The price of anything is, in the long run, determined by what it costs to produce it. This is not a contentious claim. It is how markets work. When the cost of production falls, prices follow. When the cost of production rises, they follow in the other direction. What is unusual about the current moment is that these two forces are operating simultaneously, on different categories of goods, at a speed that is visible in real time.
Understanding which category a thing belongs to is the most useful piece of analysis a thoughtful buyer can do right now.
The collapsing side
The cost of cognitive work has fallen dramatically in the last three years and is continuing to fall. Legal research that once required a paralegal for a day can now be done in seconds. Market analysis, code review, first drafts of commercial writing, financial modelling: these are not tasks that have been eliminated, but their cost of production has dropped by factors of ten or more. The effect on the people who were paid to do them is visible in employment data. The effect on the price of services that depend on them is just beginning to appear.
The same compression is coming for physical labour, with a lag. Industrial robotics is improving on a cost curve that resembles the one software followed a generation earlier. When it arrives at sufficient capability, the cost of manufacturing goods, sorting parcels, and assembling components will follow the same downward trajectory as cognitive work. This is a reasonable estimate of where we are headed, not a certainty, but the direction is well-signposted.
In a world where intelligence and labour are both cheap, most things produced by those inputs get cheaper. Software services. Professional advice. Manufactured consumer goods. Standard electronics. If it can be designed by an AI and assembled by a robot, the long-run price will reflect that.
The exception
The inputs themselves do not follow this pattern. Copper cannot be synthesised. A mine takes 15 to 17 years from discovery to production. The ore grades at existing mines are declining, meaning more rock must be processed per tonne of refined metal. At the same time, demand for copper is rising at a structural rate driven by AI data centres and electrification: banks forecast the largest supply deficit in 22 years for 2026 alone. Intelligence cannot produce more copper. Robots cannot mine deposits that do not exist.
The same logic applies across a range of materials and processes. A hand-forged axe head requires steel, heat, hammering, and the physical judgment of a person who knows their craft. The steel is not getting cheaper because logistics software improved. The heat is not getting cheaper as long as energy is priced by the commodity. The craft does not scale the way a software service does.
Heirloom seed genetics represent 10,000 years of selective cultivation. That accumulated work has no digital substitute. The seed library cannot be regenerated by a language model. The genetic diversity it encodes took centuries to develop and is actively being lost. The things that cannot be reproduced by intelligence or cheap labour are, as a category, repricing – or they will be, as the gap between the two sides of the curve becomes obvious to more buyers.
When to buy
The counterintuitive answer is: now, before the repricing is complete.
This is not a claim about timing the market or predicting an exact inflection point. It is a simpler observation: the goods in question are currently priced as if the production economics of the old regime persist. A home battery system from EcoFlow or Bluetti is manufactured today at input costs that reflect today’s copper and steel prices. A hand axe from Gransfors Bruks is priced at today’s skilled labour and Swedish steel rates. Both categories depend on finite inputs whose cost trajectory is upward.
When the supply deficit in copper widens. When the generation of toolmakers who learned their craft before the offshoring era retires. When energy infrastructure investment drives electricity prices higher in the short term. The products made from these inputs will cost more. The gap between what they cost today and what they will cost in five years is the opportunity this publication is built around.
None of this is a reason to buy things you do not need. It is a reason to buy things you do need – or will need – sooner rather than later, and to buy the version that lasts rather than the version that is slightly cheaper now and worthless in three years.
The window in which these things are priced as if the old economy still functions is not going to stay open indefinitely. That is the case for buying now.
Finite Resources earns a commission on purchases made through links on this site. This does not affect our editorial independence or the prices you pay.