1The size of the prize

2Why the residual is the whole problem

When India pays China in rupees, China ends up holding rupees. It can buy Indian goods, invest in Indian assets, or sit on them. India sells China $19.5 billion and buys $131.6 billion. After the trade nets out, China holds $112.2 billion of rupees a year with no Indian goods to spend them on.

This distinction matters because it changes what the policy actually is. Settling balanced trade in local currency removes dollar demand and costs nobody anything. Settling a deficit converts the obligation into a foreign-held rupee claim — economically identical to China buying Indian government bonds, and subject to the same question: does it want to?

The trend is the finding. The share of BRICS trade that could be settled in local currency without creating an overhang has been shrinking for ten years.

3What they actually sell us decides what can be re-invoiced

Commodities are fungible, standardised and quoted against public benchmarks. Re-denominating them is a contractual choice. Manufactures are differentiated, embedded in multi-year supply contracts and priced on specification. Splitting each counterparty's export basket into the two produces the sharpest result in this analysis.

China is the outlier and it is the one that matters. It holds 49% of the entire residual but only 8% of what it sells India is a fungible commodity. Its top three lines — electrical machinery ($46.4bn), industrial machinery ($29.5bn) and organic chemicals ($11.5bn) — are 66% of the bilateral bill and precisely the categories where re-invoicing means renegotiating thousands of specification-level contracts against a supplier with no reason to concede.

Invert it and the picture is encouraging. Of the $228.5 billion residual, $125.3 billion sits in trade that is commodity-shaped and could in principle be re-priced. Russia, the UAE and Saudi Arabia together contribute $97.9 billion of that.

Two lines carry the case. Energy at $106 billion and bullion at $30.6 billion are 77% of everything re-invoiceable. Both are areas where India is a top-three global buyer, which is the only source of leverage that matters in an invoicing negotiation. India is the world's third-largest crude importer and among the two largest gold importers.

4Securitising the residual: what the yield curve says

If partners must hold rupees, the policy question becomes what they hold them in. A zero-yield vostro balance is a tax on the seller and will be resisted. A marketable rupee security priced off the sovereign curve is an investment — the model Norway's sovereign fund and the Gulf funds already run in other currencies.

India's curve on 3 August 2026, from the Clearing Corporation's indicative yields, gives the pricing spine:

Now set that against what each counterparty earns at home. The result is the strongest argument in favour of the whole scheme — and it lands on exactly the right countries.

China, the UAE and Saudi Arabia all have policy rates and long yields below India's, and between them hold $159.2 billion — 70% of the residual. For China the pickup is 513 basis points: a Chinese institution holding a ten-year Indian government bond earns four times what a Chinese government bond pays. The rupee would have to depreciate more than 5.13% a year against the yuan before that holder was worse off.

India has also, without framing it this way, already built most of the plumbing. Foreign holdings under the Fully Accessible Route face no ceiling; the overall cap on central government securities is 6% of outstanding stock; the RBI extended the FAR to all new 15-, 30- and 40-year issuance; and interest income and capital gains on FPI holdings of government securities were exempted from tax effective 1 April 2026. Foreign investors put a record ₹39,640 crore into Indian government bonds through the route in June 2026 alone.

5Russia already ran the experiment — and quietly abandoned it

The most important evidence is not a model. After 2022, sanctions pushed India–Russia trade out of dollars at scale. India now buys $55.4 billion a year from Russia, 86% mineral fuels (HS Ch.27), overwhelmingly crude, and sells it $4.5 billion. The structural residual is $50.9 billion a year.

Rupees duly accumulated, peaking near $8 billion. Sergey Lavrov said publicly in Goa in May 2023 that Russia had billions of rupees it could not use. By September 2024 the balance had fallen to roughly $3.5 billion.

The absorption arithmetic explains why. India's commerce ministry has identified some 300 product lines where Indian exporters could plausibly serve Russian demand. India currently exports $1.7 billion across them. Russia's global import demand across the same 300 lines is $37.4 billion — about three-quarters of a single year's bilateral deficit. Even capturing the entire pool, displacing China and Turkey completely, would not close the gap. Russia's total import bill is only $283 billion, and China already takes about 41% of it.

The one genuine opening is pharmaceuticals: India is already Russia's largest supplier of generics by volume, exporting roughly $546 million against Russian global pharma imports of $9.7 billion. That is the biggest single addressable gap anywhere in the Russian basket — and it is worth about 1% of the annual deficit.

6The inverse case, and where rupees can actually be spent

Brazil is instructive precisely because it sits on the boundary — and because it moved. On FY2024-25 data India ran a surplus of about $1.3 billion with Brazil, which would have made Brazil a rupee consumer rather than an accumulator. On FY2025-26 data that reversed to a deficit of $1.0 billion, as imports from Brazil jumped 48% in a single year. The imbalance is small either way, which is what makes Brazil a plausible netting partner rather than a large accumulator. But the sign flip is itself the warning: a counterparty that can swing from rupee consumer to rupee accumulator inside one year is a weak foundation for a settlement mechanism. Egypt, Iran and Ethiopia sit more durably on the surplus side of the ledger.

