FINANCIAR
Silver after the peak — the price has halved, the deficit has stayed. What comes in 2027
After the best year and the worst day in decades, the question about silver is no longer whether it is scarce, but who still buys it at today's price.
The analysis follows five threads: how the price moved from 2024 to today, what happened to demand and supply, including at KGHM and in Romania's projects, which decisions by governments and by the exchange moved the market, how investments fared, and what olivLaw's two models say about the year ahead. Every figure has its source, and the forecast has its formula, its test on past data and the thresholds at which we would call it wrong.
The testable thesis: the silver deficit is real but small, and the 2025–2026 price was pushed mostly by borrowed money and by a temporary shortage of metal in London. That money has left; the deficit has stayed. From here, the price has long-term support, but no guaranteed short-term direction.
From 29 to 115 dollars: how the price rose
At the end of 2024, an ounce of silver (31.1 grams) cost 28.94 dollars on the COMEX exchange in New York. The first part of 2025 was quiet. Gold rose, silver did not: in April, one ounce of gold bought 105 ounces of silver, the most in years, according to the series in the olivLaw database.
The move started in the summer and sped up in the autumn. In October 2025 the London market, where most commercial stocks sit, ran almost out of available metal. The Silver Institute's annual report describes an unprecedented liquidity squeeze, with record rates for borrowing silver and metal moved into the vaults of the US exchange, according to the Silver Institute release.
The move had a political explanation. In November 2025 silver was added to the US government's list of critical minerals, and Washington was already studying tariffs on processed critical minerals. Traders shipped metal to the United States so as not to be caught by a possible tariff. Stocks in COMEX vaults reached 15,000 tonnes at the end of 2025, from 9,910 tonnes a year earlier, according to the US Geological Survey.
The price ended the year at 70.13 dollars, 142% above the end of 2024. The year's average price was 40.03 dollars, against 28.27 in 2024, according to Silver Institute data.
January 2026: the peak and the crash
In January 2026 the rise went vertical. On 26 January the futures contract closed at 115.08 dollars, the closing high in our series. On 29 January the price passed 121 dollars during the day, the all-time high, according to the Silver Institute. The gold/silver ratio fell to 44, from 105 nine months earlier.
The next day, 30 January, the closing price fell from 114.04 to 78.29 dollars, by 31.4%. The official settlement price of the March contract was 78.53 dollars, according to Barchart. It is the largest one-day fall in the olivLaw series, which starts in 2000. The previous record, from September 2011, was 17.8%.
That same day, the nomination of Kevin Warsh to lead the Federal Reserve was read as a sign of higher interest rates, the dollar strengthened, and gold also fell by about 11%, according to Wolf Street. The link is a coincidence in time, documented by the press, not a measured cause: we cannot separate how much of the fall came from the dollar and how much from the forced selling that followed. Silver fell almost three times as much as gold, which fits a market that more borrowed money had entered.
The amplifying mechanism is documented. From 13 January, the collateral required on precious-metal contracts was no longer a fixed amount but a percentage of the contract's value, according to the CME Clearing advisory. When the price rises, the collateral rises automatically. After the crash, CME raised it again, from 15% to 18% of value, from 6 February, according to Reuters, via Kitco. Those who had bought on credit had to add cash or sell.
Since then the price has come down in steps. It stayed around 76–82 dollars between February and May, fell below 60 in July and touched 55.55 dollars on 14 October, the low since the peak. On 20 October the contract closed at 57.71 dollars. Annualised volatility in 2026 is 71%, more than twice the average of a normal year.
The market balance, line by line
The deficit cannot be read from a single number. The Silver Institute and Metals Focus publish a balance in which supply also includes recycling and silver sold forward by producers, and demand also includes investment in coins and bars. The table below is the one in the World Silver Survey 2026 release of 15 April 2026, in million ounces.
