Gold’s value reflects both physical scarcity and a powerful coordination narrative — a pattern increasingly visible in AI markets.
Last week I was pondering the recent growth in the price of gold and found myself circling a simple question:
How much of gold’s current value comes from physical reality and how much comes from a shared belief we collectively hold?
Gold is an awkward asset for this question. It is physically scarce, costly to extract, and genuinely useful in some industrial contexts. And yet, none of those facts alone seem sufficient to explain why it trades where it does today.
That gap between what something does and what it is worth is where the intersubjective value lives.
A cleaner way to think about intersubjective value
Rather than arguing abstractly about whether value is “socially constructed,” I tried a more concrete approach: benchmark gold against something that shares its physical constraints, but not its narrative weight.
Enter zinc.
Zinc is geologically scarce, energy-intensive to mine, and industrially essential yet largely free of monetary, cultural, or status symbolism. In short: scarce, but boring.
This isn’t to say zinc is story-free. Industrial commodities absolutely embed narratives about construction cycles, manufacturing demand, substitution risk, and macroeconomic conditions.
The difference is which stories matter.
Zinc’s stories are local, technical, substitutable, and mean-reverting.
Gold’s narrative is global, identity-bearing, institutional, and remarkably persistent.
That difference turns out to matter a lot.
The method (no maths required)
The logic runs as follows:
Take gold’s price ten years ago.
Grow it forward at the CAGR of a similarly scarce but purely industrial metal, in this case, zinc.
That gives you a cost-anchored counterfactual: what gold might be worth if intersubjective narrative didn’t compound.
The gap between that baseline and today’s actual price is the intersubjective premium the value created by shared belief, institutional adoption, and expectation rather than physical utility alone.
This isn’t a claim that gold is imaginary or irrational. It’s a way of isolating the effect of belief that compounds over time.
What the comparison reveals
Run this comparison using current prices and a ten-year window and the result is striking:
Roughly 30% of today’s gold price can be explained by baseline, cost-anchored growth.
The remaining 70% reflects excess growth, growth that zinc never experienced.
That excess doesn’t come from new industrial uses of gold.
It doesn’t come from suddenly discovering that gold is rarer than we thought.
It comes from central banks treating gold as a reserve asset, investors using it as a hedge against monetary disorder, centuries of accumulated monetary memory, and reflexive belief: the expectation that others will value it too.
Gold has a physical floor. But its price lives well above it.
Why this isn’t just about gold
Once you see value this way, the pattern appears everywhere.
Bitcoin has almost no physical baseline; its price is overwhelmingly intersubjective.
Luxury goods have a material floor, but most of their value comes from status, signalling, and scarcity theatre.
Fine art’s materials are trivial; price reflects cultural consensus and institutional validation.
And then there’s AI.
AI is full of zincs and golds
Some AI systems are priced like zinc: fraud detection, demand forecasting, optimisation, embedded decision support. They’re bought on outcomes. Narrative barely matters.
Others are priced like gold: frontier models, “autonomous agents,” platform-scale visions, and AGI-adjacent roadmaps.
Their valuations reflect option value, institutional belief, and fear of being left behind far more than current utility.
The closer an AI product gets to operational reality, the more its value collapses toward boring baselines. The more abstract and future-oriented it becomes, the higher the intersubjective premium it can sustain.
Some intersubjective premiums persist for centuries. Others evaporate. Gold’s narrative has endured for millennia. AI’s narrative is barely a decade old.
The real question
The point isn’t whether intersubjective narratives create value they clearly do.
The harder question is when those narratives become durable, and when they collapse into bubbles.
You rarely know in advance. An intersubjective narrative always looks like hype until it proves persistent and once it does, it no longer looks like hype at all.
Hype isn’t irrational, it’s an intersubjective multiplier. Gold shows what happens when narrative compounds over centuries. AI is attempting the same leap: faster, louder, and with far less history.
Knowing which AI products are growing on baseline CAGRs, and which are compounding intersubjective premiums, may matter more than any benchmark that ignores narrative.