose world model that generates photorealistic and interactive 3D environments—and is adapted for the rigors of the driving domain. By leveraging Genie’s immense world knowledge, it can simulate exceedingly rare events—from a tornado to a casual encounter with an elephant—that are almost impossible to capture at scale in reality. The model’s architecture offers high controllability, allowing our engineers to modify simulations with simple language prompts, driving inputs, and scene layouts. Notably, the Waymo World Model generates high-fidelity, multi-sensor outputs that include both camera and lidar data. Hopefully this virtual simulation will ensure the future safety of riders, pedestrians, wildlife, and bodega cats everywhere.
CompSci Quiescence?
Enrollment in computer science programs across the University of California campuses has decreased two years in a row, declining 3% in 2024 and 6% last year. The last time collegiate CS experienced a drop was during the dotcom crash. According to Indeed data, all tech job postings have been declining since 2023; however, jobs mentioning AI are rapidly on the rise.
Agentic Engineers
Chip design company Cadence plans to transform their business model by the end of this decade from licensing design tools to renting “virtual engineers” through AI-enhanced features, according to Paul Cunningham, senior VP and GM who oversees corporate agentic AI at Cadence.
Serving Ads with Intent
Since I wrote about agentic commerce back in August 2025, we’ve seen several developments in the industry. Ecommerce merchant platform Shopify announced UCP (Universal Commerce Protocol), which allows users to checkout within an AI model like Gemini. Google has a number of examples of retailers using its latest AI services, and they recently launched the ability for users to buy directly on Wayfair and Etsy inside Gemini and AI search mode. Search advertisers are seeing an opportunity as Google shifts to a more intent-based ad landscape (as opposed to keyword-based), as that intention is allowing Google to improve the odds of conversion when they present an ad to a user. Looking back at the 25-year history of search as a utility, each subsequent improvement in search technology has increased the ad landscape in a win-win way: consumers get more relevant offers and businesses connect with more interested consumers. Likewise, AI seems to be an accelerant for the ability of search to match user intent with products and services. For example, a user might search “why is my pool water green?” and AI would provide a link to the proper chemical remedy on a spa company’s retail site. You can think of ad costs as a “tax” on advertisers, but the ROI is proven, and the Pareto efficient auctions are optimized at a scale that drives the win-win outcomes. Historically, some investors have maintained that vertical search companies for travel (e.g., Booking or Airbnb), real estate (e.g., Zillow or Apartments.com), and even ecommerce (Amazon, Shopify, etc.) may be threatened by search (or, at the very least, search has been seen as a large tax on their business; even Amazon is rumored to be Google’s largest search ad customer). However, many such vertical platforms provide unique and invaluable interfaces to messy, fragmented markets and may have significant customer loyalty. Rather than being subsumed by the LLMs, these aggregating platforms stand to benefit from more qualified leads (which will likely be the case for organic traffic as well). On the other hand, sites that are spammy or don’t provide value have, in the past, been erased by changes to the Google algorithm (e.g., Google historically demoted low-quality, local business review sites and comparison shopping engines), and we could see a wave of lower value-add digital businesses be absorbed/eradicated by AI search as well. My base case for now, however, is that most existing digital platforms providing a non-zero, network-effect outcome for their constituents will benefit from AI search.
NZS News
NZS Capital published our quarterly update in January, and Jon was interviewed for this recent Bloomberg article on Nvidia.
Stuff:
“Don’t Fight the Fed”...or the Bots
On a day-to-day basis, stock trading prices have little relation to the long-term, intrinsic value of a company. Over time, however, the proverbial “weighing machine” that ascribes merit to free cash flow tends to beat out the fickle “voting machine” of short-term market machinations. Given the recent frantic fits of volatility and AI trading bots extrapolating to the extremes, one might assume that the weighing machine has malfunctioned. However, despite these jitters, I believe we are actually seeing the underlying, historical corrective mechanism working as intended, it’s just perhaps obscured by headline hysteria of AI “haves” and “have-nots”.
What if the force that’s driving some of the market downdraft isn’t a hyperbolic fear of AI, but rather reality finally setting in? Given how top-of-mind software stocks are for investors, let’s examine whether it’s the weighing machine or voting machine that’s driving their worst drawdown since COVID-19 and the 2022 interest rate hikes. Three years ago, in March of 2023, I wrote about the risk of seat-based software disruption from AI in Mentally Reformatting for the AI Age:
Many existing software apps, which are largely sold on a per-seat or usage basis, might need to transition to being sold on a value basis, i.e., factoring in the customer's cost savings for the employees that an AI app replaces...There is growing evidence that adding AI software tools affords ~50% productivity increases, which I would estimate might imply somewhere around 30% fewer seats for those tools, ceteris paribus. Of course, the real-world changeover will be far messier and less direct, and many people will be repurposed to other, higher value tasks. Does the rapid pace of AI development also call into question the decade-long time horizon of VCs? (Or, for that matter, many other types of growth asset investing?)
