China’s August Gold Withdrawals Hit 62.1 Tonnes — the Weakest on Record
Copy link
Two weeks ago gold traded at $4,652 with Managed Money at the 90th percentile of its one-year range and Swap Dealers at the 96th, a double positioning extreme that this analysis flagged as carrying an unusually high probability of violent resolution. The resolution has arrived. By September 2 the LBMA benchmark had fallen to $4,321.79, a decline of roughly 7.1% from the COT-week peak, and the physical data released across the past week explains why the correction found so little support beneath it. The Shanghai Gold Exchange reported August withdrawals of just 62.1 tonnes — down 22.2% month-on-month and 26.9% year-on-year, the weakest August in the entire nine-year reporting series, below even the 2018 baseline, and sitting at the 8.3rd percentile of the trailing twelve months and the 2.8th percentile of the trailing three years. Year-to-date withdrawals stand at 740.3 tonnes, 13.5% below the same point last year. On the same InProved Terminal dashboards, India’s MCX gold inventory has collapsed 5,541 kilograms, or 75.5%, since July to 1,797.42 kg. The two largest physical gold consumers on earth have simultaneously stopped taking delivery, and they did so before the price broke rather than after it. That sequencing is the entire story of this week’s data.
The InProved Terminal’s SGE withdrawal dashboard presents August 2026 against every August since 2018, and the same-month comparison panel makes the deterioration impossible to attribute to seasonality. The 2018 August withdrawal was approximately 185 tonnes; 2019 and 2020 both printed near 108 tonnes; 2021 reached about 150 tonnes; 2022 and 2023 came in near 165 and 160 tonnes respectively; 2024 fell to roughly 102 tonnes; 2025 to about 84 tonnes; and 2026 now sits at 62.1 tonnes. That is a monotonic three-year decline totalling more than 60% from the 2023 level, and this year’s figure is the lowest of the nine observations by a margin of more than 20 tonnes. The percentile gauges quantify what the bar chart shows: 8.3% on a twelve-month basis and 2.8% on a three-year basis. In only 2.8% of the months across the past three years has Shanghai withdrawn less physical gold than it did in August 2026. The delivery ratio of 43.6% adds a further layer — under half of the metal changing hands on the exchange is actually being taken out of the vault system, which is consistent with a market where trading activity continues but physical offtake does not.
Setting this against the premium data that has run through the past three weeks of analysis produces a coherent and uncomfortable picture. The Chinese gold premium was reported at -0.18% on August 21, a six-month low; at -0.02% on August 28; at +0.13% on September 2; and the Shanghai spread table for August 31 showed Au99.99 at +0.04% and Au(T+D) at +0.07%. In other words, the premium has been oscillating within roughly twenty basis points of parity for a full month. A market at parity is a market in equilibrium, and equilibrium at 62.1 tonnes of monthly withdrawal is equilibrium at a demand level 26.9% below last year. The mechanism is straightforward: Chinese vaults are full — SHFE gold stocks were reported at a record 113.658 tonnes a fortnight ago — domestic demand is being satisfied out of existing inventory rather than imports, and there is consequently no bid pulling metal eastward out of London. Through 2024 and the first half of 2025, premiums of +0.5% to +2.0% meant Western paper buying was continuously validated by physical absorption that tightened the London free float. That mechanism has now been switched off for a month, and the 7.1% price decline from $4,652 is what happens when a positioning-driven rally at the 90th percentile meets a physical market that has stopped participating.
The InProved Terminal’s MCXCCL dual-panel monitor tracks Indian exchange-registered gold and silver in the same vault system, on the same axis, over the same window from July 2025 through September 2026, and the divergence between the two series over the past two months is the most instructive single image in this week’s data. Total deliverable gold stands at 1,797.42 kilograms — 57,789 troy ounces — composed of 471.00 kg of Gold Bar, 1,062.20 kg of Gold Mini, 88.02 kg of Gold Ten, 75.57 kg of Gold Guinea, and 100.64 kg of Gold Petal. That total is down 5,541 kg, or 75.5%, from its July level, and the chart shows the collapse plainly: Indian gold inventory built steadily from near zero in mid-2025 to a series of peaks above 7,000 kg through the first half of 2026, held a plateau between roughly 5,000 and 7,500 kg from January through July, and then fell off a cliff in August to the current level near the bottom of its fourteen-month range. Total deliverable silver in the identical system stands at 233,328.80 kg — 7,501,694 ounces — having surged 122,008 kg, or 109.6%, off its August lows.
