Gold’s mid-cycle correction: a reset, not a reversal
Precious metals have given back much of an exceptional advance. The structural drivers behind it are intact, prices have fallen a long way, and gold now accounts for a far smaller share of the world’s financial assets than it once did.
For private wealth investors and allocation professionals
WHERE WE STAND
Gold is still up 31% over twelve months. The retracement is large in absolute terms but unremarkable set against the advance that preceded it, and parabolic moves rarely end in gentle consolidations.
On a closing basis, gold peaked at USD 5,417 an ounce on 28 January and silver at USD 116.70 the same day, both having traded above USD 5,500 and USD 120 intraday during that week. Mining equities peaked a month later, on 27 February. The lows came in mid-July: gold closed at USD 3,976 on 16 July, 27% below its high, and silver at USD 55.52, down 52%. The NYSE Arca Gold BUGS Index reached its own low four days after that, having shed 40%.
The recovery since has been uneven. Gold has regained 10% from its low and remains 19% below its high. Silver has regained 17% and is still 45% below. Mining equities have already recovered 30% and now stand 22% below their February peak. Silver fell the furthest from its high, but it is the miners that are rebounding fastest. All market levels in these pages are as at 14 August.
WHAT CAUSED THE FALL
Three things, and none of them touches what supports gold over the long run.
The first is the turn in the interest rate debate. Kevin Warsh’s appointment to the Federal Reserve coincided with a rise in energy prices, itself caused by the disruption to shipping through the Strait of Hormuz. Expectations shifted from continued easing to the prospect of tightening, real yields climbed and the dollar firmed, which automatically weighs on an asset that pays nothing. We took the same view in February, when the appointment was first read as a change in the underlying case, and nothing since has led us to revise that view.
The second is the unwinding of positioning, which happened all at once. On 30 January gold and silver had their sharpest session since at least the early 1990s, which is as far back as our intraday history reaches. Silver lost 26% in a single day, while gold fell from its high to below USD 5,200 within hours and traded under USD 5,000 the following day. WisdomTree’s analysis of the episode finds little sign of institutional speculative excess, either in futures positioning or in exchange-traded product flows, and points instead to retail and over-the-counter activity as the amplifier.
Silver’s steeper decline over the following months, 52% against gold’s 27% at the low, is part of the same story. Industrial demand played a part, since the Silver Institute expects a fall of roughly 3% this year, with thrifting and substitution reported in photovoltaics on cost grounds. But the two orders of magnitude are not comparable: a few percentage points of industrial consumption on one side, a price cut in half on the other.
The third is profit taking. After eighteen months among the most rewarding the sector has seen in decades, many holders had substantial gains to protect.
None of this is new to the present bull cycle. The sector has paused before, most recently in October 2025, when the low took only nineteen days to form.
THE ALLOCATION QUESTION
Beneath the noise of prices, gold’s place in the world’s savings is shifting far more slowly. Investor-held gold, whether in bars, coins, exchange-traded funds or gold held over the counter, amounts to roughly USD 9 trillion. Against the USD 320 trillion or so invested in financial assets, that is close to 3%. Forty years ago the share was around 14% (World Gold Council).
The ratio describes the market, not any individual portfolio, and the World Gold Council is explicit that it should not be read as a typical allocation. What it does establish is that financial assets have grown far faster than the stock of gold held by investors.
The starting point matters too. The S&P 500 trades at roughly 41 times its average inflation-adjusted earnings of the past decade, the measure known as CAPE. The median since 1881 is 16 to 17, and only eighteen months out of nearly seventeen hundred have been higher, all of them in 1999 and 2000. The multiple on expected earnings, by contrast, stood at 20.9 times at the end of July 2026, against a ten-year average of 19.8 calculated on the same basis. That ten-year average is in any case a weak yardstick. The composition of the index has shifted towards technology, and several of the companies concerned extended the useful lives of their servers between 2022 and 2025, which lifts reported earnings without altering cash flow: Meta puts the effect at USD 2.9 billion for 2025 alone. There is nothing irregular about this. These changes in accounting estimates are disclosed in the accounts, and Amazon shortened its own useful lives the same year. Earnings per share simply do not compare well from one decade to the next. The two measures therefore say different things, one against a decade that was already expensive, the other against a century and a half. An investor weighing equities against real assets today is not starting from neutral ground.
There is a second point about size. The market for gold mining companies is far narrower than the market for the metal itself, so capital returning to it moves prices disproportionately, and re-ratings there tend to be abrupt rather than gradual.
