Software, compute and energy. These are all general purpose problem-solving tools and we will never run out of the problems that they solve. Efficiency gains in producing any of them will only increase demand. That is fundamentally different from something like food. The US population can only consume so many calories in a day, but there is no limit to our appetite for software, compute, and energy.
Software can solve essentially any problem that deals with information. Almost any business process, knowledge, or model can be represented in software. No general-purpose SaaS software will ever be specific enough. There will never be enough spreadsheets. We will never stop producing information or creating more use-cases. The backlog is endless. If AI coding tools allow us to complete work five times faster, we'll just put five times more work on the roadmap, and each engineer will create five times more value. No useful piece of software has ever been "complete".
Simple software may execute almost for free, but a lot of useful software does not. This includes AI training and inference, but also simulations, scientific research and more. Bill Gates once said nobody will need more than 640kb of RAM and he was wrong. As Moore forecasted, processing power has doubled around every two years since the 1960s. Your smartphone is a million times more powerful than the Apollo guidance computer that took us to the moon and we still want more.
When DeepSeek showed you could train an AI model with half as many GPUs as previously thought, each GPU just became twice as useful. If we could simulate a whole genome we could run a medical drug trial that would take a decade in hours. Maybe we can invent new economic theories by simulating an entire world. There's always things we want to calculate that we can't yet. We will never have enough compute.
Nearly everything that solves an economic problem consumes energy. This includes compute, but also transportation, air conditioning, sanitation, farming and essentially all forms of technology that make up modern life.
Jevons observed that, after the efficiency of the coal steam engine went up in the 1860s, English coal consumption counterintuitively soared. But it makes sense, because like GPUs, a steam engine that uses half the energy becomes twice as useful. So we want them even more. Our energy use is not limited by the available problems we have to solve with it: the applications of energy are endless.
Even today, we are still consuming so little: the daily energy reaching us from the sun alone could power all our electricity grids for hundreds of years. Energy is the root of nearly all economic value. We will never have enough energy.
And now, AI accelerates all of the above. AI software uses even more compute and energy. Previously software mainly crunched numbers, but with AI it can increasingly also be a calculator for words. Robotics transforms physical-world problems into problems that can be solved with software and compute. If AI starts being able to make useful inventions of its own, we will have even more knowledge and technology which means we'll have even more use-cases for software, compute, and energy.
We will never run out of problems to solve with software, compute, and energy. They are the tools for creating limitless value.
The better we get at making them, the more of them we want.