When Gold Stops Being a Commodity: Notes on an Interview About the Rising Price of Diversification

A long underground vault corridor, gold bars stacked into the shadows on both sides, a single shaft of light coming through a half-open door at the far end

Jade comes from Yushi, gold from the Ru and Han rivers, pearls from the Red Wilds… The former kings measured their use by their weight, and so made pearls and jade the highest currency, gold the middle currency, and knife-coins and cloth the lowest.

— Guanzi, “Guo Xu” (Warring States period; my own translation)

Two thousand years ago the point was already this: gold is currency not because of what you can do with it, but because it is far away, scarce, and was measured into that role. Its position was assigned. Positions change.

What this episode is about

The Meb Faber Show, 4 September 2026. The guest is Inigo Fraser-Jenkins, chief investment strategist at AllianceBernstein. The topic is asset allocation, but the spine of the whole hour is one idea: the way we are used to building portfolios rests on a set of conditions that will not renew themselves automatically.

The line I rewound was this one — gold is no longer a commodity; in this kind of environment, gold is money.

The main points

1. US equity exceptionalism and dollar exceptionalism are two different questions. He defends the first and declines to defend the second. For equities his reasons are checkable: US firms have a long record of exploiting IT better than rivals elsewhere; US firms can cut headcount faster (he notes this reason sounds less hopeful); and the US working-age population is projected roughly flat over the next ten to fifteen years, against about half a percent annual decline in Europe and close to one percent in China. On the dollar the ledger is more balanced — debt service crossing above defence spending in the budget is, he thinks, a big deal, but there is no alternative to the dollar, and stablecoins have created a material new buyer of short-dated Treasuries.

2. He reclassifies gold. The reason isn’t the price. It’s geopolitics and the outstanding debt positions of major economies — in that setting gold no longer functions like copper or oil, it functions like money. So he keeps gold in its own bucket, separate from commodities.

A dashed empty slot is left in the commodities box on the left as a block labeled gold moves into the money box on the right, with a horizontal arrow between the two boxes.

3. He gives no price target for gold, and admits how odd that is. He jokes in client meetings that here he is, a strategist, telling people to own an asset he refuses to price. Gold used to be priceable off TIPS. That relationship broke on the day Russia invaded Ukraine, and he doesn’t think it comes back. His substitute is a long-run return assumption: roughly 0.6% real per annum over 150 years, plus BRICS buying — call it 1% real. No income, zero correlation to equities, about 1% real. Those three numbers are the entire starting point for his position sizing.

4. Bonds may not come back as the diversifier. Twenty-odd years of negative stock-bond correlation made 60/40 a no-brainer. Extend the chart back two hundred years and positive correlation is the norm. The post-2022 experience isn’t the anomaly; it’s the return. His reason is a higher equilibrium inflation rate — deglobalisation, high starting debt levels and the pull toward debt monetisation, and climate. These push inflation up without pushing growth up.

A long timeline where most stretches sit above the zero line in the same-direction zone, with only the short span from 1998 to 2021 below it in the opposite-direction zone, returning above after 2022.

5. A chart called “false sense of security from markets that worked.” Rank equity markets by market cap in 1899: the US did spectacularly, the UK reasonably well, and the next six or seven went to zero, some more than once. He isn’t forecasting that for the US. He’s pointing out that the historical numbers we use to estimate long-run passive index returns are computed from the survivors.

6. He does the AI productivity arithmetic backwards. Humans have a poor record forecasting long-run productivity — the 2000 bubble produced permanent-uplift assumptions that had to be unwound. So instead he estimates what the other forces subtract. Demographics alone put US working-age population growth about 0.8 percentage points per year below the post-1980 rate; add a few tens of basis points for climate and you get roughly 1%. The question becomes: AI needs to add 1% a year just to keep us running at the same pace. His yardstick is the steam engine, which raised UK productivity by about 0.8% per annum in a sustained way. The average of the recent academic forecasts for AI is about 1%. So: AI probably does raise productivity, and the central case is that it holds our speed rather than increasing it.

Of eight bars only the leftmost two have height, the other six shrunk to marks on the ground, with a dashed frame above showing where they once stood.

7. Why this round of automation may differ. Overlay “sectors where AI is expected to make the biggest difference” with unionisation rates by sector, and the shape looks starkly unlike heavy industry or autos over the last twenty or thirty years — with his own industry at the top of the list. He expects some level of job dislocation near term.

Two downward bars below the zero line for demographics and climate subtract one percentage point in total, and one upward bar on the right adds it back to the zero line, ending flat.

Going further

”I’m diversified, so why does everything fall together?”

