Gold Production by Country
Global Rankings and Data

Most analysis of gold production by country treats the tonnage figures as if they represent supply. They do in a geological sense. They do not in an economic sense. A large share of global mine production is spoken for before it leaves the country of origin, absorbed by domestic jewelry demand, central bank accumulation, or in Russia's case, routed through a parallel distribution system of uncertain transparency.

China is the most extreme example. The 380.2 tonnes gets printed at the top of every ranking. What does not get printed alongside it is that China consumed over 900 tonnes of gold in 2024. The jewelry sector alone accounts for a massive share, then bars and coins, industrial use, and the People's Bank of China's 44-tonne addition to reserves that brought official holdings to 2,280 tonnes. The gap between 380 and 900 gets filled by imports. The net direction of flow across China's borders is inward, consistently, every year, by hundreds of tonnes. Shanghai Gold Exchange gold goes in and does not come back out.

China has been in this position for a long time. Even in 2016 when production peaked at about 455 tonnes, domestic consumption already exceeded output by a wide margin. Since then, environmental enforcement campaigns have shut down or consolidated many smaller mining operations, dropping output 16% from the peak. The government evidently considers cleaner mining more important than maintaining the production crown, and the supply gap gets plugged through imports. There is no indication this dynamic is going to reverse.

The implication is straightforward. China's 380 tonnes contributes nothing to the gold available for purchase outside of China. It is entirely consumed internally, and the country reaches into international markets for hundreds of additional tonnes each year on top of that. If you are trying to estimate how much new gold supply the rest of the world can access in a given year, China's output subtracts from the pool rather than adding to it.

Russia presents a different version of the same problem. Output grew about 5% in 2024, driven by Polyus operations in Siberia and the Far East. The gold is real. Where it goes is murkier. After 2022 when Russian refiners lost their spot on the LBMA Good Delivery List, the direct path into London and New York clearing was cut off. Gold bars without a Good Delivery stamp cannot settle on COMEX, cannot be held by physically-backed ETFs, and most Western-aligned central banks will not accept them.

What happened was not a halt in Russian gold flows, but a rerouting. The metal goes to the UAE and Turkey primarily, gets melted down and recast at local refineries, picks up a new refiner's stamp, and re-enters the international system. There is nothing exotic about the method. Gold is gold. Melt it and recast it and whatever information its previous form carried is gone. You cannot look at a bar made in Dubai and determine that the gold originated in the Magadan region. UAE gold export data has been elevated in patterns that people in the bullion trade read as indicative of substantial re-routed Russian material, but the volume attributable specifically to Russia versus other sources cannot be disaggregated from published customs figures. That is by design.

Work through the rest of the ranking. Australia at 284 tonnes is a relatively clean contributor to international supply. Most of it gets exported. Canada similar. The US at 160 is partially offset by domestic consumption, though the net export position shifts around. Ghana's 141 mostly leaves the country. Keep going down the list, subtracting what each country absorbs domestically, and the amount of mine gold available for genuinely free international trade is probably somewhere between 2,000 and 2,500 tonnes per year. Maybe less. The lower bound is hard to pin down because absorption data for smaller producing countries is unreliable or unpublished.

That range, 2,000 to 2,500, is what should be compared against international demand when assessing market balance. Using the 3,661 headline instead overstates available supply by a third to a half. The gap is not a secret, exactly. Analysts who work in the physical gold market are well aware that headline production numbers are not the same as tradeable supply numbers. But most public commentary on gold production uses the headline, and the casual reader of a gold production ranking article comes away with the impression that the world produces 3,661 tonnes and that is how much gold is available. It is not even close.

Ore grade is the variable that determines the ceiling on global gold production over any multi-decade time horizon. It is a slow-moving variable, hard to perceive year to year, easy to ignore when gold prices are rising, and the one that will ultimately matter more than exploration budgets, political stability in mining jurisdictions, or any other factor people tend to focus on.

Global average gold ore grade was about 8 grams per tonne in 1970. In 2023 it was about 1.2. S&P Global, which tracks this in detail, has the decline at 13.4% just since 2010. Those are industry-wide averages that obscure enormous variance between individual mines. Some underground operations still process material above 5 g/t. Plenty of open pits are running at 0.5 to 0.8. The average is what it is, and the direction is down, and it has been down for decades without interruption.

What does this mean in practice for a mine? Take a hypothetical operation processing one million tonnes of ore per year. At 8 g/t, that million tonnes contains 8,000 kilograms of gold, call it about 257,000 ounces. At 1.2 g/t, the same million tonnes contains 1,200 kilograms, about 38,600 ounces. To maintain 257,000 ounces of annual output at 1.2 g/t, the mine needs to process not one million but about 6.7 million tonnes of rock per year. The crusher is the same. The haul trucks are the same. The mill is the same, or rather it is not, because you need a much larger one. The fleet of trucks is not the same either, because you are hauling almost seven times as much material. The tailings facility receiving the processed waste is not the same. The cyanide consumption is not the same. The power bill is not the same. The water draw is not the same. The workforce may or may not be the same, depending on how much automation you can throw at the problem.

The economics of this are not linear. It is not the case that a 50% decline in grade doubles the cost. At low grades the flotation and leaching processes become less efficient, so a higher fraction of the gold in the ore fails to get recovered. Recovery rates that might be 92 or 93 percent at 4 g/t can slip to 80 or lower at very low grades depending on ore mineralogy. You move seven times as much rock, your reagent consumption per ounce is up, your energy consumption per ounce is up, and you are recovering a smaller percentage of the gold that is nominally there. It compounds.