| Line | 2024 | 2025 |
|---|---|---|
| Mine production | 823.6 | 846.6 |
| Recycling | 194.5 | 197.6 |
| Producers' forward sales (net hedging supply) | — | 44.7 |
| Net official-sector sales | 1.5 | 1.5 |
| Total supply | 1,019.6 | 1,090.4 |
| Industrial, total | 679.0 | 657.4 |
| of which electrical and electronics | 460.9 | 449.5 |
| of which photovoltaics | 197.5 | 186.6 |
| of which brazing alloys and solders | 49.7 | 50.5 |
| of which other industrial | 168.4 | 157.4 |
| Photography | 25.5 | 24.2 |
| Jewellery | 205.1 | 189.3 |
| Silverware | 53.5 | 42.1 |
| Coins and bars | 190.9 | 217.7 |
| Producers' hedge buybacks (net hedging demand) | 3.5 | — |
| Total demand | 1,157.4 | 1,130.6 |
| Deficit (demand minus supply) | 137.8 | 40.2 |
Three things stand out. First, the deficit shrank more than threefold in a year, from 137.8 to 40.2 million ounces. Second, a large part of the fall comes from hedging supply. As the price rose, producers sold forward 44.7 million ounces of future output to lock in the price. Without that line, the 2025 deficit would have been about 85 million ounces. Third, recycling rose to 197.6 million, a 12-year high, because people sold jewellery and silverware at high prices.
On the demand side, industry fell by 3%, mostly because of solar panels: photovoltaic use fell to 186.6 million ounces, from 197.5, as cell makers cut the amount of silver per panel. Jewellery lost 8% and silverware 21%, mostly in India. Coins and bars gained 14%. For 2026 the Silver Institute expects total demand to fall by 2%, to about 1.11 billion ounces, a marked fall in photovoltaics, an 18% rise in coins and bars and a deficit of 46.3 million ounces.
The figure needs context. A deficit of 46 million ounces is about 4% of annual demand. The gap is covered from stocks built up over past decades, which are large but are not all where they are needed, as London showed in October 2025.
Supply: mines grow slowly, even at high prices
Mexico remains the largest producer, with about 6,300 tonnes in 2025, followed by Peru with 3,600 and China with 3,400, according to USGS estimates. Known reserves are about 610,000 tonnes, roughly 23 years of output at the current pace. The largest are in Peru, Russia, Australia and China.
Why does supply not respond to price? Because most silver comes out as a by-product of lead, zinc, copper and gold mines. A copper mine does not produce more silver because silver has become more expensive. The Silver Institute expects almost flat output in 2026, of about 844 million ounces.
The companies confirm it. Pan American Silver kept its target of 25–27 million ounces for 2026, according to its second-quarter report. Hecla lowered the top of its target to 16.1 million ounces because of the Keno Hill mine in Canada, according to its August results. Fresnillo, the world's largest silver producer, reported first-half output 11.4% lower and cut its investment. Discovery Silver received in 2026 the environmental and land-use permits for its Cordero project in Mexico, one of the largest undeveloped deposits, but building it takes years.
KGHM, Europe's silver producer
The European Union has a single large silver producer: Poland's KGHM Polska Miedź, which extracts silver from the copper ore of Lower Silesia. The group produced 1,347 tonnes of silver in 2025, against 1,341 tonnes in 2024, according to the company's annual presentation, and its stated aim is to remain among the world's top three producers. Its output equals about 5% of the silver mined worldwide.
The high price showed up at once in the accounts, not in volumes. In the second quarter of 2026 KGHM produced 383 tonnes, 14% more than a year earlier, and revenue rose, according to the company, "mainly due to significantly higher metals prices (above all silver and gold)", in its quarterly release. KGHM's share on the Warsaw Stock Exchange rose from 115 to 280.80 zlotys in 2025, by 144%, reached a closing high of 392 zlotys in June 2026 and stood at 311.30 zlotys on 21 October: up 10.9% this year and down 20.6% from the peak.
KGHM shows why European silver does not respond to price: the company plans its output around copper. A doubled silver price raises its profit, not its tonnage.
Romania: known deposits, no production
Romania has no meaningful silver production today. Its best-known deposits are gold with silver, in the Apuseni Mountains, and none is close to being mined.
At Roșia Montană, the World Bank arbitration tribunal rejected all of Gabriel Resources' claims in March 2024 and ordered the company to pay costs. Gabriel asked for the award to be annulled in July 2024, and the hearing set for 22–23 January 2026 was postponed, according to the company; no new date has been announced. The mine has no licence, and the area is a UNESCO World Heritage site.
Euro Sun Mining's Rovina project was included in 2025 among the EU's 47 strategic projects for critical raw materials, according to Mining Weekly. It is a copper and gold deposit without an environmental permit, with a memorandum signed in September 2026 for financing of up to 400 million dollars. The gold and silver project at Certej, in Hunedoara, has not reached construction either. No Romanian project will add meaningful silver to the market before 2028.
