If there was decision point to inform career and investment choices I would recommend understanding contemporary exponential trends or technology trends. By exponential trends I am mostly referring to changes generally produced by new technology be it the size of a user base, production/service costs, or something else. Making the right decision here could be the difference between starting your career at BlockBuster or Netflix in 2002. Or investing in Sears rather than Amazon in 2002.
Technology innovation is often a precursor for major historical events that remake social and economic modes of organization. Imagine how different the world would look with the combustion engine, gunpowder, steel, computers, printing press, and internet. The introduction of a new technology alone can reshape economic futures. Whaling used to be a major market because blubber was used for lamps, but that entire profession has disappeared because of the use of petroleum. Then again entire industries around oil lamps disappeared because of the introduction of electricity and the incandescent light bulb. Even still incandescent light bulb manufacturers faced economic peril if they didn’t eventually migrate to manufacturing more energy efficient bulbs like LED. Each innovation created and destroyed fortunes, though at each transition more value was created than lost as the cost structure improved.
When invention first ushers in the beginning of disruption, technology adoption tends to happen more slowly than people expect. The birth of the internet is debatable depending on how we define it, but let’s say 1990. Even in 2022, only 60% of the world has internet access globally. I think a good rule of thumb today is about 30 years to go from inception to mass adoption (maybe 60-80%).
Not shown in this graph are technology disruption cycles. Print advertising like newspapers and magazines has been disrupted for online advertising like Facebook and Google. For print advertisers, the years 2000 – 2010 erased the prior 50 years of revenue gains. For each technology adoption cycle there is some complimentary technology disruption cycle, as each new technology is ultimately replacing some less efficient solution. For example, Uber’s rise came at the expense of Taxis and Airbnb’s rise came at the expense of hotels.
I don’t have good data on the length of technology disruption cycles but I’ll estimate 10-20 years. Disruption cycles tend to be shorter than adoption cycles because adoption cycles tend to create new markets, in addition to disrupting existing ones.
There are many active technology adoption cycles ongoing today. You are doubtlessly part of this either directly by using some emerging consumer technology (e.g. online dating apps) or indirectly by prioritizing cost/performance in your purchase decisions (e.g. buying a phone with a long battery life). Exponential trends tend to precede widespread technology adoption or act as precursors to other forms of technology adoption. The increasing ability to send more and more data over the internet (bandwidth), observable in the 90s, was a harbinger for the eventual spread of high-bandwidth internet applications such as streaming video and video conferencing.
So if we want to identify future technology adoption cycles, we can just identify existing exponential trends. Then we can reason about the where these exponential trends might lead us.
A non-exhaustive list of exponential trends today:
Moore’s law
I mention Moore’s law because it’s the most well-known example of an exponential trend, even though it’s mostly no longer true today. It concerns the number of transistors that can be placed on a computer chip known as a CPU. More transistors = more compute. For decades computers were getting so fast that old computers were almost immediately obsoleted in compute power. This pattern eventually tapered off due to physical limits. More transistors = more friction = more heat. If the computer chips get too hot, the transistors get damaged. So heat dissipation prevented making computers faster by just making transistors smaller and stuffing them into the same chip.
Nielsen’s law
Nielsen’s law concerns the bandwidth, or the rate of data transmission, in an average consumer internet connection. With low bandwidth it might only make sense to send text, so applications like email. With more bandwidth you might be able to download/upload images and maybe short video (social media). With more bandwidth still you could stream long-form video (youtube, netflix, and Zoom). With even more bandwidth you could stream high-bandwidth video in real time (think VR). This progression is predictable and it will end when physical limits are hit in how much data can be streamed over fiber optic cables, but we are nowhere near this limit.
Huang’s law
GPUs are computing devices that are really efficient at parallelization. There are a few applications for such kinds of compute: gaming, deep learning, and scientific computing. Of particular importance is deep learning, which you might know as artificial intelligence. As GPUs get exponentially more efficient at a fixed price point, the cost to train a deep learning model decreases exponentially. That is, artificial intelligence training is largely GPU-compute bound. Unless physical limits are hit soon, you can expect artificial intelligence to become dirt cheap.
