Reading the Shanghai-COMEX Silver Spread as a Physical Market Signal

In the eight months since late December 2025, silver has effectively carried two different prices at once: one in Shanghai, one in New York, and the gap between them has refused to close. On 1 September 2026, the Shanghai Gold Exchange benchmark sat roughly $8 above COMEX, a premium that had peaked as high as 21.9% in late January 2026.

Persistent double-digit divergence of this kind between two of the world’s most liquid commodity exchanges is not supposed to last eight months.

Regional price premiums in commodity markets are ordinary enough. They reflect freight, taxes, tariffs, and local supply-demand balances. What makes this particular spread analytically significant is its persistence, its scale, and the chain of physical dislocations that preceded it: a supply crunch in India pulled metal from London, London drew from COMEX and China to top itself back up, and the resulting tightness in Chinese physical markets gave the Shanghai premium both a trigger and a rationale.

The spread has since outlasted the acute shortage that created it. That is where the interpretive challenge begins.

What follows here is a framework for reading regional silver premiums as physical market signals: separating what the data confirms from what it merely suggests, and marking clearly where genuine analytical disagreement remains. The read you take from this should leave you with a more grounded view of silver’s supply-demand fundamentals than futures pricing alone can offer.

How the Shanghai premium built: a cascade of physical dislocations

The spread did not appear from nowhere. It was the endpoint of a sequence of physical events, each one feeding the next, and reading them in order is what turns an apparent anomaly into something closer to inevitability.

The cascade ran in four stages:

1. An overt silver shortage developed in India, the world’s largest silver market, pulling physical metal toward the subcontinent.
2. London drew down its own inventories to satisfy that Indian pull.
3. To replenish London, metal was pulled from both COMEX warehouses and China.
4. The resulting tightness in Chinese physical markets gave the Shanghai premium its trigger, and the premium held.

Each link was a physical event, not a paper market move. That distinction matters, because a spread built on the movement of real metal behaves differently from one built on speculative positioning.

The warning had actually arrived more than a year earlier. During summer 2024, a China premium of roughly $3 appeared, modest in absolute terms but highly unusual, and large enough to make shipping silver into China economically worthwhile. Chinese smelters responded by contacting Latin American silver producers directly, bypassing exchanges to lock in supply.

By late December 2025, that early tremor had become a structural shift. Daily Shanghai premiums pushed above 10%, then kept climbing.

The Shanghai premium peaked at roughly 21.9% on 30 January 2026, one of the widest regional silver divergences on record.

When a $3 premium becomes a 20%-plus premium in eighteen months, the story is no longer about arbitrage economics. It is about procurement reality diverging from paper pricing.

When manufacturers call miners directly

The clearest evidence that the pressure had reached the producer level came from a datable company event. Between Christmas and New Year of 2025, solar panel manufacturers in both China and India contacted Kuya Silver directly, offering 8% to 10% premiums above market to lock in forward production agreements.

Public sources do not widely document the specific terms of those approaches. But the behaviour pattern itself, industrial buyers paying forward premiums to secure metal straight from a miner, is recognised by market analysts as a supply-security signal.

That is the tell. When major industrial users bypass the exchange system entirely and pay a premium to reach the mine, it means paper market pricing has decoupled from the physical procurement reality those manufacturers actually have to navigate. Reading the sequence, rather than the snapshot, is what explains why the premium persisted: each link created its own local tightness, and the effects compounded rather than cleanly resolved.

What the inventory data actually shows across COMEX, Shanghai, and London

If the cascade explains why the premium emerged, exchange inventory data shows how much physical stress sat behind it. The COMEX drawdown is the most quantitatively dramatic reading available.

COMEX total silver inventory fell from roughly 532 million ounces in October 2025 to 366.25 million ounces by 20 February 2026, a decline of nearly 31% in about four months. Registered stocks, the metal actually available for delivery, slipped below 90 million ounces, hitting 88,191,059 ounces in late February.

One withdrawal stands out. The first week of January 2026 reportedly saw a single-week outflow of 33.45 million ounces, though this figure is unverified and should be treated as indicative rather than confirmed.

The March 2026 physical delivery crisis provides a concrete stress test of how thin registered stocks interact with futures open interest at contract expiry, a mechanism that sat directly beneath the most acute phase of the COMEX drawdown documented here.