More recently (about a year ago), I softened my fears for select software companies that may become more valuable in the age of AI as they shift to value-based pricing and capitalize on AI agents’ usage of their systems and data:
In terms of value destruction or creation in cloud platforms and apps, there remain numerous unanswered questions regarding how AI agents will interact with these legacy systems. For example, will AI agents need “seats” in the old tools like Salesforce and Microsoft Office in order to be productive? Will they need the same tools like Okta and antivirus software?...Will the current generation of SaaS apps become a “system of record” for AI agents, with incremental value created by new apps that ride on top of them? Or, will the new AI platforms become the new systems of record, displacing legacy cloud apps? To be determined.
Even if some cloud software companies successfully transform into AI-driven, value-based tools, we cannot ignore the fact that these same companies have largely reached maturation. The fact that the top of their S-curve coincides with the dawn of AI creates a scenario much like that of their predecessors in on-premise software, which faced maturing revenue growth during the rise of cloud-based software as a service (SaaS). Thus, our starting point is a group of maturing businesses that are potentially facing a widening range of outcomes (both good and bad) due to AI. Perhaps the voting machine is throwing the babies out with the bathwater and overreacting; however, underpinning that response is a typical market adjustment that reflects the slowing revenue growth rates and widening terminal value risks for a set of previously higher growth companies.
Maturing growth businesses often have several potential paths toward value creation – provided they are willing to work actively and be adaptable. (There are two visuals I’d recommend glancing at in Redefining Margin of Safety to better understand companies that are nearing the top of their S-curve growth). For example, here are six ways legacy SaaS companies could create shareholder value today while continuing to create value for their customers: 1) As mature companies, they must shift to GAAP accounting, which means replacing most of the sharecount dilution and stock-based compensation with cash. Often, the value of a cash equivalent to employees is at an “accounting” value vs. Black-Scholes calculations of stock comp. 2) Legacy SaaS companies need to recapitalize their balance sheets. Specifically, they need to take on an appropriate amount of debt that reflects sustainable revenues even in the face of disruption risk. Responsible debt should be used to fund buybacks or M&A, which leads me to the third path: 3) The incumbents will need to realize material M&A for expense synergies (as opposed to M&A for products or features, which should be lower priority). 4) Accomplishing the aforementioned three goals often requires C-Suite and/or board-level changes to shift strategies. 5) Companies need to cut costs, e.g., by leveraging AI productivity tools themselves, rightsizing headcount, and dramatically improving margins. 6) Finally, these companies need to prove that AI is more opportunity than risk by modifying their seat-based pricing model to incorporate usage. For example, if a user is running multiple AI agents accessing software on the same login, then that value-creating productivity for customers needs to be value priced by software companies in some form of a “seat”. In sum, while there is hysteria that software will be vibe-coded away by AI, the reality is that SaaS is simply going through the typical transition of a maturing growth industry. This transition naturally entails a widening range of outcomes and will require each company to implement some combination of the six strategies listed above to stack new S-curves in order to maintain relevance and success. Despite its proclivity for turbulence and distortion, AI is likely to ultimately provide more opportunity than risk for many – but not all – legacy SaaS companies. However, it’s an open-ended question as to which platforms will capitalize on the AI transformation, thus leading to the market’s pricing in that uncertainty for the sector as a whole.
Next, let’s zoom out from software stocks to a broader view of growth stocks. With the election of Trump in November 2024, the market began pricing in the probability that rates would finally reverse their record-setting rapid rise (this commentary largely applies to US stocks; other factors around self-sufficiency and geopolitical risks have driven other markets around the world). Indeed, for much of the US’ growth-stock-driven rally since the election, it appears to me that the market was assuming long-term rates would be materially lower to justify higher valuations. At the moment, that expectation seems somewhat misplaced given the Fed’s emphasis on both inflation and the jobs market. Currently, the labor market remains tight in the US despite a meaningful decline in job openings for entry-level and white-collar positions. Further, labor pool growth is weakening daily due to a combination of demographics and immigration policies (the apparent net negative immigration in 2025 is expected to continue in 2026, a phenomenon that hasn’t happened in the US in over 50 years). Demographics in particular are conspiring against the labor pool: Boomers are retiring, and there are fewer new entrants to the workforce due to depressed birth rates (from two decades ago) and a lack of immigrants to make up the difference (the US population aged 24 and under is shrinking from here on out). If labor remains tight, interest rates may be poised to hang steady rather than decline. For growth stocks, where much of the value is ascribed to the distant future of the company, higher-for-longer interest rates weigh on their terminal values due to higher discount rates on future cash flows. Further, the risk premium – or, as we call it at NZS Capital, the range of outcomes – for many growth stocks has risen due to fears concerning AI’s impact on various businesses. Therefore, in spite of (or in addition to) fears of AI, the overall stock market for growth stocks may simply be on a traditional path of pricing in a new reality of higher rates enduring for longer than expected. I could dive into longer-term risks of the US dollar, fears of stability in the US, the ramifications of a shrinking labor pool causing a wider social security funding gap that would drive higher US borrowing needs (not to mention the potential for an AI-agent-caused layoff wave for white collar workers leading to an increased welfare state), and other factors that could hold long-term rates higher, but let’s for the moment ignore those fears, as I don’t think they are necessary to invoke in order to understand today’s stock market for higher growth companies. Simply put, the range of outcomes has widened for many growth stocks. This is good news for active investors, because the market’s “shoot first and ask questions later” approach will be wrong in its assumptions on some of those stocks, thus creating opportunities.