Energy density of Lithium Ion Batteries
Li-Ion batteries are what powers your smart phone and your laptop. They’re the predominant battery type for consumer electronics and electric vehicles. Higher energy density means smaller batteries. Smaller batteries means lighter batteries, or if you prefer, longer battery life with a constant size. Unless this trend his physical limits soon, we can expect such things as:
- phones you charge once a week
- cars you charge once a week
- consumer drones that can fly for hours at a time (currently flight time is up to 20 min)
- potentially electric planes
Decrease in storage costs
HDD and SSD are two types of computer storage drives. HDD is older and uses a spinning magnetic disk. SSD is newer and has no moving parts. As storage becomes cheaper, it means we can afford to store more things for the same cost. You might have used this trend to predict the rise of cloud storage (think of saving your phone’s images on Apple’s servers). You could predict the rise in telemetry (monitoring some data point such as temperature) and consequently the rise in BigData (analytics on the vast quantities of historical telemetry/data). As this trend continues, you can expect to see longer historical record keeping, such as car dashcams never deleting data. And consequently this high-quality data can be used for other purposes such as training self-driving cars. This trend is analogous to the storage trends that followed the decreasing cost of paper.
Growth in money supply from central banking
Today’s national currency comes from central banks, like the Federal Reserve. Central banking is a relatively new phenomena in human history, only being around for the last 100 years or so. With central banking, an institution actually has the ability to print money. As history proves, governments with the ability to print money tend to operate their economies on credit and are rarely stingy. There are many interpretations you can make from an exponentially increasing money supply, and one of them is that the exchange rate for less inflationary assets like stock or gold will increase in nominal values.
Cost to send a kg into space
Space is a pretty expensive place to send materials. The main thing we send into space today are satellites (think GPS or weather satellites). As costs decrease, we can send more material. That might mean bigger, more sophisticated satellites. Or it could mean a new form of travel even faster than air travel. The longest flight in the world is New York to Singapore at 18 hours 30 minutes. With a rocket this could be done in less than an hour. It’s technologically possible today, just too expensive for anyone to do. That is, until it isn’t. Cheap transport of materials into space opens up new opportunities for high-precision manufacturing (dust in the air can result in impurities) and research I can only imagine about.
Cost of DNA sequencing
Large classes of disease are hereditary. A DNA test could identify genetic diseases immediately, or responses to certain drug classes, or propensity to develop certain diseases later in life (e.g. diabetes, cardiovascular, addiction). But before we can even get to that point, there needs to be more genomic research, and that’s only possible with more data (DNA sequencing patients with known diseases). Both both genomics research and clinical practice, cheap DNA sequencing is needed. DNA sequencing is currently around the $600 price range, which is still not economical at scale. That said, these falling costs could be used to predict the rise of companies like 23andMe which do partial DNA sequencing.
Final words
As I said this is a non-exhaustive list of trends and I’ve only chosen some of my favorite examples. Even to one trend there are often many dimensions. Li-Ion batteries aren’t just becoming more energy dense, they’re also becoming cheaper with better performance characteristics like recharging speed. There are many important trends I’ve probably missed that should have made my list too. And these trends need not be purely about technology. You could just as well make a chart with fentanyl overdoses and predict its emergence as a national issue years ahead of schedule. These trends need not be exponential, either, but exponential trends naturally tend to have the largest impact.
There were a huge number of historical exponential trends I could identify even with cursory research. Imagine the decreasing cost of rail transport, or decreasing accident rate from elevators / airplanes and how they enabled skycrapers / air travel, or the decreasing costs of lighting. There are simply too many to list and it would only be scratching the surface. The point being that these trends are everywhere, all the time. Especially the moment you read this sentence. The exact exponential trends will change but the cycle of technology adoption changing economic and social modes of organization will continue. Identifying technology adoption cycles ahead of schedule by monitoring exponential trends can be huge beneficial for fun and profit.