The decisions that moved the market
On 7 November 2025 the US Department of the Interior published the final list of critical minerals, in which silver appears for the first time, alongside copper and uranium, according to the Federal Register. China announced on 26 October 2025 that, in 2026 and 2027, silver may be exported only by approved state-trading firms with output of at least 80 tonnes in 2024. China produces about 13% of the silver mined in the world; the measure does not ban exports, but puts them under control.
On 14 January 2026 the White House issued Proclamation 11001, which finds that imports of processed critical minerals threaten national security but imposes no tariffs and calls for negotiations, according to the published text. Silver is not named in the proclamation.
India, the largest physical buyer, changed its rules three times. From 1 April 2026, banks may lend against silver jewellery, up to 10 kilograms, and coins, up to 500 grams, but not bars, under the Reserve Bank of India's directions. On 13 May 2026, to defend the rupee, the government raised the import duty on silver from about 6% to 15%: the basic customs duty doubled, from 5% to 10%, through Notification 15/2026, and an additional 5% levy was added alongside, according to The Tribune. The duty has no expiry date.
Three days later, on 16 May, the foreign-trade directorate moved silver bars from free to restricted imports, except for units that work for export. On 2 June the restriction was extended to other forms of the metal. In May, silver imports fell to 76 million dollars, from 411 million in April, according to the Indian press. In October 2026 the authorities sped up import licences, but did not cut the duty.
Investments: the funds stayed, the speculators left
Money came into silver along three paths, and each behaved differently in the fall.
The first path is the funds that hold physical metal in a vault. The largest, iShares Silver Trust, held 528.7 million ounces at the end of 2025, according to the fund's quarterly report, and 498.5 million on 20 October 2026, according to the issuer's page. The fund lost about 6% of its metal while the price fell by 18%. Investors in these funds largely stayed put.
The second path is futures contracts bought with borrowed money. This is where the exit happened. Net positions of speculative funds on COMEX reached 29% of open contracts in March 2025 and 24.5% in September 2025. After the crash they fell to 3.5%, and on 13 October 2026 they stood at 5.2%, according to the weekly reports of the US regulator, the CFTC. Total open contracts fell from a peak of 184,831 in June 2025 to 108,433.
The third path is mining shares. They amplify the price in both directions, because much of their cost is fixed. In 2025 Hecla's share rose by 291%, First Majestic by 203% and Coeur Mining by 212%, against 142% for the metal, according to closing prices in the olivLaw database. From the 2026 peak to today, the same shares lost between 44% and 53%. The instructive exception is Wheaton Precious Metals, which runs no mines but finances other companies' mines in exchange for part of their output at a fixed price: it gained 109% in 2025 and is among the few silver stocks up in 2026, by 11.7%.
What it means for an investor in Romania
In lei, the 2026 fall was smaller. An ounce went from about 304.5 lei at the end of 2025 to about 274 lei on 20 October, or −10%, against −17.7% in dollars, because the leu weakened against the dollar over the same period, from 4.3417 to 4.7516 lei per dollar, at the BNR reference rate.
Unlike investment gold, silver bars and coins are not exempt from VAT in Romania, so the standard rate of 21% is paid on purchase. Anyone who buys physical metal thus pays about a fifth above the exchange price, plus the dealer's margin, and only gains once the price rises above that threshold. No silver funds trade on the Bucharest Stock Exchange; exposure is taken through brokers that offer access to foreign exchanges, including KGHM's share in Warsaw.
What comes next (1): the statistical model and its test on past data
The first model is a statistical yardstick, also published in the olivLaw Intel Hub. The formula: the price in t trading days is today's price multiplied by exp(σ·√t·Z), where Z follows a standard normal distribution. The median stays at today's price, so the model has no direction. σ is daily volatility, computed over the last 252 days with exponential weighting (λ = 0.97, each day's weight falls by 3% a day), so the −31% day of January now weighs less than 1%. Today σ is 38.6% a year.
We tested it on past data: at every month-end from August 2001 onward, we rebuilt the fan using only the prices known then and compared it with the realised price. The 90% band should contain the price in 90% of cases. In fact it did so in 87.7% of cases at one month, 87.0% at three months, 88.2% at six months and 84.1% at one year. The 50% band covered 47–57%. The model therefore slightly underestimates extremes.
In crises the underestimate grows. At one month, the 90% band contained the price in 67% of the months of 2011, when silver rose to 48 dollars and crashed, and in 80% of the months of the 2025–2026 liquidity squeeze. On the ±10% and ±20% thresholds the model was no better than the historical frequency of the same thresholds: the Brier score was practically equal (a difference below 1.5%). We tried three variants, with fatter tails, with volatility multiplied by 1.2 and with the empirical distribution of past moves; none was better at every horizon, so we keep the simple model and publish its coverage figures.
