Separating AI Reality From Hype: Lyn Alden
This issue focuses on the topic of artificial intelligence, and provides a framework for separating datacenter AI from portable AI (i.e. robots) from an investment perspective.
Today I’m extremely happy to be bringing you the latest from my friend Lyn Alden.
Lyn’s background bridges the fields of engineering and finance. She holds a bachelor’s degree in electrical engineering and a master’s degree in engineering management, specializing in engineering economics, systems engineering, and financial modeling. Her early career included roles as an electrical engineer and later an engineering lead at the Federal Aviation Administration’s William J. Hughes Technical Center.
Alden has been a passionate investor for years. From 2010 to 2015, she ran her first investing website as a part-time venture, eventually selling it to a larger publishing company. In late 2016, she founded Lyn Alden Investment Strategy, a research firm that grew significantly, leading her to leave engineering management in 2021 to focus on finance full-time.
Now an independent analyst, Alden aims to deliver institutional-level research in clear, accessible language for both professional and retail investors. She also serves as an independent director on the board of Swan.com and is a general partner at the venture capital firm Ego Death Capital. In 2023, she published the best-selling book Broken Money, exploring the history, present, and future of money through a technological lens.
Lyn is an investor I always read and always love to hear from, so I was grateful she gave me permission to share her research with my subscribers. I’m sure you’ll find it as valuable as I do.
AI’s Impact on the Investing Landscape
March 27, 2025
This newsletter issue focuses on the topic of artificial intelligence, and provides a framework for separating datacenter AI from portable AI (i.e. robots) from an investment perspective.
AI: Setting the Stage
Artificial intelligence is increasingly impacting things socially, economically, and in terms of investing, but it can be hard to separate reality from hype.
The number of AI mentions on corporate earnings calls skyrocketed in recent years, and a lot of AI-related startups are pulling in capital with big visions of the future. Startup tech company CEOs by their nature often have an incentive to overpromise or overhype their outcomes, since they need to raise venture capital and have to err on the side of bullishness rather than bearishness to be successful.
It has impacted politics to some degree as well. Andrew Yang, for example, was one of the leading Democratic Party primary candidates in 2016, and he made self-driving cars and universal basic income a center theme in his policy framework. He often described driverless trucks as being a quickly-approaching reality that would impact millions of jobs, and yet as we look at that concern nearly a decade later, there hasn’t been meaningful impact there.
That’s not to dunk on Mr. Yang per se. In our science fiction, we often over-estimate how much tech growth we will have by a given date, especially outside of electronics. The original 1982 Bladerunner adaptation, for example, was set in 2019 and included mainstream flying cars, off-world colonies, and human replicants, which we haven’t achieved yet. Its physical skylines were also way more massive and dramatic than ours are now. But our electronic screens and other electronic devices turned out superior to what were shown in that film. We overshot on the electronic and software side, and undershot on the physical side.
Similarly, this website includes a list of expectations that Elon Musk made about full self-driving from 2014 through 2022. The general theme was that it was always 1-2 years away, at a level that is supposedly safer than humans, for a decade now.
In October 2016, Musk said that Tesla would demonstrate full cross-country automated drive without a human touch, including charging, by the end of 2017. In March 2017 Musk said FSD was about two years away from being safe enough that you could sleep in your car and wake up at your destination. In December 2020 Musk said he was 100% confident that Tesla would have level 5 self driving in 2021, while today the company is still around level 2.
A major complication is that in-field automation is orders of magnitude more difficult than automation in a controlled environment in terms of both mechanical precision and programming/processing complexity. Creating robotic assembly arms in a clean manufacturing facility with technicians on standby to deal with edge cases is a vastly easier problem to solve than creating robots that can go out into the world and deal with all of the countless hazards and unexpected scenarios they can encounter there, and function at or above the level of a human by themselves.
And while I do tend to think that Musk overrepresented expectations for Tesla (TSLA) to raise capital and pull forward car sales, and I have been critical of that approach, I’d be the first to say that I’m a big fan of his other company SpaceX and the work that’s happening over there. SpaceX has tended to meet or surpass various intermediate-term expectations, and is one of today’s sources of technical optimism that inspires engineers the world over.
AI and Energy: the Numbers
An important area that I view as not getting enough discussion is the difference between datacenter AI and portable AI.