Gold Fell 20% in Q2. The 13Fs Show Who Kept Buying.

Most of the coverage from the August 14 13F deadline focused on the same place it always does: megacap tech. Reuters ran the numbers across 6,371 institutional filers and found something close to a stalemate — roughly 44% of filers trimmed their Magnificent Seven exposure while about 42% added to it. That is not a signal. That is a coin flip dressed up as positioning data.

The more interesting story in the Q2 2026 filings is in a corner nobody was watching, because gold had just done something it hadn't done in years: it fell hard.

The quarter gold investors would rather forget

Gold set a record near $5,600 an ounce in January. By late June it had given back more than 20% — a genuine bear-market drawdown in the asset that spent 2025 as the consensus hedge for every macro worry on the board. Anyone who piled in at the top spent Q2 underwater.

That timing matters for reading 13Fs. The June 30 snapshot captures institutional gold exposure at close to the worst possible moment in the cycle. Every position you see in these filings was marked at the bottom, which means the interesting question isn't "who owns gold" — it's "who was still adding while it was falling."

Gold has since rebounded roughly 10% off that late-June low, breaching $4,400 in August after a cooler-than-expected CPI print pushed rate-hike expectations back out. Futures briefly touched $4,500. Goldman Sachs lifted its target to $4,900 by December. So the funds that added into the drawdown are already being paid — and the ones that capitulated in June sold the low.

Paulson's split decision

John Paulson has been the highest-profile gold bull in the hedge fund world since 2008, which makes his Q2 filing worth reading closely — and it's not a simple story.

He exited Agnico Eagle Mines entirely, selling all 783,561 shares, a position that carried roughly a -5.11% impact on his reported portfolio. That's a meaningful reduction in a name that had been a core large-cap gold miner holding.

At the same time he added to International Tower Hill Mines, buying another 4.9 million shares to take his stake to about 104.5 million shares, worth roughly $208 million. That's not a hedge fund abandoning gold. That's a hedge fund rotating from producing majors into development-stage optionality — a very different bet on the same underlying commodity, and one that only makes sense if you expect the price to be higher, not lower, in a few years.

You'll see this pattern more often than you'd expect once you start comparing quarter-over-quarter position changes on the investor pages: the headline "sold gold" and the actual behavior point in opposite directions.

The miners' fundamentals didn't get the memo

Here's what makes the Q2 selling look more like price-chasing than analysis. While the metal fell 20%, the businesses that dig it up had one of their best quarters on record.

Operating cash flow across the GDX top 25 miners rose 52.5% year over year to roughly $17.2 billion — the third-highest quarterly figure ever recorded. Their combined cash treasuries climbed 72.7% year over year to a record $41.3 billion. These are companies with balance sheets that look nothing like they did in the last gold drawdown, when the sector was over-levered and diluting shareholders to survive.

Meanwhile the structural bid stayed intact: central banks bought an unprecedented 289 tonnes in Q2, with Poland adding 51 tonnes and China adding 33 — Beijing's strongest quarterly purchase since late 2023. Central banks are not tactical traders. They were buying the same drawdown the fast money was selling.

How to actually use this

The lesson isn't "buy gold miners." It's about what 13Fs are good for.

A single filing tells you a fund's position on one day. The signal comes from comparing that snapshot against price action during the quarter — because that's what separates conviction from drift. A fund that held a falling position was probably passive. A fund that added to a falling position made a decision. Those two look identical in a holdings table and completely different in a change table, which is why the flow view is usually more useful than a portfolio list.

The same logic applies to the tech stalemate. When 44% sell and 42% buy, the aggregate tells you nothing — but the specific funds on each side, and what they did during the drawdown rather than at the end of it, tells you plenty. That's also the cleanest way to spot names that multiple credible managers converged on independently, which is what the overlap tool is built for.

Gold spent Q2 as the most hated trade among funds that had loved it six months earlier. The August tape suggests the ones who didn't flinch had the better read. Whether that continues depends on the Fed, and the macro consensus view is where that argument is currently being had.

---

Data sourced from public SEC 13F filings. Educational research only — not investment advice.