Shanghai told a quieter but equally revealing story. Early in 2026, combined visible inventories across the Shanghai Gold Exchange and the Shanghai Futures Exchange (SHFE) were estimated at roughly 700 tonnes, strikingly low for the market that consumes the bulk of the world’s industrial silver.

Then came a partial recovery. By 31 August 2026, COMEX warehouse silver had rebuilt to 338.1 million ounces total, split into 101.3 million registered and 236.8 million eligible. SHFE stocks, meanwhile, stood at roughly 1,407 tonnes as of September 2026.

The temptation is to read the rebuild as a return to normal. It is not. Registered stocks at 101.3 million ounces remain well short of the levels that would let the spread compress without continued physical premium pressure.

There is also a caveat that has to sit alongside every one of these figures:

• Declining exchange stocks are not the same as declining above-ground supply. Metal can move into private, off-exchange storage and vanish from the data while still existing.
• Registered and eligible are different categories. Only registered metal is available for immediate delivery; eligible stock sits in the warehouse but is not committed to the market.
• “Invisible inventory” moving to private vaults is statistically indistinguishable from true depletion in the published exchange numbers.

Even with those caveats, inventory data is harder to manipulate than futures pricing, which makes it one of the cleaner signals you have for separating genuine physical tightness from speculative froth. The gap between the October 2025 total and the August 2026 total, even after the partial rebuild, remains large enough to matter.

Two frameworks for reading the same spread

Here the analysis has to slow down, because credible specialists genuinely disagree about what all of this means. There are two coherent frameworks, and each deserves its strongest form.

The structural deficit view reads the evidence as proof that something durable has shifted. Registered COMEX stocks have fallen and only partially recovered. Industrial users are securing metal via direct offtake at premiums. Solar demand keeps expanding. And the spread has persisted for eight months rather than closing in days. Under this framework, Western futures markets are underpricing scarcity risk, while local physical exchanges in China and India are pricing it accurately.

Paper market pricing systematically understates physical procurement costs when the arbitrage channels that normally keep futures and spot prices aligned are structurally impeded, a dynamic that the structural deficit view treats as a persistent mispricing signal rather than a temporary basis trade anomaly.

The market friction view reads the same data and sees distortion rather than deficit. China’s capital controls complicate arbitrage between onshore and offshore markets, which prevents the gap from closing even when it is economically large. Domestic value-added tax (VAT) on silver structurally pushes Chinese wholesale prices above international benchmarks. SHFE contracts are renminbi-denominated and physically delivered on different terms from COMEX’s US dollar system. And India’s May 2026 import curbs show how a policy decision alone can manufacture a localised shortage regardless of global supply.

India offers the sharpest illustration of how fast policy can reshape a market. Prior to restrictions, silver briefly traded at a discount of up to $5.50 per ounce. Then, on 17 May 2026, the government moved silver bars of 99.9% purity into the “restricted” import category, and the direction reversed hard.

India’s silver premium swung from a $5.50 discount to roughly $6.50 per ounce by July 2026, and to a reported $10.89 by 1 September 2026, though sources conflict on that final figure.

Demand growth sits underneath all of it. Analysts projected in 2024 that India’s silver imports could double, driven largely by the solar sector. The competition for metal is not abstract, either: at current prices, a $2 billion outlay buys roughly 18 million ounces, which puts industrial users, investors, and exchange-traded funds (ETFs) in direct contest for the same finite pool of deliverable silver.

So which framework wins? Neither cleanly. Both are partially correct at the same time, and the analytical task is to weight them rather than pick one. For an investor acting on physical silver signals, the friction view is a reason to discount the spread’s magnitude, not a reason to dismiss it. The policy constraints that stop arbitrage are themselves a form of physical market risk.

India’s premiums add a second physical market stress signal

India matters here for a specific reason: its premium is analytically independent of the Shanghai spread. As the world’s largest silver market, India’s pricing carries weight in its own right, and when it runs a large premium for different reasons than China does, the two together form a stronger signal than either alone.

The policy timeline is worth tracking in sequence:

1. Late 2025: Indian silver traded at an estimated 16.3% premium over global prices, based on a global price of $51.69/oz and an India spot price near ₹185,000/kg, though this baseline is unverified.
2. May 2026: Before the restrictions bit, silver briefly traded at a discount of up to $5.50 per ounce.
3. 17 May 2026: Silver bars of 99.9% purity were moved into the “restricted” import category, aimed at cutting the import bill and supporting the rupee.
4. July 2026: The resulting shortage pushed premiums to roughly $6.50 per ounce, more than 10% above domestic benchmarks.
5. 1 September 2026: A reported premium of $10.89 per ounce, surpassing China’s, though sources conflict and subsequent research suggests the figure may be lower.

