I’m Optimistic: My Thesis on Startups, Venture & Technology — Part 1
This is the first post in a series. You can find my other later posts here:
- I’m Optimistic: Extending my Thesis to Early-Stage Venture — Part 2.0
- I’m Optimistic: Extending my Thesis to Kanata’s Strategy — Part 2.1
On March 10, 2023 Silicon Valley Bank collapsed and the startup and venture industry thought we were going to experience our own version of Lehman Brothers and the 2008 GFC. I started working at a multi-family office in April 2008, shortly after JP Morgan acquired Bear Sterns for $2 a share, so I wasn’t excited about living through Lehman Brothers 2.0.
Thankfully the startup and venture industry didn’t experience any financial contagion. Reflecting now just over a year later, on the surface it’s not difficult to find reasons to be pessimistic or concerned:
- For most of 2023 and early 2024 risk-off has become the status-quo position for most venture GPs and LPs;
- the world is experiencing increased persistent geopolitical volatility and supply chain disruptions;
- restrictive monetary policy, high cost of capital, and revaluing of venture portfolios, all of which makes any venture investor’s job much harder for the foreseeable future, and
- the new technology paradigm of Generative AI being the one bright corner of venture, feels very hyped up while investors and entrepreneurs are stumbling to figure out where value will accrue.
Why I’m Optimistic About Technology, Startups & Venture
For 20 years edge.org used to ask a group of intellectuals an annual question. In 2013, that question was “What Should We Be Worried About?“. I’ve always loved Craig Venter’s response (quoted below), and for the last +10 years it’s stuck with me. So, as an Entrepreneur, a Technologist and an Investor I have to be an optimist, it’s in my DNA. Otherwise I’d be in the wrong profession and probably drive myself into some kind of Nietzschean style madness. Regardless of which of the three hats I’m wearing, it’s not my job to obsess over the endless list of things that could go wrong. My job is to squint and ask “How do we make this go right? And what does that outcome look like?”
“What — Me Worry?”
As a scientist, an optimist, an atheist and an alpha male I don’t worry. As a scientist I explore and seek understanding of the world (s) around me and in me. As an optimist I wake up each morning with a new start on all my endeavors with hope and excitement. As an atheist I know I only have the time between my birth and my death to accomplish something meaningful. As an alpha male I believe I can and do work to solve problems and change the world.
Reason for Optimism #1: Economic Reset & Start of a New Economic Cycle
1.1: Reset in Venture Market Activity
Starting in 2019 through to the second half of 2022 it seems like all asset prices became disconnected from reality. Prices of everything from crypto, to used cars, to early-stage start-ups just kept going up rapidly and arbitrarily. Since late 2022 to today we’ve been experiencing far more sustainable market dynamics, specifically regarding valuation exceptions and speed of due diligence.
Chart: Range of Pre-seed Pre-money Valuations (US$M) 2014 to Q1 2024
Chart: Range of Seed Pre-money Valuations (US$M) 2014 to Q1 2024
After reflecting on the frothiness and frenzy of 2021 and 2022, I personally feel a reset of founders and investors exceptions has been a healthy revert back to the long-term market mean.
1.2: Start of a New Business Cycle & Long-term Asymmetrical Upside
I first got my first entrance into the start-up world in September 2009 when I was running a student-led angel group during my undergrad at TMU. I was fortunate enough to be at the right place and the right time to ride an 11 year bull market, which is the longest economic expansion since the 1850s, while simultaneously participating in a computing generation (Mobile & SaaS) from almost start to finish.
I have no idea how long the next economic expansion will last, but considering it will most likely be underpinned by Generative AI (Transformer architecture) in the short-term, and any newly invented Neural Network architecture in the long-term, I’m comfortable assuming it will have some durability.
1.3: Some of the Best Venture Funds Were Launched in Similar or Worse Market Conditions as Today
Union Square Ventures started raising their first fund in the summer of 2003, which went on to become one of the best-performing venture capital funds in history, apparently returning +67% annualized.
“It took us a year and a half to raise that fund. We traveled all around the country (and to the UK too which was a total waste of our time and money). We spent our own capital raising that fund. We believed in ourselves and our thesis, but it was very hard to get others to understand it. Sometime in the spring of 2004, someone got it. And then a few others did. And by the summer, we had a group together and we were able to build to a first close in November of 2004. We had our final close in February 2005, eighteen months after we started. That was a hard raise.”
Andreessen Horowitz raised their first fund in 2009 during the GFC, which apparently returned 44% annualized as of 2019. Scott Kupor, Managing Partner at a16z was the first employee and was interviewed on the Origins podcast. He considers the timing of their fund launch to be an asset to the firm’s success:
“The other benefit we had tremendously was timing.”
