The bid that broke

Spot gold settled Friday near $4,073 an ounce, up 1.2% on the day, and the bounce flattered an otherwise rough week. This was the fourth consecutive weekly decline. At the lows the metal traded near $4,000 and had handed back essentially all of its 2026 gain, which means a position opened in January is now roughly flat on the year. For the desks that bought kilo bars at $4,700 in May, the drawdown is real money, and the floor talk has shifted from chasing the rally to defending cost basis.

The cause is monetary, not physical. Kevin Warsh's Fed delivered a hawkish pause last week, and futures now price the odds of a December rate hike near 80%, with a September move running around 60%. A central bank that markets believe is more likely to hike than cut is the worst backdrop a non-yielding asset can face. Thursday's PCE inflation print landed in line with expectations, which was enough to pull the dollar and Treasury yields back and let gold rally into the close, but in line does not give the Fed a reason to ease. It simply removes the case for an emergency hike. That distinction is the whole trade right now.

The official bid never left

Under the speculative selling, the structural buyer kept buying. The People's Bank of China added 9.9 tonnes in May, its nineteenth straight month of accumulation, lifting reserves to 2,313 tonnes. Poland added 31 tonnes in the first quarter to reach 582, and Uzbekistan added 25 to reach 416. Central banks took in an estimated 244 tonnes in the first quarter, and the World Gold Council still projects 700 to 900 tonnes for the full year. This is the tell that separates a price correction from a broken thesis. The hands buying gold for reserve diversification do not care about the December meeting. They are accumulating into the dip, and they have been the floor under every pullback this cycle.

It is worth keeping the scale in mind. A 9.9 tonne monthly purchase from Beijing is small against daily futures volume, and it will not by itself reverse a rate-driven selloff. What it provides is a persistent, price-insensitive bid that compounds month after month. Nineteen straight months of it has put a visible floor under every dip since the streak began, and there is no sign the program is ending.

That split, central banks accumulating while leveraged longs liquidate, is exactly what a healthy correction looks like. The weak hands are being shaken out while the strong hands add. It does not tell you the bottom is in, but it tells you the metal is not in structural trouble. The trouble is concentrated in the part of the market that bought on margin expecting cuts that are not coming.

The physical desk reads it differently

On the bench, a lower spot is a mixed blessing. Cheaper metal eases the sticker shock that pushed jewelry buyers toward 10-karat and hollow product through the first half, and a sustained move back toward $4,000 would help the back half of the year at retail. But falling prices also dry up scrap, because nobody wants to sell old gold into a declining market when they believe a rebound is coming. Refiners describe kilo bar premiums as steady and scrap flow as thin, the familiar pattern of a market that is falling without conviction that the bottom has arrived.

The watch trade feels the same pressure from the other direction, which I cover in this week's watch column. Rolex raised retail on gold models an average of 5.0% on June 1, a hike priced when bullion was higher. If spot holds near $4,000 rather than $4,700, the case for the next precious-metal increase weakens.

What to watch

The path from here runs through two more inflation prints before the September meeting. If they come in soft, the September hike odds fall and gold gets a reason to rally. If they run hot, the December hike firms toward a certainty and $4,000 becomes a ceiling rather than a floor. The full cross-asset read is in this week's trade wrap. The number that matters is simple: can gold hold $4,000 with a hawkish Fed in the chair and central banks still buying? That is the question the next two weeks will answer.