The allocation question is a different matter, and only the investor concerned can answer it, in the light of their own constraints. Exposure to this sector is rarely, as far as we can see, the product of a recent decision: it is usually the legacy of a choice made years ago, under a different monetary regime, and never revisited since. Whether that share is still the one that would be chosen today, whether that means raising it or cutting it, is worth asking. A correction is a good moment to ask it.
FISCAL PRESSURE AND REAL ASSETS
Six months on, the fiscal argument is stronger rather than weaker. The United States is running deficits of a size usually seen only in wartime or deep recession, and the annual interest bill on federal debt, now above USD 1 trillion, ranks among the largest items in the budget. Once debt service absorbs that share of revenue, monetary policy gradually ceases to be independent of debt management.
History offers a short menu. A study by Carmen Reinhart and Belen Sbrancia, published by the International Monetary Fund, sets out five routes out of excessive debt: growth, fiscal adjustment, default or restructuring, a surprise burst of inflation, and financial repression accompanied by steady inflation. The last two work only on debt denominated in domestic currency, and they usually appear together. Between 1945 and 1980, real interest rates were negative about half the time in advanced economies. In the United States and the United Kingdom, the debt liquidated this way averaged 2 to 3.5% of GDP per liquidation year. In our judgement the last of these routes remains the least costly politically, and it erodes both the real value of the debt and the purchasing power of savings. That is where the most durable argument for real assets lies.
The counter-argument deserves a hearing. There is no reason this fiscal pressure has to show up in prices in the near term: governments can defer its effects for years, and a central bank that managed to hold real rates positive would deprive gold of one of its principal supports. That last point needs qualifying. The inverse relationship between the ten-year real yield and the gold price, pronounced for a decade, has loosened since 2022: the yield moved back above 2% and gold rose regardless, supported by official buying and by demand for a hedge, as the World Gold Council notes. The years 1979 to 1981 are a reminder that an aggressive tightening cycle coexisted with historic highs for gold. These two observations do not rest on the same measure: the ten-year yield is a market rate, whereas the lesson of 1979 to 1981 concerns the policy rate set against observed inflation. We read them together as pointing to the same thing, namely that what weighs on gold is not the level of rates in itself but the extent to which monetary policy lags inflation, bearing in mind that how inflation itself is measured remains disputed. That scenario nonetheless looks hard to sustain for long, given an interest bill already above a trillion dollars a year. For an investor, the risk is not so much being wrong as waiting a long time.
SUPPLY THAT CANNOT KEEP PACE
Mine production rose about 2% year on year in the second quarter, to 966 tonnes, helped by new output in Canada and Chile, while recycling fell 6% despite high prices as holders preferred not to sell.
Supply has been structurally unresponsive for a decade. S&P Global’s data shows the rate of major discoveries still declining despite sustained exploration spending, and puts the average interval between discovery and production at around fifteen years. Permitting has lengthened over the same period, and producers have favoured profitability over volume growth.
Higher prices therefore generate little additional output within the horizon of an investment decision. Recycling and the mobilization of existing stock respond faster and can take the edge off a price spike, but they add nothing to the world’s total holdings. On new production, the sector’s ability to respond to higher prices is limited.
DEMAND THAT HAS BROADENED
Total gold demand in the second quarter was flat year on year at 1,269 tonnes, taking the first half 2% higher to roughly 2,522 tonnes, or some USD 380 billion (World Gold Council).
Central banks bought a net 289 tonnes over the quarter, 62% more than a year earlier, with Poland the largest single buyer and the People’s Bank of China adding 33 tonnes, its biggest quarterly addition since the end of 2023. The broader picture is more measured. At 345 tonnes, official demand in the first half was the weakest since 2022, and the World Gold Council expects a slightly slower pace than over the past four years. Those purchases were made during a quarter in which prices fell, though the published data does not allow them to be placed within it.
Asked by the same body what leads them to hold gold, 90% of central banks cite its performance in a crisis, a record for that answer, 84% its role as a long-term store of value and 83% its diversification properties. The geopolitical hedge divides respondents: 85% of emerging market central banks consider it relevant, against 56% of those in advanced economies.
Custody arrangements are worth setting alongside those motives, since gold held abroad is more exposed to the jurisdiction hosting it. The Bank of England remains the most used vaulting location, cited by 57% of respondents, ahead of domestic storage at 49%. Both readings are down on last year, but the proportion of respondents declining to answer rose from 8% to 20% over the same period, so the declared percentages are not a reliable measure of where the metal sits. The answers on changes actually made are more reliable: over the past twelve months, 9% of respondents increased domestic storage and 10% diversified their overseas locations, against 5% and 2% the year before. Official buying serves several motives at once, and reserve managers are changing where they keep the metal faster than these percentages suggest.