Probably the most common question of the past few years. You hold US equities, some international, some bonds, a few ETFs. It looks diversified. Then 2022 happens and it all falls together.

Listening to this, I think the problem is that we treat “owning different things” as diversification, when the actual unit is “driven by different forces.” Stocks and bonds offset each other because the dominant shocks were demand shocks — demand weakens, equities fall, central banks cut, bonds rally. That was a set of conditions that happened to fit, not a property of the bond as an asset. The shock sources he names next — deglobalisation, constrained supply chains, geopolitics — are supply-side. Supply shocks hurt growth and raise inflation at the same time, and in that world stocks and bonds move together.

In the demand shock panel on the left the stock arrow points down and the bond arrow up, offsetting each other; in the supply shock panel on the right both arrows point down.

So when I want to check whether I’m actually diversified, counting names doesn’t help much. The useful question is: would these things break on the same day? His test for gold is worth stealing — it has no cash flows and no real use, so there’s no reason to expect its correlation to equities to be anything other than zero. The reasoning matters more than the conclusion: whether something should move with your core position depends on what it earns from, not what it’s called.

He also explains why gold sold off hard in the first half of this year — retail and CTA flows last year pushed gold and equities into positive correlation, which was abnormal, and then unwound. That’s its own reminder: the same asset can temporarily lose its function while the flow structure around it is distorted.

”It’s already run this far — am I buying the top?”

Gold, AI, anything that has run. I get stuck here too.

What’s interesting is that his ordering is the reverse of the intuitive one. We tend to judge the price first and then decide whether to own it. He decides what job the thing does in the portfolio first, then talks about price. He says it plainly: his gold view is not a standalone view. He likes gold because he wants a strategic overweight in equities, and the industry is struggling to say what diversifies that equity position now that bonds don’t. Gold is one of the things that goes into that empty slot.

The ordering matters because it determines what should change your mind. If the reason you own gold is “it goes up,” then when it falls you have no reason left to hold — only willpower. If the reason is “it doesn’t move with my equities,” then the thing that should change your mind isn’t the price, it’s the correlation. The first has a floating failure condition. The second has a testable one.

On the left a rising and falling line with a dashed threshold trailing beneath it that it never touches; on the right the same line crossed once by a fixed horizontal line, with the crossing marked by a dot.

His stance on valuation strikes me as honest. He doesn’t dismiss it — on a Shiller PE basis the market looks fully valued, and equities aren’t alone, most assets look expensive versus history. But valuation alone is dangerous: used against history it would have kept you out of stocks for years, which would have been a terrible call. The role he ends up giving it is narrow — valuation doesn’t time your entries, it forbids you from assuming further multiple expansion. In his framing, that’s the whole job.

A bar in three segments; the first two, earnings growth and dividends, are solid, while the last, multiple expansion, is drawn in a dashed frame and crossed out.

“Is AI going to change everything?”

The backwards arithmetic is the part of this episode I’d most want to keep.

Most AI discussion asks what AI will add. He asks first what everything else subtracts — 0.8 points a year from demographics, some tens of basis points from climate. Once you have that floor, “AI adds 1% a year” means something else entirely. That isn’t acceleration. That’s holding level.

The habit underneath it: we judge something new by comparing it to now, when the right comparison is the future without it. And that future is often sloping downward. Same number, different baseline, and “takeoff” becomes “break-even.”

On jobs, I’d copy his method too. He states the counter-case fully first: across two hundred years of automation scares, more jobs were created than destroyed. That’s a record, not reassurance. Only then does he offer specific evidence that this time may differ — not a feeling, but a shape you see when you lay two charts on top of each other. Saying “this time is different” carries no weight by itself. The weight is in naming which variable changed.

Where to look next

The episode itself (The Meb Faber Show #649, 4 September 2026); the show posts notes for each conversation
Inigo Fraser-Jenkins publishes through AllianceBernstein’s website, and says most pieces also go up on LinkedIn
For the 1899 markets that went to zero, search “survivorship bias equity markets 1899” — the underlying dataset is documented in the academic literature
Robert Shiller’s public data page at Yale hosts the long-run CAPE series for free

One thing to take with you

One idea: ask what job a thing does for you before you ask what it’s worth.

If you can’t state the job, a low price won’t help — you won’t know when to let go. If you can state it, you get a checkpoint for free: is it still doing that job?

Something I’ve tried, nothing to do with investing, and you could do it tonight. Write down the three things that make you feel steady this month — a job, a person, a routine. Then write one line beside them: could these three break on the same night?

The first time I did it, all three lines had the same answer.

Three boxes on the left labeled job, person, and habit send three lines converging into one circle on the right holding the same single source.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.