Historically, the overlapping of a new economic cycle and an a new technology cycle has created some of the best venture fund vintages in history. This is partially because when market conditions are tough there is less competition and more attractive opportunities. There is no substitute for skill and hard work, but good timing and riding a long tailwind up always helps.
Counterpoint or Bear Case
In the spirit of intellectual honesty, I think it’s important to identify counterpoints to my argument where appropriate. I touched on a bear case earlier in this essay: monetary policy has entered a new paradigm, fundamentally increasing the cost of capital over the long term. This means the free money era that fuelled growth rates and inflated valuations during the 2010s are long gone.
Reason for Optimism #2: New Technological Paradigm & Start of a New Technological Cycle
Generative AI has been compared to the start of the internet in the early 1990s. In his recent Shareholder letter, Jamie Dimon compared AI to be as transformative as “the printing press, the steam engine, electricity, computing and the Internet”.
2.1: Is This a New 15 Year Cycle or a New 45 Year Cycle?
General-purpose technologies (different GPTs), are defined as technologies that have the potential to change entire societies and economies. These technologies often act as the “Big-bangs” that initiate an industrial revolution, which historically have lasted between 30 to 60 years in duration.
Table: Carlota Perez’s List of Industrial Revolutions (with my update)
According to Carlota Perez’s work, the most recent Technological Revolution (Age of Information and Telecommunications) started 53 years ago in 1971 when Intel launched their 4004 Microprocessor. It’s important to appreciate that the invention of the Microprocessor was made possible by:
- the invention of the transistor by William Shockley at Bell Labs in 1947, and
- the invention of Information Theory by Claude Shannon also at Bell Labs in 1948.
Both of these technologies needed to gestate for 10 to 20 years (Fairchild Semiconductor was founded in 1957 and Intel was founded in 1968), before igniting the Information Age of the last +50 years. I can’t help but draw similar parallels to the last ~15 years of Generative AI gestation with:
- Nvidia launching CUDA in 2007;
- the AlexNet team popularizing the use of GPUs and Neural Networks for image recognition in 2012, and
- the “Attention Is All You Need” paper in 2017.
The Information Age of the last +50 years can be further broken down into at least three smaller Computing Cycles that have lasted approximately 15 to 20 years.
Table: Key Milestones of the Past Four Computing Cycles
It’s obvious that Generative AI, General Pre-Trained Transformers (GPTs), the Transformer Neural Network architecture, GPUs, and Neural Networks in general have caused a new technology cycle. The most interesting and impossible to answer question: Is AI (collectively all of its definitions) a successor to the ~15 year Mobile & SaaS Computing Cycle? Or is it transformational enough that have we started an entirely new 45 year Industrial Revolution to succeed the Microprocessor?
I don’t have the answer to this question, however already I think there are a couple identifiable themes in this new technological paradigm:
- Transition from deterministic coding of software to emergent learning of Neural Networks. Examples of this are Tesla’s FSD 12 and Covariant’s RFM-1.
- Transition from purely retrieval of existing human created content and data stored on internet servers, to retrieval plus on-demand machine generated content.
- The value of content and data repositioned for some aspect of model training. This should change the way enterprises view the value of their data (example & example), change business models of media companies (example, example & example) and give rise to new data tooling solutions (example).
In either case, I’m excited to see how this next cycle plays out.
“Every aspect of the computer has fundamentally changed, and so everything from networking, to the switching, to the way the computers are designed, to the chips itself, all of the software that sits on top of it and the methodology that pulls it all together. It’s a big deal because it’s a complete reinvention of the computer industry.”
“We’ve got we’re in the beginning of a brand new generation of computing. It hasn’t been reinvented in 60 years.”
2.2: AI Adoption Should Further Improve the Capital Efficiency of Startups
If you started a software start-up in 2004, before the launch of Amazon Web Services, you needed to raise millions of dollars in investment to buy servers to be installed in a colocation data centre before writing any code. Post-2006, the SaaS business model became incredibly capital efficient due to the adoption of 1) cloud service providers, and 2) compounding recurring revenue models.
However, for the last 10 to 15 years, the scarce and expensive resource limiting a startup’s product development and speed to market was technical and product talent. I don’t think ChatGPT code will replace a skilled CTO, but Founders and early teams should be able to leverage AI Agents, co-pilots, etc to hack together an early working version of their product and generate early customer validation on almost no investment.
Counterpoint or Bear Case
Again, I touched on a bear case earlier in this essay: AI feels very hyped up while everyone is stumbling to figure out where value will accrue. I think we are still in the early stages of this AI cycle, but eventually the technology will not be able to live up to expectations, the hype will evaporate and the Trough of Despair will set in.
