During the early computer era, the most difficult and expensive problems to solve were in hardware. The profits flowed there.
In recent years, compute was cheap compared to the value provided by quality software - and the software could run on a variety of hardware systems. So software captured most of the value.
There are some new problems that are difficult and expensive to solve that must be solved in hardware. This is making investment and profits flow back to hardware makers. Furthermore the years of commoditization of hardware resulted in consolidation and survival-mode until there were only a handful of serious players in various layers of the manufacturing stack. This will get resolved but will take a long time (decades).
AI businesses constantly need to improve and retrain models. This is extremely compute intensive and time consuming. Therefore we need hardware that is optimized to this specific type of number crunching. This hardware is novel and needs smart and talented people (therefore investment).
As an investor (both of capital and of your labor) you need to analyze where the expensive problems are, with the highest need to be solved, with those that need it most having the highest ability to pay. If you optimize on that spectrum your investment has a better chance of success.
Computer people are drawn to nostalgia, especially considering the entire history of our craft has been meticulously recorded by enthusiasts. We get drawn to certain types of problems and environments and every 10 years the entire world changes. Our profession is one of constant disruption and change.
What makes me rest easy is that the fundamentals have remained constant since the earliest theoreticians. We must use the best algorithms to reduce compute as much as possible. If you do that you win, at the most fundamental level. It’s another way of saying that you must solve the problem as simply as possible (and no simpler).
As an addendum, I don’t fault any talented and ambitious person who goes into the industry where they can make the most money as fast as possible. Some people lament that the best and brightest don’t go into, say, NASA or public school teaching or some such sacrificial act. The bursting of the tech bubble in 2022 and the drying up of VC capital to dime-a-dozen nonsense startups will help drive talented people back to hard problems.