Tearing apart the gold ownership transfer

Gold.

CNBC ran the chyron for six straight weeks: "Gold suffers worst quarter since 2013." Bloomberg's take was nearly identical — rate-hike fears under Warsh's hawkish Fed, a stronger dollar, the usual inputs-to-dashboard reading. Gold dropped 14.1% in Q2. Case closed.

Except the dashboard is lying to you about who's actually in the market.

While Western ETFs dumped 45 tonnes in Q2, central banks bought 289 tonnes. That's not a rounding error. That's 6.4 times the volume, moving in the opposite direction — the largest quarterly sovereign purchase on record for any second quarter, executed during gold's steepest drawdown in thirteen years.

The financial press is treating this like a sentiment cycle. Momentum cools, gold dips, wait for the next catalyst. That framework made sense when ETF flows and speculative positioning drove price discovery. It stopped working about four years ago. The marginal buyer of gold has changed. And the new buyer doesn't read your dot plot.

I shorted gold into the 2022 rate-hike cycle myself, expecting the old playbook to hold. It didn't. That was the last time I let the macro dashboard override what the physical flow data was telling me.

Story off. Numbers on.

Central banks bought 1,082 tonnes in 2022. Then 1,037 in 2023. Then 1,045 in 2024. Three consecutive years above a thousand tonnes — absorbing roughly a quarter to a third of total annual mine production before a single jeweler, ETF manager, or coin dealer gets their allocation.

Mine production hit a record 3,672 tonnes in 2025, growing roughly 1%. Sounds like plenty until you subtract that sovereign wall. Recycled supply contributed another 1,404 tonnes — but grew only 3% despite gold rising 65% that year. At $4,000 an ounce, people still aren't melting grandma's jewelry at the rate the models predicted.

The buyers are specific and trackable. Poland: 82 tonnes in 2026 alone, reserves now 632 tonnes, the single largest sovereign buyer on the planet. China's PBOC: a 20-month buying streak, 14.93 tonnes added in June — the biggest single month since 2023. Total PBOC reserves: 2,346 tonnes. India's RBI: 880.52 tonnes as of late June. And a record 45% of surveyed central banks say they plan to buy more.

Here's the flow:

The flow is one-directional. Gold is moving from Western financial products into sovereign vaults that don't lend, don't lease, and don't sell. LBMA London vault holdings stood at 9,534 tonnes at end-July. COMEX registered inventory: 14.19 million ounces. These numbers aren't flashing red yet. But the rate of sovereign absorption means the buffer between comfortable and tight is thinner than the headline inventory suggests.

Every ounce a central bank buys is an ounce removed from the available float — permanently.

Here's the part nobody in the sell-side research notes is modeling.

China's gold reserves sit at 2,346 tonnes. Sounds large. It represents less than 10% of total PBOC foreign exchange reserves. The average for developed-economy central banks runs around 34%. The PBOC doesn't need to match that — but even moving to 15 or 20% would require thousands of additional tonnes accumulated over a decade or more. They bought during Q2's worst drawdown in thirteen years. They will buy at $5,000. They are not trading gold. They are restructuring reserves.

Saudi Arabia. 323 tonnes. 2.6% of total reserves. For a nation sitting on roughly $495 billion that just joined BRICS+, that allocation is absurd. A move to just 5% would require purchases equivalent to the entire projected central bank demand for 2026 — from one buyer. They haven't confirmed anything publicly. But they're at the table. And everyone else at that table is buying.

Seventy-four percent of central banks surveyed expect the US dollar's share of global reserves to decline over the next five years. That's not a fringe view. That's nearly three-quarters of the institutions that actually manage sovereign money telling you — on the record — that they're diversifying. BRICS+ nations now hold 17.4% of global gold reserves, up from 11.2% in 2019. The trend line doesn't wobble.

Managed money is net long 130,766 COMEX gold contracts. Not extreme, but the speculative layer is trading on a completely different thesis than the physical layer. The specs see Warsh, see rate-hike odds, and position accordingly. The central banks see sanctions risk, de-dollarization, and reserve diversification on a multi-decade horizon. Two games on the same board. When one thesis breaks, the other doesn't flinch.

The choreography writes itself.

Here's the sequence. Warsh talks tough, but inflation data softens or growth stumbles. Rate-hike odds compress. The dollar gives back some of its recent strength. Western ETF investors — the same ones who dumped 45 tonnes in Q2 — reverse course. It doesn't take much. A single strong monthly inflow of 30–40 tonnes.

Now you have ETF buying and central bank buying hitting the market simultaneously. Mine supply grew 1% last year. It can't respond. New mining projects take seven to ten years to permit and build. Recycling barely budged at $4,000 gold — it won't surge at $4,500 either. The supply side is geologically inelastic. I've heard that word used loosely before. In gold mining, it's literal.

Gold at $4,432 this morning. Up 10.5% in the last month. Up 31.8% year-over-year. The Q2 drawdown is already being absorbed. China imported 820 tonnes in H1 2026 — nearly double the year-earlier pace. The machines see a bounce. The sovereigns see exactly what they've been seeing for four years: an undervalued reserve asset in a world where the dollar's share of global reserves has declined 18 percentage points in a decade.

Where does the capital go? Not into the broad gold ETFs — those are the products central banks are quietly draining from the other side. The edge, if there is one, sits in mid-cap gold producers with unhedged production, low all-in sustaining costs, and reserves in stable jurisdictions. Companies running AISC of $1,200 to $1,600 per ounce while spot sits above $4,400. That margin — roughly $2,800 to $3,200 per ounce — is among the widest in the industry's history. Most of the sell side is still modeling "conservative" price assumptions of $2,500 to $3,000. It's like pricing fire insurance using last decade's rainfall.

Financial layer says overextended. Sovereign layer says just getting started. I know which one has the bigger balance sheet. And I know which one doesn't panic-sell in June.