Private investment demand has redeployed rather than disappeared. Gold-backed exchange-traded funds saw 45 tonnes of outflows over the quarter, though first-half flows remain slightly positive. Bar and coin demand fell only 3% year on year and remains well above the level recorded in the first half of last year. Over-the-counter buyers also absorbed selling, to the tune of 327 tonnes over the quarter according to the World Gold Council, which attributes part of it to Asia. That figure is less solid than the others, over-the-counter activity being reconstructed rather than directly measured. Lastly, jewellery volumes are falling under the weight of high prices even as the amounts spent continue to rise.
The market therefore rests on several distinct sources of demand rather than a single flow, which makes it markedly less fragile than past cycles dominated by investment demand.
MINING EQUITIES
They took the hardest hit, having entered the correction on multiples already below those of the major listed sectors. The discount is now considerable: the enterprise value of gold mining companies stands at 7.0 times their expected EBITDA over twelve months, against 20.0 times for global technology, 15.8 for industrials and 14.6 for healthcare. On trailing EBITDA the gap widens further, to 7.9 times against 29.1 for technology. A similar gap appears on earnings: the miners trade at 10.7 times expected profits, against 20.9 times for the S&P 500.
That discount sits oddly with the condition of the companies we follow. Their balance sheets are in better shape than through the previous cycle, leverage is low relative to the cash they generate at current prices, and dividends and buybacks have become settled policy rather than an occasional gesture. The multiples do not yet reflect any of this. The net margin of the mining index now exceeds that of the S&P 500, 26.9% against 12.6%, having been far below it ten years ago. This advantage owes much to the level of the gold price, and so to an external factor, with no guarantee that it holds if the metal falls back. Their share prices have yet to catch up with that level of profitability.
Whether the capital discipline these companies show today survives the next leg of the cycle is the open question. The sector’s record at previous peaks is poor: acquisitions struck at the top, capital committed to marginal projects on the assumption that prices would keep climbing. Among the companies we follow, spending is going into brownfield extension and asset quality rather than volume growth, but the real test will come when prices have risen and temptation returns. That is precisely why it pays, in this sector, to choose companies rather than take exposure to the whole.
One objection is often raised to them: miners are equities, and should therefore fall with the market when it corrects. Recent history does not bear that out. Across the six S&P 500 drawdowns of more than 15% since 2007, the NYSE Arca Gold BUGS Index held up better every time. In 2008 it lost 31% while the American index gave up 57%. At the height of the pandemic, 26% against 34%. In 2018 the miners gained 12% while the S&P lost 20%, and in the spring of 2025 they finished the episode flat while the index fell 19%. What follows the trough is far less consistent, and we draw no conclusion from it. One possible explanation lies in where their revenue comes from: a producing mine’s top line is driven mainly by the price of the metal rather than by the economic cycle, so a phase in which gold rises while equities fall can widen its margins at the very moment those of the wider market are compressing.
THE APIS PERSPECTIVE
We read the present phase as a reset within a secular uptrend rather than its end. We would therefore rather pick companies one by one than take exposure to the sector as a whole. The criteria we weigh are asset quality, the jurisdiction in which reserves sit, cost of production, balance sheet strength and management’s record on capital allocation. In an environment of rising metal prices, the choice of operator does more to determine the outcome than the entry price does.
That reading shows in how our strategies are positioned. We added exposure twice, first as gold approached USD 4,400 and again between USD 4,000 and USD 4,100, on both occasions while prices were still falling rather than after the rebound had begun. We build positions in tranches rather than in a single move, which is what makes it possible to add while the trend is still against us. The investment policy of some of our sub-funds allows the use of options, and we have sold put options at lower levels. If prices return there and the options are assigned, our exposure is increased at the price we had set; if they do not, the premium collected pays for the wait.
Since those trades, the macroeconomic picture has developed as we thought it might. The Federal Reserve has so far proved more vocal than active. As for the recent intervention in the yen, we see it as a form of monetization carried out on behalf of a state whose debt leaves it little room. That is our reading of the episode, not its stated purpose. Gold has started rising again, without yet having found the level at which it will settle.
The banks have gone the other way. Every forecast was cut over the summer. Goldman Sachs moved from USD 5,400 to USD 4,900 on 19 June, having removed from its scenario the rate cuts it had expected in 2026. J.P. Morgan made the largest revision, bringing its fourth-quarter average down from around USD 6,000 to USD 4,500 on 3 July, this time on demand rather than rates. HSBC and Deutsche Bank followed. HSBC held its year-end target at USD 4,750 while cutting its annual average, and Deutsche Bank reiterated USD 4,600 in early August. None of these forecasts, however, sits below current prices, the lowest of them still above the USD 4,376 of 14 August.
