Who is actually buying

Strip away commentary and the 2026 gold market reduces to four flows. Central banks bought 244t in Q1 (+3% y/y, above the five-year average) — Poland +31t toward an approved 700t target, the PBOC accelerating from 1t/month to 8t in April, its 17th+ consecutive monthly addition^[1]. Asian investors set records — ETF inflows +84t in Q1, Chinese net imports tripling to 317t, Chinese bar-and-coin demand +67% to an all-time-high 207t^[3][4]. Western funds took profits — a record ~$13bn US ETF outflow in March, May outflows in North America, the CFTC managed-money net long washed out to ~64k contracts, its lowest in recent history^[2][6]. And price-sensitive demand retreated — jewellery −23% y/y^[4]. The composition matters more than the sum: the flows that are structural (official sector, Asian savings) held or accelerated; the flows that are cyclical (Western momentum money) left. That is what a correction inside an intact regime looks like.

The defining flow pattern of 2026 is a geographic handoff. Q1 global ETF inflows slowed to +62t (from +230t a year earlier) only because record Asian inflows were offsetting the largest US monthly outflow ever recorded^[2][5]. The Shanghai premium tells the same story from the physical side: most of late 2025 it sat at a discount of $17–36 (weak Chinese demand at highs); in the first week of January 2026 it flipped positive, and through the correction it has held a modest premium (≈ +$4.5 on June 11), with a six-month range of −$40 to +$167^[7]. Physical gold migrates to whoever bids hardest; right now that remains the East.

For 2025 as a whole the West was the marginal buyer — a record ~$89bn / 801t of ETF inflows^[8]. That money is faster in both directions, and its exit is most of the 24% drawdown. The asymmetry that matters for the forecast: Western money can return on a single catalyst (Fed cuts, equity stress); official-sector money leaves only if the geopolitical arrangement that motivated it reverses. One of those is plausible by 2027. The other would require un-freezing the world.

WGC central-bank figures are model-based estimates that include unreported buying; IMF-reported Q1 flows were only ~16t^[1]. The gap is mostly deliberate opacity (China historically reported in batches years apart), but it means part of the structural-support thesis rests on estimated rather than declared purchases. We flag it because an honest framework should: if the unreported component were materially overstated, the floor under this market is weaker than consensus assumes. The WGC's track record (its 2025 estimate of 863t was within ~2% of SSGA's independent tally^[8]) argues the estimates are sound — but it is an estimate, not a ledger.