South Africa illustrates a second obstacle that is political rather than arithmetic. India is already South Africa's second-largest source of imports at roughly 7.3%, and the product fit is unusually good: South Africa has lost almost all of its refining capacity, importing about 60% of its fuel, which is precisely the gap Jamnagar could fill. Yet Pretoria has declined the rupee-first model, and so has Brasília. Both favour a multi-currency arrangement in which traders settle in whichever BRICS currency suits them. India's proposal is not the bloc's consensus position, and the two members whose trade is most balanced with India are the ones resisting it.

This points to the missing piece in most local-currency proposals. A rupee is only useful to someone who wants Indian goods, so the global capacity to absorb rupees is bounded by India's exports. Across India's major partners, the surplus relationships — the United States ($33.8bn), the Netherlands ($11.7bn), Bangladesh ($8.8bn), Nepal ($5.5bn), Spain, Sri Lanka, Kenya, Mexico and others — total roughly $85 billion a year of demand for rupees. Against a BRICS residual of $228.5 billion, the arithmetic does not close within the bloc. It only closes if rupees earned in Moscow or Shanghai can be spent in Dhaka or Nairobi, which requires a genuinely multilateral clearing mechanism rather than a set of bilateral vostro arrangements.

7What the broader trade research says

The McKinsey Global Institute's 2026 update on the geometry of global trade provides useful corroboration and one important corrective. Goods trade grew about 6.5% in 2025, roughly in line with global output, even as supply chains re-routed — trade is being redirected rather than reduced. Economies are trading more with geopolitically aligned partners, and India is named among the principal beneficiaries alongside Vietnam, Indonesia and Mexico. US–China trade fell around 30% between 2024 and 2025.

The corrective is the concentration finding: roughly 10% of global trade by value is "globally concentrated", with three or fewer economies supplying more than 90% of a given product. That is the structural reason the China line in Exhibit 3 is so hard to move. Where a supplier is one of very few sources, the buyer has no invoicing leverage regardless of how large its purchases are. India's leverage exists in energy and bullion, which are deep and diversified markets, and is close to zero in the electronics and machinery categories that dominate the Chinese bill.

8Country focus: Russia, and what seventy years of rupee–rouble trade already taught us

Russia deserves separate treatment because it is the only counterparty where the entire experiment — a collapsing export ratio, a rupee overhang, and a bilateral settlement mechanism — has already been run twice, seventy years apart.

The proximate cause is the war. India's imports from Russia sat at $9.87 billion in FY2021-22. One year later, as sanctions redirected Russian crude away from Europe and toward the few buyers willing to take it, they reached $46.21 billion — a 4.7-fold increase in a single fiscal year. They peaked at $63.81 billion in FY2024-25 and were $55.37 billion in FY2025-26, still ten times the FY2016-17 level.

Two things changed and one did not. Russia went from supplying roughly 2% of India's crude in FY2019-20 to about 35% by FY2024-25, and Indian refiners captured a discount for taking barrels the sanctioned market would not. What did not change was India's ability to sell anything back: exports to Russia rose only from $3.25 billion to $4.49 billion across the same period, so export cover of imports collapsed from 33% to 6.8% and has recovered only to 8.1%. The rupee overhang was not a policy failure. It was the arithmetic consequence of a war-driven import surge with no matching export channel.

India–Russia trade was $59.86 billion in FY2025-26. Removing the dollar as intermediary saves the double-conversion spread on that flow, and nothing else:

The defensible figure is $150–350 million a year — real, but a rounding error on a $60 billion relationship.

The conclusion is uncomfortable but clear. Pursue rupee–rouble settlement with Russia for sanctions resilience and optionality, which are genuine and strategically valuable. Do not pursue it for savings. The savings are small, the friction being removed is currently earning India more than it costs, and the one mechanism history offers as precedent ended with sixteen years of unwinding an unusable balance.

9Slowing the fall: the instruments India is already using

De-dollarising trade settlement, as the preceding sections show, does very little for the rupee's level. But India is not short of instruments that do act on it — and on 5 June 2026, with the rupee under pressure and portfolio money leaving, the Reserve Bank and the government deployed most of them at once. The package is worth reading closely, because it is a textbook demonstration of the difference between changing the currency of a payment and changing the demand for the currency itself.

Every one of these acts on the capital account, and that is the point. The rupee's level is set by net flows, not by the invoicing currency of trade. Settlement reform changes who holds which claim; these measures change how many claims foreigners want to hold at all.