Amara’s Law: We tend to overestimate the effect of a technology in the short run, and underestimate the effect in the long run.
Reason for Optimism #3: Growth in Technology Versus the Overall Economy
In 2011, Marc Andreessen published an Op Ed in the Wall Street Journal titled “Why Software Is Eating the World”, in which he predicts that:
“More and more major businesses and industries are being run on software and delivered as online services” and“entrepreneurial technology companies that are invading and overturning established industry structures”.
3.1: Growth in Aggregate Total Addressable Market (TAM) for all Technology
The heart of Andreessen’s thesis is that the aggregate Total Addressable Market (TAM) for all technology is bigger than most venture investors were appreciating at the time. This insight informed his firm’s strategy and allowed them to seemingly overpay to win deals.
Marc Andreessen was not the first person to come to this conclusion, Mike Moritz also figured this out in the mid-to-late 90s when he assumed leadership at Sequoia. In fact, the hosts of the Acquired podcast coined the concept the “Mike Moritz Corollary to Moore’s Law”.
Here’s the thesis: as the inputs into the cost of computation drop exponentially, meaning chips get exponentially faster, more energy efficient, and cheaper (Moore’s Law), and the cost of energy goes down, then the general cost of technology drops exponentially as well. As the general cost to build, adopt and deploy technology drops exponentially, technology is able to permeate further into other non-tech industries (for example: manufacturing, construction, automotive, etc.), which increases the Aggregate Total Addressable Market (TAM) for all technology.
Lastly, there is an additional flywheel element as technology is able to permeate further and is adopted into other non-tech industries, those industries become more operationally efficient as well.
As the Aggregate TAM for all technology increases, aggregate revenue of all technology companies should continue to grow, which supports higher valuations and future exit outcomes back to venture investors. David Clark, the CIO of Vencap explains this well:
“You need to do is to compare fund sizes today with the exit sizes in 10 to 15 years because that’s when those companies are ultimately going to become liquid and if you look at how those exit sizes have increased over the last 15 years are you saying to me that you don’t think technology outcomes are going to get bigger over the next 15 years?”
3.2: Tech Versus Non-Tech & The Overall Economy
As a simple example to illustrate that Aggregate Technology TAM is growing and becoming a larger portion of the overall economy, the table below shows annual US GDP and total annual revenue for the six largest US technology companies for years 2019 to 2023 inclusive. The insights I find most striking from this analysis are:
- While US GDP had a nominal growth rate of 27.7% over five years, the total annual revenue for these six companies grew over 83% over the same period.
- The ratio of total annual revenue for these six companies versus US GDP grew from 4.25% to 6.12% over five years.
Disclaimer: by only including only the largest US technology companies, and only five years of data, this analysis has a relatively small sample size and has inherent survivorship bias.
If you assume growth in market capitalization as a relative proxy for growth in revenue and/or growth in TAM (assuming little to no multiple expansion), then these high growth rates for technology is not a recent trend. The table below shows the average annualized public market return for these six companies is approximately 30% across a variety of founding and IPO vintages, over the course of almost 50 years.
Disclaimer: by only including the largest US technology companies this analysis also has inherent survivorship bias.
Altimeter did a similar and more extensive analysis, and they came to the same conclusion. Altimeter’s analysis shows that since 2014 S&P500 technology companies have compounded earnings at 13% annually, versus 6% for non-tech S&P500 companies.
“I would suggest to you is it’s almost certain that Tech’s going to be a larger portion of the global GDP in 10 to 15 years because technology is becoming more important every day in every company’s life every person’s life the second thing I would tell you is that I think that technology in aggregate will continue to out earn non-technology companies”
Closing Thoughts
I left Bay Street in March 2017 to build INFINITI Lab and run Multiplicity full-time. Since then, I’ve co-founded three companies, two of which I run as CEO:
- Multiplicity, an innovation consulting firm servicing corporate innovation teams in multi-national enterprises, and
- Kanata, a venture fund focused on international pre-seed and seed B2B enterprise startups.
For the last 7 years I’ve been betting every aspect of my career and my livelihood on long (buy) technology and innovation. Looking into the future, I’m excited to keep this bet running as I believe the risk/return has become even more attractive:
- I’ve been through one Computing Cycle (Mobile & SaaS) from almost start to finish (2009 to present)
- In my opinion, at this point in time (April 2024) there is 40% to 60% probability that AI (collectively in all of its definitions) is a General-purpose technology and transformative enough to initiate a new Industrial Revolution successive to the Age of Information & Telecommunications started in 1971.
And I think Bill Gurley said it best:
“I just have immense appreciation for disruption, I just think it’s so cool that you can build something out of nowhere and completely change an industry”