The direction of travel tells you the important part. India’s premium overtaking China’s on 1 September 2026 does not mean the stress has eased; it means the pressure point has migrated. The stress is still in the system, just concentrated in a different geography.

India’s silver demand growing 192% creates a demand floor that policy interventions like the May 2026 import restrictions can only partially suppress, because the underlying industrial procurement need does not disappear when import channels narrow; it reprices into the premium.

Solar demand as the floor beneath both premiums

Beneath the policy shocks sits a structurally growing demand source that keeps premiums elevated even after acute disruptions fade. Analysts projected in 2024 that India’s silver imports could double, driven largely by solar manufacturing. That growth is what puts industrial users in direct competition with financial investors for deliverable metal, and it is the strongest single support beneath the structural deficit case.

Two major markets running significant premiums over Western prices, for different but compounding reasons, is a higher-order signal than either premium in isolation. If you are assessing silver’s supply fundamentals, both spreads deserve tracking, not just the COMEX-Shanghai pair.

What a calibrated investor reads from eight months of spread persistence

The point of all this is not a verdict of “buy silver” or “ignore the spread.” It is a diagnostic posture: knowing which variables will tell you whether the spread reflects durable structural change or is slowly resolving back toward parity.

Start with the current baseline, and note that even this is contested. On 1 September 2026, one source put COMEX at $65.50 against Shanghai at $73.16, an $8.21 premium, while MetalCharts recorded $63.89 on COMEX and a $72.25 SGE benchmark. Treat the live premium as a range in the region of 13%, not a single precise figure.

Across the full eight-month window, the spread averaged roughly $8. That is not noise; it is a durable signal that physical silver in the world’s largest manufacturing market has been priced structurally above the Western futures price.

The task is to distinguish persistence driven by structural deficit from persistence driven by policy and friction that will eventually normalise. These are the variables that separate the two:

• COMEX registered stock trajectory. Watch whether the rebuild past 101.3 million ounces (August 2026) continues toward late-2025 levels or stalls.
• The Shanghai-COMEX daily spread. A steady grind back toward parity favours the friction view; sustained double-digits favour the deficit view.
• India’s premium versus its historical baseline. A second geography holding a large premium strengthens the structural read.
• Direct industrial offtake news. More miners fielding forward-premium approaches signals genuine supply anxiety, not tax artefacts.
• Solar capacity installation pace in China and India. Rising installations sustain the demand floor beneath both premiums.

The 21.9% peak of 30 January 2026 stands as the reference point for what acute stress looks like. Measuring current readings against that peak, and against the $8 average, is how you convert this framework into an actual monitoring posture rather than a one-off reaction to spot price.

Reading the spread as a map, not a forecast

The honest conclusion is that both frameworks are operating at once. The structural deficit view and the market friction view are not rivals to be resolved into a single answer; they are two forces acting on the same market, and the investor’s job is calibration, not a binary choice.

What the analysis upgrades is the toolkit. Regional spreads become physical market stress indicators that sit on top of Western futures pricing, not instead of it. A spread that has held for eight months across two of the world’s most liquid silver markets, averaging roughly $8, is telling you that something structural has shifted in how physical silver moves globally, even if the precise nature of that shift stays contested.

By September 2026, the acute stress of the January peak has partly normalised, yet the spread has not closed to historical norms. That gap is the story.

The clearest near-term test is COMEX registered stock. If it keeps rebuilding toward late-2025 levels, the friction view is winning. If the recovery from 101.3 million ounces stalls, the structural view gains ground.

Track that one number, alongside the Shanghai and India premiums, and you hold a more complete picture of physical silver than spot price alone will ever give you.

For investors translating this spread analysis into a physical silver position, our dedicated guide to physical silver liquidity risks examines the execution gap between paper price and actual realised proceeds when selling into a dislocated market.

This article is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions.

Past performance does not guarantee future results. Financial projections are subject to market conditions and various risk factors. Several figures referenced above are noted as unverified or subject to conflicting sources, and should be treated as indicative rather than confirmed.