What makes this divergence analytically valuable is that it controls for almost every variable that normally confounds cross-market comparison. Same country, same exchange, same clearing corporation, same warehouse network, same regulatory and duty environment, same reporting methodology, same two-month window. The only variable that differs is the metal. Indian gold consumption is dominated by jewellery fabrication, which accounts for roughly 75% of annual demand and which is acutely price-sensitive — a fact that explains the collapse, because gold spent August between $4,300 and $4,650 and Indian jewellers do not restock into record pricing. Indian silver demand, by contrast, is increasingly industrial, driven by the country’s solar installation programme, and industrial buyers procure against production schedules rather than price levels. The 75.5% gold drawdown against the 109.6% silver build is therefore a clean measurement of price elasticity, and it tells gold investors something the futures market cannot: at current levels, the world’s second-largest gold consumer has withdrawn from the market entirely, and the world’s largest has cut its offtake to the weakest August on record. Both of Asia’s physical engines are idle at the same time, and neither will restart on sentiment — they will restart on price.
For bullion dealers, the shift from two weeks ago is decisive and it favours patience. The advice at $4,652 with Managed Money at the 90th percentile was to sell into strength and run inventory lean; that call has been vindicated by a 7.1% decline, and the question now is when to reverse it. The physical data argues for waiting rather than catching the falling knife. A Chinese August at the 2.8th percentile of three years and an Indian inventory down 75.5% since July mean the two demand centres that historically arrest gold corrections are not currently bidding, and neither shows any sign of returning at $4,300. Dealers should expect two-way retail flow over the coming weeks — buyback enquiry from customers who bought near the highs, and bargain-hunting enquiry from those who waited — and holding cash capacity to make markets in both directions is worth more than inventory right now. The constructive note is that the correction is doing exactly what a healthy correction does: it is the mechanism by which price falls to the level at which Asian physical demand re-engages, and when Chinese withdrawals turn back up from a 2.8th-percentile base the recovery in offtake will be substantial. A working planning range of $4,150 to $4,550 through the fourth quarter, with restocking staged in tranches below $4,250 rather than committed at any single level, is the disciplined framework.
For conservative investors, nothing in this week’s data damages the long-term case for gold, and it is worth stating that plainly before drawing the tactical conclusion. Central bank accumulation, reserve diversification away from dollar assets, and the fiscal trajectory underpinning the debasement argument all operate on horizons measured in years and are unaffected by a month of soft Shanghai withdrawals. Gold’s year-on-year return of +38.42% recorded in late August confirms the secular thesis has been working. What the physical data does change is the near-term entry calculus, and it changes it favourably for anyone who has been waiting. Two weeks ago this analysis argued against deploying new capital at the 90th percentile of speculative positioning with the Chinese bid absent; the positioning has since unwound meaningfully and the price has come to the buyer. The remaining caution is that the Asian physical bid is still missing, which means the correction may not be finished. The disciplined approach is to begin staged accumulation into the $4,150 to $4,350 zone rather than committing in full, and to treat a recovery in the SGE withdrawal percentile above roughly 25% or a sustained Chinese premium above +0.30% as the confirmation that physical demand has re-engaged. A twelve-month target of $4,900 to $5,150 remains reasonable and sits below the 52-week high of $5,415.17.
For active traders, the correction from $4,652 to $4,321.79 has resolved the double positioning extreme that made gold untradeable a fortnight ago, and the market is now considerably cleaner — but the physical data argues against getting long too early. The key insight is that the buyer of last resort is absent: with Shanghai withdrawing at the 2.8th percentile of three years and MCX inventory down 75.5%, there is no physical bid to absorb a further leg of liquidation, which means downside overshoots remain more likely than the technical picture alone would suggest. Immediate support sits at $4,250, the prior consolidation shelf, with a deeper zone at $4,100 to $4,150 where the July base was built; a weekly close below $4,100 would signal the correction has become a trend change rather than a positioning unwind. On the upside, $4,450 is the first resistance and a sustained close above $4,550 would indicate the unwind is complete. The higher-quality trade remains the relative-value expression rather than the directional one: long silver against gold continues to be supported by the physical divergence documented across both the SGE withdrawal data and the MCX vault data, where silver’s demand indicators are strengthening in precisely the periods gold’s are deteriorating. Traders should size gold longs at half of normal until either the Chinese premium or the SGE withdrawal percentile confirms that Asian physical demand has returned.
Hugo Pascal’s observation about the AU9999 contract hitting a 10-week volume high underscores the increasing significance of physical gold trading on the Shanghai Gold Exchange. This trend not only highlights robust domestic demand in China but also reflects broader shifts in the global gold market toward physical-backed assets.
Latest articles
Tool and strategies modern teams need to help their companies grow.
Invite users to stay updated with exclusive insights and market trends by subscribing to the newsletter.
Start stacking your bullion. Download the InProved app and build your legacy in gold and silver.
+65 8899 3720
support@inproved.com