SPDR Shares closed at $406.77 on September 4 and has been sliding with the metal ever since. With at $4,365.50 an ounce Thursday morning — down $35.30 on the session after reaching an intraday high of $4,434 — the shares are trading near $402.
The reversal happened inside a few hours and it was entirely macro. Gold opened the day back above the $4,400 handle, up roughly 0.5% in early trading, with spot printing near $4,414 and the four-hour chart showing $4,424. Then August producer prices landed at 8:30 a.m. ET, coming in at 5.4% annually against a 5.3% forecast and up from 4.8% in July.
Gold gave back the entire European session. By 10:36 a.m. ET the metal had stabilised at $4,381.32, roughly $53 below the high. December traded $4,415.20, down $45.50 or 1.02%, having been at $4,427.40 earlier for a 0.75% loss.
The mechanism behind the decline is worth stating plainly because it inverts the intuition most GLD holders bring to the fund. A hot inflation print sent a gold ETF lower. That happens because bullion is priced off real yields rather than headline inflation: when a CPI or PPI surprise raises the probability that the Federal Reserve tightens, nominal yields rise faster than inflation expectations, real yields rise, and gold falls.
The benchmark climbed 8 basis points to 4.90%, its highest since November 2023. The recovered from an intraday low of 98.71 to 99.10. Market-implied odds of a hike at the September 15–16 meeting moved to between 62% and 64%.
Silver took the harder hit, with futures down 3.41% at $66.31 an ounce.
GLD carries roughly $149.09 billion in assets under management and an expense ratio of 0.40%. Over the past month the shares have gained 1.96% and over twelve months 19.68%, with NAV returns of 0.69%, 5.74% and 20.50% across one-month, three-month and one-year windows respectively.
The gap between price return and NAV return over a month is a tracking artefact, not a structural problem.
What GLD Actually Holds: 100% Allocated London Good Delivery Bars
The structure matters more for this fund than for almost any other ETF, and it is unusually clean.
GLD is a grantor trust holding physical gold bullion in London vaults. At June 30, 2026, the amount of gold owned by the Trust and held by the Custodians was 32,314,227.7 ounces, 100% of which is allocated gold in the form of London Good Delivery bars. Allocated means specific, serial-numbered bars assigned to the Trust rather than an unsecured claim on a pool. The trustees cannot lend the gold.
NAV is determined using the LBMA Gold Price PM, formerly the London PM Gold Fix, which gives the fund an extremely close relationship to spot. There is no futures roll, no contango drag and no counterparty exposure beyond the custodian.
The creation and redemption mechanism runs through baskets of 100,000 shares exchanged for physical metal. When an authorised participant creates, the Trust receives gold. When it redeems, the Trust delivers gold. That is why GLD flow data is a genuine measure of physical demand rather than a paper claim — every dollar of net creation corresponds to bullion moving into a vault.
Expenses are paid in metal. In the three months ended June 30, 34,843.1 ounces of gold were sold to cover them, following 32,336.2 ounces in the first quarter. That is the mechanical cost of the 0.40% expense ratio: the ounce count per share declines slightly and continuously over time, which is why long-horizon GLD returns lag spot.
The launch history explains the fund’s scale. GLD was the first US-traded gold ETF and the first US-listed ETF backed by a physical asset, and that first-mover position has translated into the deepest secondary-market liquidity in the category.
One structural drawback deserves naming. The IRS treats GLD as a collectible, which means long-term capital gains are taxed at the collectibles rate rather than the standard long-term rate — a material consideration for taxable accounts holding a position with a 19.68% twelve-month gain.
The SEC filings carry the full ledger.
The Holdings Ledger: 33.63 Million Ounces in March, 32.31 Million in June
The tonnage record is the most honest measure of what has been happening in this fund, and it shows a quarter of steady bleeding.
At March 31, 2026, the Trust held 33,634,221.4 ounces with a market value of $154,998,264,698 based on the LBMA Gold Price PM. At June 30, holdings had fallen to 32,314,227.7 ounces with a market value of $130,098,696,246.
That is a decline of 1,319,993.7 ounces — roughly 41 tonnes — across a single quarter, and a $24.9 billion drop in market value driven by both redemptions and the fall in the gold price.
The share activity underneath explains it. In the first quarter, 39,000,000 shares were created across 390 baskets in exchange for 3,584,534.3 ounces, while 46,800,000 shares were redeemed across 468 baskets in exchange for 4,300,883.2 ounces. Net: 78 baskets redeemed, 716,348.9 ounces out.
The second quarter was worse. 18,500,000 shares were created across 185 baskets for 1,698,714.7 ounces, against 32,900,000 shares redeemed across 329 baskets for 3,020,610.7 ounces. Net: 144 baskets redeemed, 1,321,896 ounces out. Creations fell by more than half sequentially while redemptions stayed elevated.
Those two quarters cover the period when gold fell from its January record toward the $4,000 floor established in July. ETF holders sold into the decline, which is what they always do.