Today's result: in six months, the price would sit with 50% probability between 48 and 69 dollars, and with 90% probability between 37 and 90 dollars. In a year, the 90% range stretches between 31 and 109 dollars. The test says these ranges should in practice be read as covering about 85% of cases, not 90%.
What comes next (2): what the internal olivLaw engine says
The second model is olivLaw's forecasting engine, which runs internally and which we publish here for the first time, as an experiment. We ran the question "how likely is silver to touch a threshold in the next 6 or 12 months" on 22 October 2026, with a starting price of 57.71 dollars. The engine answers a different question from the statistical model: the chance that the price touches the threshold at least once in the window, not that it ends there. That is why both directions can have high probabilities at the same time.
| Threshold touched at least once | Within 6 months | Within 12 months | Without the historical trend, 12 months | Random walk, 12 months | Historical frequency, 12 months |
|---|---|---|---|---|---|
| +15% (above $66.36) | 61% | 78% | 65% | 75% | 66% |
| +10% (above $63.48) | 75% | 86% | 73% | 83% | 75% |
| +5% (above $60.59) | 84% | 91% | 83% | 91% | 87% |
| −5% (below $54.82) | 75% | 81% | 86% | 91% | 78% |
| −10% (below $51.94) | 54% | 67% | 71% | 81% | 58% |
| −15% (below $49.05) | 43% | 62% | 61% | 71% | 44% |
How to read it. On the upside, the engine sits slightly above the random walk and above history. On the downside, it sits well below the random walk and close to the historical frequency. The column without the trend shows where the difference comes from.
The engine learns its tables from silver's monthly history since November 2003, the only break found in the series, and over that period silver rose on average by 0.88% a month. With that trend removed, the chance of a 10% fall within a year rises from 62% to 71%, and that of a 10% rise falls from 82% to 73%, on the same table. The tilt toward recovery is therefore mostly the trend of the past 23 years, not a reading of the situation after the crash.
What the engine sees and does not see. Since 22 October, three actors reach the silver price in the run: the Federal Reserve, through its policy rate, which enters the numbers; the CME exchange, through the collateral it requires, which remains only an annotation; and the speculative funds on COMEX, through the positions reported to the CFTC, also as an annotation. China's Ministry of Commerce and India's government are still missing, because the engine's rules require a citable official text for each actor, which we could not save.
The engine has two more limits: it does not recognise the turmoil of 2025–2026 as a new volatility regime, because its statistical test asks for stronger evidence than the data so far give, and it answers only the touching question, not the one about the final price. The results are marked "diagnostic" internally and have not yet been scored on past data.
What comes next (3): three scenarios for 20 October 2027
The scenarios are defined on the same variable and on the same date, so that they can be checked: the closing price of the COMEX silver futures contract on the last trading day up to 20 October 2027. The three ranges cover every outcome, and the probabilities add up to 100%.
- A · Recovery, above 65 dollars (40%): the Federal Reserve stops raising rates, India eases imports, the deficit is confirmed in the April 2027 report, and physical-metal funds add to their holdings.
- B · Flat, between 50 and 65 dollars (30%): the deficit stays small, speculators do not return, industry keeps saving silver, and India keeps its duty.
- C · A new fall, below 50 dollars (30%): higher US rates, a strong dollar, a slowing world economy or outflows from physical funds.
Where the numbers come from. The statistical model gives 38% for A, 27% for B and 35% for C. We move five points from C to A for fundamental reasons, not because of the internal engine, whose tilt comes mostly from the historical trend: speculative positions are almost emptied, so they have little left to sell, the deficit continues, and the gold/silver ratio of 71 is above its average of the past year. We round to 40, 30 and 30.
The causal chain
Links 2, 3, 5, 7 and 8 are documented from primary sources or exchange data. Link 4 is measured in the CFTC positions. Link 6 is a connection in time, not a measured cause: we know what happened on the same day, not how much each factor contributed. Link 9 is our conclusion, which we put to the test through the predictions below.