Gold prices can, for extended periods, overwhelm this problem. At $2,500 an ounce, ore at 1 g/t with an 80% recovery rate yields revenue of about $64 per tonne of ore processed. That can cover costs at many operations. At $1,100, the same ore yields about $28 per tonne, which does not cover costs at most operations. Between 2013 and 2015 gold went from around $1,800 to below $1,100 and the result was a rash of mine closures, asset impairments, dividend cuts, and distressed sales at companies that had been getting by on price support with thin grade margins. Today's global cost base is higher than 2013's because grades have fallen further in the intervening decade, which means the gold price level at which a comparable wave of closures would be triggered is higher too. Nobody knows exactly where that trigger price is. It is somewhere. If gold were to spend a year below $1,500, a lot of mines currently shown as producing on the ranking table would stop producing.

S&P Global published data on the gap between mined head grades (what actually gets dug up and processed) and reserve grades (the average of everything still in the ground). In 2017 this gap was 0.24 g/t. In 2023 it was 0.05. This is a subtle metric that means something specific: in 2017, mines were preferentially extracting the richer material in their ore bodies, running head grades well above the reserve average. This is called high grading and it is a rational short-term strategy when you need to make a production target or keep costs per ounce under control. The problem is that it degrades what is left. Every time you pull the richest ore out first, the average of the remaining ore body drops. By 2023 the gap was almost zero, meaning mines were no longer selectively targeting better material. There was no better material left to target. They were working through their ore bodies evenly, grinding whatever the mine plan delivered, without a high-grade buffer to lean on.

This has consequences for future flexibility. A mine with a wide head-grade-to-reserve-grade gap has room to temporarily boost output by selectively mining its best stuff. A mine where the two numbers have converged does not have that option. Production is mechanically tied to the average quality of the remaining ore body, and the average quality is trending down.

The Witwatersrand Basin in South Africa is grade decline taken to its endpoint. Over 1,000 tonnes of annual gold production in 1970. About 110 in 2024. The initial grades in the early workings were 20 to 30 grams per tonne. The mines now push nearly four kilometers underground. Rock temperature at that depth reaches 60 degrees Celsius. Every other cost category is correspondingly extreme. This is not some kind of cautionary parable. It is a physical description of what happens when a mining district's grades fall from extraordinary to depleted over the course of half a century. The arc takes a long time to complete. The direction only goes one way.

Nevada's Carlin Trend is partway along the same arc, from a much lower starting grade. Western Australia's major districts are showing signs of grade fatigue. Newer African producing regions still have room to run. Nobody starts with 20 g/t anymore. The best new projects being brought online today might start at 3 or 4 if they are lucky.

How grade interacts with exploration is worth spelling out because the two trends reinforce each other. S&P Global counted zero new gold deposits above 2 million ounces discovered in either 2023 or 2024. Six meaningful finds since 2020, total around 27 million ounces. In the 1970s through 1990s every decade produced at least one discovery above 50 million ounces. The ratio of grassroots exploration spending (new frontier work) to total exploration spending was about 50% in the mid-1990s and is 19% now. The major mining companies have compensated partly through acquisitions. Between 1998 and 2011, about 42% of reserve growth at the top gold firms came from buying juniors rather than finding new ounces organically. That reshuffles ownership of already-known gold without adding anything to the global total. The average time from discovering a new deposit to producing the first gold from it is 20.8 years.

These trends have not yet produced a visible production crisis because the existing inventory of known, partially-developed, and operating deposits is large enough to sustain current output levels for now. The issue is what the mine pipeline looks like in the 2040s. Given the discovery drought of the past five years and the two-decade development lag, the pipeline for 2040s production is much thinner than it was for the current decade. Whether this translates into an actual cliff or a gradual downslope or even gets rescued by currently unknown deposits or technology changes is not something anyone can confidently predict. The inputs are there for a problem. Whether the problem materializes or gets papered over depends on too many variables.

The other piece of this picture that the production ranking format misses: about 65 to 70 percent of global gold refining happens in Switzerland, a country that mines zero gold. Four refineries in Ticino and Neuchâtel, Valcambi the biggest at 1,500 tonnes annual capacity, process the doré from mines worldwide into 99.99% bars that bear LBMA Good Delivery stamps. Three of the LBMA's five quality referees are Swiss refineries. The concentration is a historical artifact of Italy's 1970s jewelry industry sourcing from across the Ticino border. It persists through inertia and reputation.

The Grasberg copper mine in Papua, Indonesia, incidentally produced 52.9 tonnes of gold in 2023 alongside 680,000 tonnes of copper. That byproduct gold exceeds many countries' total output. Freeport-McMoRan's CEO has stated that gold revenue at Grasberg covers total operating costs at around $1,800 gold, making the copper pure margin. Around 7% of US gold production is similarly byproduct from copper refining, recovered from anode slime during electrolysis. Part of global gold supply tracks copper economics rather than gold economics.

Recycled gold added over 1,300 tonnes to supply in 2024, about a quarter of the total. The cumulative stock of gold ever mined is about 216,000 tonnes, three to four times larger than remaining underground reserves of 54,000 to 64,000 tonnes. Central bank net buying was 1,037 tonnes in 2023. China's central bank alone took 225 tonnes in 2023 before slowing to 44 in 2024. That buying removes metal from the tradeable pool.

African gold production is projected above 700 tonnes in 2025. The official number probably undercounts by 20% or more because artisanal mining output is largely unrecorded.

Annual global mine output likely stays in the 3,500 to 3,700 tonne range for some number of years. The grade problem caps the top end. High gold prices support the bottom.