The G-sec channel is the one with the most room left in it, and it connects directly to the securitisation argument in Section 4. India has spent three years making its government bond market accessible to foreigners: the Fully Accessible Route carries no per-investor cap, index inclusion in J.P. Morgan's GBI-EM from June 2024 phased India to a 10% weight, and the tax exemption now removes the last major deduction from a foreign holder's return. Foreign inflows into FAR securities hit a record ₹39,640 crore in the single month of June 2026.

The headroom is real and quantifiable. The overall ceiling remains 6% of outstanding central government securities (and 2% of state development loans), and actual foreign holdings sit far below it. That gap is the country's most underused instrument: it is a standing, pre-authorised channel for foreign capital that requires no new negotiation with any counterparty — unlike every settlement mechanism discussed earlier in this report, which requires a partner to agree.

Set against the alternative, the logic is stark. Selling gold to buy foreign assets exports Indian savings and finances someone else's deficit. Selling rupee-denominated debt to foreign investors does the opposite: it finances India's deficit with foreign savings, in India's own currency, with the exchange-rate risk sitting on the buyer. That is the structural privilege reserve issuers enjoy, and the FAR route is the closest thing India has to it.

This is the most instructive table in the report, and it comes straight off the Bank of Russia's own site. Over eleven months Russia's total reserves rose 4.5% — but the entire increase came from gold, which grew 17.3%, while the foreign-exchange component actually fell 3%. Gold now sits at 41.5% of Russian reserves, having peaked at 48.3% in January.

So the widely repeated claim that Russia "moved into the yuan" needs splitting in two. Russia transacts in yuan because sanctions left it little choice — and that pool keeps running dry, which is why swap rates hit 40% and the central bank is contemplating mandatory holdings. But Russia stores value in gold. A currency that has to be rationed by decree is a working balance, not a reserve asset.

India, on the Reserve Bank's own reporting, is doing a milder version of the same thing: gold rose from 13.92% to 16.70% of reserves over the year to March 2026. The revealed preference of both countries is identical — when you distrust the incumbent system, you accumulate the one asset that is nobody's liability. Neither has chosen to hold the other's currency, and that is the honest ceiling on intra-BRICS reserve diversification.

Electronics is the single largest line India pays for in dollars anywhere in its BRICS trade: $46.37 billion in one year, growing 21.9%. If any bilateral flow could justify a dedicated rupee–yuan settlement arrangement on transaction-cost grounds alone, it is this one — a single counterparty, a single currency pair, enormous and rising volume, and two central banks that both run CBDC pilots. At a realistic 25 basis points of round-trip conversion cost, moving the whole China bill off the dollar would save on the order of $330 million a year; on electronics alone, about $115 million.

The realistic sequence follows from that. Rupee–yuan settlement on electronics is worth negotiating for transaction cost and sanctions resilience, and it should be paired with the FAR channel so that any yuan-side rupee accumulation has a yielding instrument to sit in rather than a dead vostro balance. It is not worth negotiating as a way to defend the rupee: at $131.6 billion a year the whole China bill is about $0.36bn/day, roughly 0.28% of daily USD/INR turnover. And the underlying problem in this exhibit is not a currency problem at all. A bill rising 16% a year in the one category India cannot substitute is an industrial-policy problem, and no settlement mechanism will fix it.

Read down that table and the lesson for India is unusually precise. China succeeded emphatically at one thing — getting its own trade settled in its own currency, from 13% to 30% in six years. It has barely moved the other two: third countries still invoice only 1.5% of their goods imports in yuan, and the currency remains 2% of world reserves.

Those are three different ambitions, and they are not equally available. Getting foreigners to hold your currency requires them to want it as a store of value, which is the thing neither the yuan nor the rupee has achieved. Getting third countries to invoice with each other in your currency is harder still. But settling your own trade in your own currency requires only that your counterparties accept it in payment — and China has demonstrated that a large, persistent trading nation can take that from 13% to 30% in six years.

10The four-country lens: who actually holds whose currency

The cleanest test of every argument in this report is to stop asking what governments say and start reading what their central banks publish. Take the four counterparties where the question is live — Brazil, Russia, the UAE and South Africa — and ask two things of each: how much renminbi and how much rupee is in your reserves, and how is your trade splitting between India and China. The answers line up with uncomfortable precision.

Three findings fall straight out of this table, and each one cuts against a claim that circulates freely in the de-dollarisation debate.