Beyond the traditional growth stocks, many other sectors of the market, e.g., industrials and healthcare, await a return to normalized growth, i.e., the type of growth we were accustomed to before COVID. However, those old growth rates were fueled by extremely low (effectively zero) interest rates and a more stable global economy (not to mention no one was worried about AI a few years ago!). With higher rates, declining globalization, and rising global instability, it’s perhaps not reasonable to assume these sectors will ever return to their prior growth rates. Again, perhaps the market is simply adjusting to the reality of a new normal.
If the above speculated fears are to become our actual experienced future reality, value creation in the market is more likely to come from productivity gains driven by AI rather than from topline growth from low interest rates and economic expansion. Under these circumstances, it’s imperative that management teams make their own weather. Capital allocation, reallocation of resources, efficiency gains (in particular in the area of labor), and focusing on known levers for value creation are critical. Companies cannot wait around for a free-money, global growth landscape to emerge; instead, they need to take action now, predicated on the reality of a tight labor market, higher long-term interest rates, and the existential need to take advantage of the opportunity set provided by AI.
Now, in typical SITALWeek hypocritical fashion, allow me to take the other side, in particular with respect to the US’ current tight labor market and how AI may free up a wave of workers. AI agents appear poised to take over most of rote white-collar work in the short term. Further, they also appear ready to enhance the productivity of more creative, non-rote work in short order. Rather absurd sounding predictions from the head of Microsoft AI (and former co-founder of Google’s DeepMind) put this notion into perspective: "White-collar work, where you’re sitting down at a computer, either being a lawyer or an accountant or a project manager or a marketing person — most of those tasks will be fully automated by an AI within the next 12 to 18 months.” If Mustafa is correct, we are forced to imagine a scenario where Microsoft will soon have virtually no human customers for their core suite of productivity tools. And, if they aren’t able to charge for AI agents’ seats/usage, or if AI agents don’t need the suite of Microsoft tools, it will be hard to imagine Microsoft remaining a going concern for too many more years. Imagine, a multi-trillion-dollar company, one of the five most valuable in the world, relegated to a low-margin infrastructure provider for the AI agents who ripped out the heart of its business. That seems to be the implied base case according to Microsoft’s own head of AI in that FT interview. However, I don’t think we need to be quite that hyperbolic to see the potential for a meaningful headwind to white-collar jobs as AI-driven productivity rises. For example, Goldman Sachs has been working with Anthropic to develop autonomous AI agents aimed at automating a number of roles at the firm. While Goldman points out it’s “premature” to say this could lead to job losses, they may start eliminating third-party providers (indeed, a canary for many white-collar jobs can be seen in the deteriorating jobs market for IT-BPO professionals in India where unemployment has hit a four-year high). I believe the market is currently mistakenly focused on AI agents disrupting software when the real focus should be on the much more detrimental displacement of human workers.
We can understand the necessity of white-collar job eliminations by looking at how much AI is going to cost corporations. Currently underpinning the multi-year trillions of dollars of data center spending is the assumption that AI will generate meaningful consumer and enterprise revenues for the model providers. On the consumer side, we have seen that AI is augmenting the win-win outcome of businesses like search, while also increasing the monetization of social media in the latest earnings reports from Alphabet and Meta. And, on the enterprise side, we can look at Anthropic’s anticipated ~$30B revenue run rate by the end of 2026. If there were $100B in enterprise AI revenues across all of the big AI model makers in two to three years, that would be a meaningful negative impact to the profits of their corporate customers (assuming AI is used solely for productivity and not revenue acceleration). The reality is that the assumptions underpinning the massive data center investments demand many hundreds of billions in revenues from corporations generated by AI agents. To put that number in scale, the Fortune 500 has around $1.9T in annual profit. At the moment, we have a stagnant job market for desk jockeys, but what if that turns into a loss of a million jobs a year, (or many multiples of that) in order to pay for the data center buildout and human workers’ AI agent replacements? All things equal, for every $100B in AI agent or AI API revenue, more than one million workers need to be fired to keep corporate margins steady using an average Western world employee cost per year. Such a wave of layoffs could hit the core spending cohort that drives the economy in the West, wreaking havoc on growth, significantly increasing the burden on the US’ welfare and social security systems, and, dare I say it, even meriting a meaningful reduction in interest rates (at least temporarily…). Unfortunately, that economic depression and rampant unemployment would also significantly increase the US’ debt needs, which, in turn, would ultimately require higher interest rates to lure global investors into buying treasuries, particularly if there were fears of US economic instability. This interest rate paradox is not easy to reconcile.