Current AUM of $149.09 billion against the $130.1 billion market value at June 30 implies the fund has recovered roughly $19 billion since — a combination of the metal rallying from below $4,000 toward $4,600 in August and creations returning in the third quarter.
The ounce count is the number to watch when the September 30 filing lands. Dollar AUM can rise on revaluation alone. Only rising tonnage confirms new metal entering the vault.
$95.24 Billion of Cost Against $130.1 Billion of Market Value
The cost basis disclosed in the filings is one of the more useful pieces of information in the entire gold complex, and almost nobody uses it.
At June 30, 2026, the Trust’s gold carried a cost of $95,236,108,571 against a market value of $130,098,696,246 — an unrealised gain of approximately $34.86 billion, or 36.6% above cost. At March 31 the figures were $96,197,525,739 of cost against $154,998,264,698 of market value, a 61.1% unrealised gain.
Watch what happened between those two dates. Market value fell 16.1% while cost fell only 1.0%. The Trust redeemed metal at prices well above its blended cost basis, which mechanically raised the average cost of what remained relative to the declining market price.
The implied average cost per ounce at June 30 works out to roughly $2,947. That is the blended entry price of every bar still in the vault, accumulated across two decades of creations.
Two implications follow. First, the aggregate GLD holder base remains substantially profitable even after a 22% drawdown from the January record — there is no systemic underwater position forcing capitulation at the fund level. Second, the marginal holder is a different story entirely.
Approximately 298 tonnes of gold inside the global ETF complex is held at a loss at prices around $4,000, up from 270 tonnes when gold was still above $4,250. Those positions were established during the late-2025 and January-2026 melt-up and they are the supply that meets every rally.
That is the tension inside the flow data. A twenty-year cost basis near $2,947 says the fund is fine. A recent cohort underwater at $4,000 to $5,000 says the next $400 of upside gets sold into.
Which cohort dominates depends on the price. Above $4,474, the recent buyers get their exit and supply arrives. Below $4,000, they capitulate and the long-term base absorbs it.
Flows: $954 Million in Five Days, $6.52 Billion in a Month, Minus $7.44 Billion in Six
The flow picture across different windows tells three completely different stories, which is exactly what a market in transition looks like.
Over the trailing five days, GLD recorded net inflows of $954.32 million. Over one month, $6.52 billion. Over three months, $3.76 billion. Over six months, negative $7.44 billion. Over twelve months, roughly $9.15 billion of net creation.
Reconcile those and the shape becomes clear. The six-month window captures the first-half redemption wave documented in the quarterly filings — 41 tonnes out of the vault in the second quarter alone. The one-month and five-day windows capture the August recovery, when gold ran from below $4,000 in late June to above $4,600 in late August.
A fund that shed $7.44 billion over six months and added $6.52 billion in the most recent one has effectively round-tripped. The twelve-month net of $9.15 billion is positive because the period includes the tail of the 2025 melt-up.
The pattern is the same one visible in every physically-backed gold product this year: heavy redemptions from March through June as the Federal Reserve’s calculus shifted, then a sharp reversal once the metal based near $4,000.
Global data confirms the turn was not GLD-specific. Gold-backed ETFs recorded net outflows of 16 metric tonnes in May 2026 and continued bleeding into the first half of June, before a $1.1 billion inflow snapped four straight weeks of redemptions. July then delivered $3 billion of global inflows, reversing two consecutive months of outflows, with all regions contributing and European-listed funds leading.
Positive flows plus a higher gold price lifted global gold ETF AUM by 1% to $530 billion in July.
For the current session, flows will lag price. Redemptions triggered by Thursday’s reversal from $4,434 to $4,365.50 will show up in next week’s data, not today’s.
The August 17 Billion-Dollar Day and the August 18 Reversal
Two consecutive sessions in mid-August captured everything wrong and everything right about US gold ETF demand this year.
On August 17, GLD took in $1.01 billion of net creations in a single day. That figure was more than fourteen times the entire net inflow North American gold ETFs recorded across the whole of July.
On August 18, GLD recorded $767.8 million of redemptions — reversing roughly three-quarters of the previous session’s inflow inside twenty-four hours.
A two-day swing of that magnitude in the world’s largest gold fund is not allocation. It is a tactical trade being put on and taken off, and it demonstrates that a meaningful share of GLD’s flow is trading capital rather than strategic positioning.
That distinction matters for anyone reading flow headlines. A billion-dollar creation day generates enthusiastic coverage. The redemption the following session generates none, and the net across the two is roughly $242 million — a respectable but unremarkable two-day figure.
The pattern has precedent. GLD logged its biggest daily inflow ever during the 2025 rally, when US-listed gold ETFs pulled in $32.7 billion across the year and global inflows reached $57.1 billion. That year gold surged 42% to a record above $3,700, following a 27% gain the prior year, surpassing its inflation-adjusted 1980 peak and posting its best calendar year since 1979.
The comparison establishes what genuine, sustained demand looks like: $12.9 billion into GLD alone in 2025, against a 2020 record of $15.1 billion in annual inflows. The 2026 figure of $9.15 billion over twelve months is a fraction of that pace and includes the tail of the prior cycle.