Forecast table
| Prediction | Horizon | Probability | How it is checked | Verification source |
|---|---|---|---|---|
| The COMEX silver futures contract closes above 60 dollars an ounce | 31.03.2027 | 40–50% | Settlement price of the front contract on the last trading day of March 2027 | cmegroup.com |
| The Silver Institute reports a deficit for 2026, i.e. demand above supply | by 30.04.2027 | 75–85% | The balance in World Silver Survey 2027 | silverinstitute.org |
| Photovoltaic silver use reported for 2026 is lower than in 2025 | by 30.04.2027 | 85–95% | The “photovoltaics” row in World Silver Survey 2027 | silverinstitute.org |
| iShares Silver Trust holds more than 500 million ounces | 30.06.2027 | 45–55% | Holdings published by the issuer on that date | ishares.com |
| The gold/silver ratio, June 2027 monthly average, stays above 60 | 30.06.2027 | 70–80% | Monthly average of the ratio between COMEX futures prices | cmegroup.com |
| Net positions of speculative funds exceed 15% of open contracts in a CFTC report | by 31.03.2027 | 30–40% | Weekly Commitments of Traders report, COMEX silver | cftc.gov |
| The US imposes a customs duty on silver imports | by 30.06.2027 | 5–15% | Proclamation or order published in the Federal Register | federalregister.gov |
| India lowers its total import duty on silver below 15% | by 30.06.2027 | 30–40% | Notification by India's Ministry of Finance | cbic.gov.in |
| Scenario A: COMEX silver closes above 65 dollars | 20.10.2027 | 35–45% | Close of the front contract on the last trading day up to 20.10.2027 | cmegroup.com |
| Scenario B: COMEX silver closes between 50 and 65 dollars | 20.10.2027 | 25–35% | The same close | cmegroup.com |
| Scenario C: COMEX silver closes below 50 dollars | 20.10.2027 | 25–35% | The same close | cmegroup.com |
Correction: the first version of this analysis, also published on 21 October, said India had doubled the duty "from 5% to 10%, until 31 March 2027" and carried a prediction about extending that duty. The total duty on silver in fact rose from about 6% to 15% and has no expiry date. That prediction has been voided in our forecast register, and the question about a duty cut has taken its place.
When we would call the analysis wrong
We set the thresholds in advance, so as not to interpret them after the fact.
- The deficit thesis fails if World Silver Survey 2027 shows supply at least equal to demand for 2026.
- "Mines do not respond to price" fails if reported mine output for 2026 exceeds 870 million ounces, 2.8% above 2025, against about 844 million expected by the Silver Institute.
- "Physical funds stayed" fails if iShares Silver Trust holdings fall below 450 million ounces, 10% below the level of 20 October, on any day up to 30 June 2027.
- "Speculators have left" fails if their net positions pass 20% of open contracts, the level of autumn 2025, before the April 2027 report.
- The statistical model is too narrow if the price leaves the six-month 90% band, that is below 37 or above 90 dollars on 20 April 2027.
Method, definitions and figures
The price series. We use the daily close of the COMEX silver futures contract nearest to expiry, as published by Yahoo Finance under the symbol SI=F, from August 2000 onward. The series moves from one contract to the next before expiry, by a rule Yahoo does not publish.
We mark the days suspected of a contract change: a day inside the expiry window (the last 30 trading days before the contract's last day) with a move larger than 3.5 typical deviations. The silver series has 89 such days since 2000. Of the days cited in the analysis, only 23 September 2011 falls in this category, a real market crash that coincided with the December contract's window. The days marked at the end of December 2025, with jumps of 8–11% in both directions, were real moves: the exchange's official settlement prices show the same jumps, and the 31 December close, from which we compute the 2026 change, differs from settlement by less than 1%.
Close and settlement. The close is the last price of the day in the series. The settlement price is the official price set by the exchange at the end of the session and can differ: on 30 January the close was 78.29 dollars and settlement 78.53. Every change in the analysis is computed on closes, and settlement is cited only where a public source gives it. The price in lei is computed with the BNR reference rate.
Demand and supply. The balance is the one in the World Silver Survey 2026 release; the 2026 figures are estimates from the same source. Output by country and reserves are USGS estimates, which use a different method from Metals Focus. KGHM's output and share performance come from company reports and daily closes, without dividends.
Probabilities. The statistical model, the internal engine and the scenarios are three different things and we name them separately. The probabilities in the table are editorial estimates, in bands of at least ten percentage points, and will be scored publicly when due.
Limits of the analysis
We do not have access to daily stocks in London vaults, to the full World Silver Survey data or to the official tables of CME collateral, which we give as reported by the press. India's notification on the additional 5% levy could not be read on the official site, and India's import figures after May 2026 come from the press. The internal engine has only three actors for silver, one of which enters the numbers, and the test of the statistical model has, at the one-year horizon, only about 24 independent observations.