Our own year-end expectation is above all of those figures. If Jackson Hole and the September meeting pass without action, against a backdrop of rising energy costs, we expect gold to trade back above USD 5,000 before the year is out. UBS published a comparable level on 7 August but placed it in the first half of 2027, so the disagreement is about timing rather than destination. We would rather put a figure forward and stand behind it than settle for a direction.
WHAT WE TAKE FROM THIS
The case for holding precious metals over several years rests on four supports: a fiscal trajectory leading more probably to financial repression than to restraint, a supply side unable to respond with new production, demand that has broadened, and a market whose size remains modest against the capital that could enter it. None of those has given way in the past six months. What separates January from today is the price, not the thesis, and on that basis our conviction is firmer than it was at January’s high.
None of which says anything about the coming weeks. Positioning is cleaner than in January, but the path of interest rates remains the variable that could tip the scenario, and a September rate rise, or a de-escalation in the Middle East, could produce another leg down before our year-end view is confirmed or refuted. The World Gold Council’s mid-year assessment describes a market likely to stay broadly within its current range while acknowledging that the conditions for a break are in place. We are confident about the direction, and conditionally so about the level, but we do not claim to know the route the market will take.
What history does suggest is that consolidations within secular uptrends have been the moments when long-term positions were built rather than unwound. In most sectors one can set a buying price and wait patiently for it. Here the mechanics differ: the market is too narrow for the amount of capital that can enter it, and once the rise starts it moves quickly. Whoever waits for confirmation usually finds the move has come and gone without them. That is what sets this sector apart from others, and what explains why positions in it are built during the correction rather than after the recovery. We will return to this after Jackson Hole and the September meeting.
DISCLAIMER
This document has been prepared by APIS Asset Management for information purposes only. It does not constitute an offer to subscribe, a solicitation to invest, or investment advice, and shall not be construed as such. The analyses, opinions and estimates it contains reflect the judgement of their authors as at the date of publication and are subject to change without notice. Past performance is not indicative of future performance. Investments in precious metals and related financial instruments involve risks, including market, volatility, liquidity and capital loss risks. This document does not take into account the particular financial situation, investment objectives or specific needs of any recipient. It is the reader’s responsibility to conduct their own analysis and, where appropriate, to consult their professional advisers before any investment decision. Reproduction or distribution of this document, in whole or in part, is prohibited without the prior authorisation of APIS Asset Management.
SOURCES
• World Gold Council, Gold Demand Trends Q2 2026
• World Gold Council, Gold Mid-Year Outlook 2026
• World Gold Council, Gold Market Primer: market size and structure
• World Gold Council, Central Bank Gold Reserves Survey 2026
• World Gold Council, You asked, we answered: are fiscal concerns driving gold?, June 2025
• WisdomTree, analysis of the volatility episode of 30 January 2026, for intraday levels on 29 and 30 January
• Silver Institute, on industrial silver demand and substitution in photovoltaics
• S&P Global Market Intelligence, on gold discovery rates and discovery-to-production lead times
• S&P 500 Shiller CAPE, data from Robert Shiller, level as at 1 August 2026
• Annual reports on Form 10-K of Meta, Microsoft, Alphabet and Amazon, for the useful lives of servers and the disclosed effects of the changes in estimate
• Bloomberg, internal extractions as at 14 August 2026, including the six S&P 500 drawdowns of more than 15% since 2007. Gold spot (XAU Curncy), silver spot (XAG Curncy), NYSE Arca Gold BUGS (HUI Index) and MSCI World sector indices for information technology (MXWO0IT), health care (MXWO0HC) and industrials (MXWO0IN). Trailing and forward twelve-month EV/EBITDA, forward twelve-month P/E and net margins as at 31 July 2026, the forward figures based on sell-side consensus estimates
• US Congressional Budget Office, for the US deficit and interest expense
• Carmen M. Reinhart and M. Belen Sbrancia, The Liquidation of Government Debt, IMF Working Paper 15/7 (January 2015). The figures quoted come from this revised version, not from the 2011 working paper
• Published gold forecasts: Goldman Sachs (19 June 2026), J.P. Morgan (3 July 2026, fourth-quarter average), HSBC (9 July 2026), Deutsche Bank (3 August 2026), UBS (7 August 2026, first half of 2027)
• Unless otherwise stated, the prices and changes quoted are closing prices, calculated close to close, from a single Bloomberg extraction as at 14 August 2026