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: AI technology platforms are seeing significant gains in revenue per employee; bipedal robots arrive in Amazon warehouses and a robotic quadruped learns to talk; progress in metalenses; squeezing light to see gravitational waves; and more below. 

Stuff about Innovation and Technology
Digital Warehousing; ChatBOT
Amazon is using Agility Robotics’ bipedal humanoid “Digit” in its warehouses to autonomously reposition empty product tote bins. Agility plans to make 10,000 of the robots, although it’s not clear if all of those are for Amazon. Recall that Digit’s backward-bending knees sent me on a journey in #390 to understand why many bipedal robots have that anatomical oddity (forward-bending human-like knees are more versatile, whereas the opposite appears to give more stability, albeit with less range). It’s notable that Amazon is working with an outside company for bipedal bots, and their internal progress in robotics continues to disappoint. I believe that the biggest market by far for technology over the next decade will be the embodiment of LLMs and other forms of AI into a large array of robots and automated machines. Boston Dynamics recently infused their robot dog Spot with ChatGPT, which you can see in this video. 

Artificial Efficiency
Leading AI companies might be seeing an impact on their headcount from applying the technology internally. Let’s start with Google parent Alphabet: with the exception of a short period during 2009’s Great Recession, Google has not had its headcount decline in any quarter on a year-over-year basis until the most recent quarter*. Indeed, since 2009, Google’s headcount has consistently grown around 10-20% per year, with the exception of a pandemic hiring binge that exceeded 20% growth. In June, Google’s headcount declined by ~9,000 people sequentially, but still grew 4.5% y/y. In September, however, their headcount declined by ~2% to 182,000, similar to 2009. And, sequentially, the company only added around 500 net new employees in September over June. The 11% revenue growth reported by Google with a 2% decline in employees resulted in a 13% growth in revenue per employee. Google’s $420k/employee quarterly revenue is the highest I could find in the company’s history, with the exception of the pandemic boom period from June through December 2021 when it peaked at $481k/employee. Speaking on their recent earnings call, Google CFO Ruth Porat noted: “we have engineering work streams underway to improve productivity across Alphabet. Given the magnitude of investment in our technical infrastructure, we have a superb team focused on efficiency of our operations there. We are also making progress in streamlining operations across Alphabet through the use of AI.” Is Google a one-off? I looked at how revenue per employee is progressing over at AI-leader Microsoft, which has been building and deploying OpenAI copilots. Historically, Microsoft has only given annual headcount numbers. Their revenue per employee per year has risen steadily from around $500k/employee in the 1990s to over $800k in the 2010s. In the fiscal year ending June 2023, Microsoft reported 7% revenue growth on approximately zero headcount growth resulting in a record $958k in revenue per employee. And, referencing a 600 bps sequential increase in operating margins at AWS in Q3, Amazon’s CFO noted that it was “primarily driven by increased leverage on our headcount costs.” While I hesitate to declare that early adopters of AI like Google, Microsoft, and Amazon are all seeing strong AI-driven productivity gains given the gyrations of the pandemic, it’s a noteworthy trend that could have large ramifications for corporate margins and productivity across the economy. Such a boost would accelerate the deflationary forces of technology that in part helped keep interest rates low and falling for most of the last few decades. Companies will be reluctant to broadcast that they are replacing people with AI or bots, so one way to keep tabs is via revenue-per-employee or profit-per-employee trends…at least until AI bots gain personhood status! This rise in people displacement is likely to lead to a robot/AI tax on the extra savings accruing to corporate profits. 
(Note: all data for this paragraph come from SEC filings and conference call transcripts.
*In 2013, Google shrunk headcount related to layoffs for manufacturing of Motorola phones; given this was not the core business and was related to an acquisition, I excluded it from my analysis.)

Nanoimprinting Sensors
A couple weeks ago, Canon grabbed a few headlines for their latest nanoimprint semiconductor manufacturing equipment, with the market speculating that the Japanese tech company might catch rivals. While the equipment is capable of making advanced chips down to 2nm, it has some drawbacks, in particular concerning multi-layered chips, which heavily limit its uses. However, there is a different use for nanoimprint that caught my eye, and that’s for making metalenses. Recall what we’ve previously learned (#397) about this new form of sensor:
New metalenses are poised to disrupt some of the image sensor market, and ultimately might find their way into ultra-thin smartphones. Developed at Harvard and commercially produced by Metalenz and chip companies like ST Micro, the new devices offer a host of improved sensor functionality for applications like distance sensing. Importantly, these relatively simple chips have nanostructures capable of detecting not only visible light but also polarization: “Using this technology, we can replace previously large and expensive laboratory equipment with tiny polarization-analysis devices incorporated into smartphones, cars, and even augmented-reality glasses. A smartphone-based polarimeter could let you determine whether a stone in a ring is diamond or glass, whether concrete is cured or needs more time, or whether an expensive hockey stick is worth buying or contains micro cracks. Miniaturized polarimeters could be used to determine whether a bridge’s support beam is at risk of failure, whether a patch on the road is black ice or just wet, or if a patch of green is really a bush or a painted surface being used to hide a tank. These devices could also help enable spoof-proof facial identification, since light reflects off a 2D photo of a person at different angles than a 3D face and from a silicone mask differently than it does from skin. Handheld polarizers could improve remote medical diagnostics—for example, polarization is used in oncology to examine tissue changes.” This type of sensor could be very useful in the coming robot revolution as AI embedded in automatons of all types becomes a reality.
During the recent Canon Expo 2023, the company showcased both its new metalenses and announced that their nanoimprint machines can be used to make the exciting new sensors.

Miscellaneous Stuff
Subsiding Rage
While there continues to be unthinkable tragedies around the US and the world, homicides in the US dropped 6% in 2022, and they could show an even bigger decline (potentially 7-10%, based on FBI crime stats) for 2023. A 10% drop would be the largest y/y decline on record. Violent crime overall also dropped back to 2019’s pre-pandemic level. It’s not all good news, as auto thefts have soared thanks to TikTok videos showing how easy it is to steal some car models. Manufacturers have responded by opening up software upgrade clinics across the US to fix the issue. It’s hard to conceive that in some ways things are improving far more than your social media feed might lead you to believe.

Light Squeeze
Thanks to our quantum universe, particles like photons can randomly appear and disappear from our observable physical realm, creating a background of quantum noise. These (de)materializations don’t impact us at the macro level; however, if you are trying to measure things like gravitational waves, this noise limits instrumentation precision. Researchers working on LIGO, which can detect gravitational waves from phenomena such as black hole collisions, have surpassed the quantum limit that would ordinarily be imposed by this noise. This feat was accomplished by the frequency-dependent “squeezing” of light in tubes around 300 meters long in such a way that measurements can be made accurately across the entire spectrum of interest. The actual methodology is like science fiction come to life: “This is accomplished with the help of specialized crystals that essentially turn one photon into a pair of two entangled (connected) photons with lower energy. The crystals don't directly squeeze light in LIGO's laser beams; rather, they squeeze stray light in the vacuum of the LIGO tubes, and this light interacts with the laser beams to indirectly squeeze the laser light.” One of the new use cases will be to detect neutron star collisions to help determine a more accurate makeup of the celestial participants in these events (which are believed to be the superdense, collapsed remnants of supergiant supernovae) to gain insights into the composition of their black hole cousins and the origins of heavy elements in the universe. 

What to Take, and What to Leave
Speaking of black holes and other hard-to-comprehend ideas, cosmologist Carlo Rovelli has a great essay on how we can learn new things, particularly when we lack the ability to see them clearly, adapted from his upcoming book White Holes.✌️-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.