In this widespread layoff scenario, workers would need to reskill rapidly to growth areas in the job market, such as healthcare labor, which is increasingly in demand to take care of the aging population (which is growing at an even faster pace than expected thanks to the health benefits of GLP-1s). Another potential growth area is skilled labor (e.g., electricians, plumbers, etc.), as it will be decades before robots make any sort of cost-effective progress in those areas. And, as inhumane as it might sound, we will need a lot of construction workers to build the AI data centers that are displacing the white-collar jobs. According to the latest US jobs data, the vast majority of job growth is isolated to construction and healthcare. From accountant and attorney to nurse and plumber? Do we really think that transition could happen smoothly and quickly? Nursing alone is typically a three-year degree.
Thus, what do we make of this thought exercise of two different paths for the economy? Either the labor market remains tight, causing rates to remain high or even go higher, or millions of people that drive the US economy will suddenly face economic ruin at the hands of AI agents. Well, if you’re looking for answers, I will now apologize. I don’t have a strong instinct which economic path is more likely. In the near term, however, I would guess that a tug-of-war between these two tensions is likely to persist for a while. If we somehow land on a Goldilocks-like measured pace for the transition into the AI Age, we may be able to accomplish retraining of desk jockeys into Boomer caretakers without massive economic upheaval. However, the market may continue to anticipate that AI bear bots will do away with the Goldilocks scenario. If so, the market may soon move on from its apoplectic concerns about software stocks to the more serious concern that jobs could dissolve at a rapid rate. If so, the outlook will be far more detrimental than merely pricing in the impact of higher interest rates on growth stocks. We have our way of approaching uncertainty at NZS Capital based on our Complexity Investing framework. Regardless of what framework you use for investing, adjusting position sizes to the ranges of outcomes is critical when there is a widening range of outcomes system wide, both positive and negative.
Back in December 2025, before the market’s AI-driven panic, I wrote the following:
I’ve found it much more productive to view the world through the lens of skeptical optimism. As I’ve noted in the past, neither cynicism nor pessimism have been winning long-term investment strategies. Skepticism is always prudent, but optimism is the only way to make money in the long term. In many ways, it’s nearly impossible to have a successful investment strategy or philosophy that expresses pessimism when the world is optimistic. It does work occasionally, just as a stopped clock is right twice a day. However, by being optimistic when the world is pessimistic, you are more likely to collect the treasure that’s lying around the economic landscape. You don’t even need to find the end of the rainbow. So, now is a good time to prepare to be optimistic for when the world turns pessimistic. It was Buffett who famously said: “be fearful when others are greedy, and be greedy when others are fearful”. I think the first part of Buffett’s aphorism is a tough way to make a living as an investor, given that things are always getting better over time, and the second half sounds a little too cynical to me. So, with a great deal of unmerited hubris, let’s rewrite that quote as: “Be appropriately skeptical when others are optimistic, be optimistic when others are pessimistic, and never be cynical”.
I still believe that optimism is our best path forward. I think AI will bring near-term disruption, and we may find ourselves on that jobless path above, but this will be far outweighed by a wave of scientific discovery and the creation of new ways for the economy to thrive over a longer time horizon, and, dare I say, perhaps even create jobs. In the meantime, active investors should be able to uncover myriad opportunities in the market by identifying those businesses willing to take an active role in ensuring their own future success.
Now for a healthy dose of skepticism: the only thing I know for sure is the odds favor me being wrong more than right on most of what I wrote above (hopefully I am right about that last optimistic bit!). As we know from complex adaptive systems, predicting the future in normal times is virtually impossible; and, in times of disruption, attempting to do so is about as effective as putting all your money on one number on the roulette wheel. As we stand here today, no human on planet Earth knows which way things will play out. The best any of us can do is assess the data objectively as they come in and adapt as needed. I hope that the above exercise gives you a little insight into how I think about market periods such as what we find ourselves in currently, and that it helps you frame the range of probabilities regarding the various scenarios the world faces today.
✌️-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.