First, no rupee. Brazil and South Africa both publish complete, itemised currency lists. Brazil's runs to nine lines — dollar, gold, euro, renminbi, sterling, yen, Canadian dollar, Australian dollar, and a Korean won position added in 2025. South Africa's runs to six. The Indian rupee is on neither. It is not a small allocation; it is not an allocation. Russia and the UAE publish nothing, so the rupee cannot be confirmed or denied there — but Russia's rupee balances are held in commercial-bank vostro accounts under the SRVA mechanism, which is a settlement arrangement, not a reserve holding. The distinction matters more than any number in this report: a vostro balance is a claim a trading counterparty is waiting to spend; a reserve holding is a claim a sovereign has chosen to keep. India has produced the first and none of the second.

Second, the yuan's reserve share is not even rising uniformly. Brazil's went up, from 5.31% to 5.94%. South Africa's went down, from 6.0% to 5.6%. Both sit within a percentage point of each other and neither has moved much in three years. Set against that, look at the gold column: Brazil roughly doubled its gold share in a single year (3.55% to 7.19%), South Africa took its from 18.7% to 23.9%, and Russia is at 41.5%. Every one of these central banks bought materially more gold than renminbi. That is the same revealed preference documented for Russia in Exhibit 9, now visible in four institutions on three continents. The move away from the dollar is real; the move into any rival currency is not.

Third, Brazil's dollar share fell 6.45 percentage points in one year — from 78.45% to 72.00% — which is the single largest de-dollarisation step anywhere in this dataset. Gold took 3.64 points of that — well over half — against about six-tenths of a point into renminbi. If a country actively reducing its dollar exposure at that speed still puts nearly six times more of the released money into metal than into China's currency, the ceiling on currency substitution is not a matter of political will.

Rank the four by how much renminbi they hold and you get roughly the order in which they trade with China. Rank them by rupee holdings and the exercise is trivial, because the answer is zero everywhere. But the last column is where India's actual position becomes clear: the one country in this group that trades with India on something close to parity with China is the UAE — 1.07 to 1 — and the UAE is also the only one to have signed a local-currency settlement arrangement with India. Brazil trades twelve and a half times more with China than with India, and has no rupee arrangement at all. The pattern is not ideological. It is gravitational.

11Rechecking the rupee–rouble saving against the exchange rate itself

Section 8 priced the saving from removing the dollar as intermediary in India–Russia trade at $150–350 million a year on $59.86 billion of flow — the double-conversion spread, and nothing else. That calculation is correct as far as it goes, but it prices only one side of the trade. Eliminating the dollar leg does not eliminate currency risk; it relocates it, from a pair India transacts in constantly to a pair that barely trades. What that relocation costs is not a matter of opinion. It is in the exchange rates.

Over ten and a half years the rupee lost 43.6% against the dollar and the rouble lost 17.5% — which, read as a headline, makes the rouble look like the sounder currency. The path says the opposite. The rupee's line is a slope; the rouble's is a seismograph. Between March 2022 and September 2023 the rouble went from 84.09 to the dollar, to 57.41 under capital controls, to 97.42 when the controls loosened — a 32% appreciation and then a 70% depreciation inside eighteen months.

The number that matters for a trade invoice, though, is neither of those. It is the cross.

Set that against the saving. Removing the dollar leg on $59.86 billion of trade is worth roughly $150 million a year at a realistic 25 basis-point corporate spread. Taking the rouble exposure means carrying a cross whose annualised volatility is 23.1% since 2016 and 31.4% since the war began — between five and nine times the volatility of the rupee–dollar pair the trade currently runs through.

None of this argues against the mechanism. It argues against the justification. Rupee–rouble settlement buys India something genuine — a payment rail no third country can switch off, and continued access to discounted crude that a dollar rail cannot legally carry. Those are strategic goods and they are worth paying for. What the exchange rate makes clear is that India is paying for them. The mechanism is insurance, priced at roughly a billion dollars a year, and the honest case for it is that the insurance is worth the premium — not that the premium is a saving.

12Does the money actually come back? Testing the thesis on Sberbank

Section 4 argued that the way out of the rupee overhang is to give the holder something to buy — to convert a trapped settlement balance into a yielding claim on India. Since late 2025 that argument has stopped being hypothetical. Sberbank, the largest Russian institution in India, has announced a fund giving Russian investors Nifty 50 exposure, an expansion programme, rupee lending products for Indian firms, and the Reserve Bank has quietly rewritten the rule that governs the whole thing. It is the closest we will get to a controlled experiment, and it is worth separating what actually moves capital from what merely looks as though it does.

The consequential item on the list is the one that made no headlines. On 5 August 2025 the Reserve Bank issued A.P. (DIR Series) Circular No. 08, which reads, in its operative sentence:

A week later, on 12 August 2025, Circular No. 09 completed it: persons resident outside India holding an SRVA “may invest their rupee surplus balance in the aforesaid account in Central Government Securities (including Treasury Bills),” with the operating detail folded into the Master Direction on non-resident investment in debt instruments.

Those two sentences are the entire securitisation mechanism described in Section 4, enacted. The first removes the approval gate that made every new rupee-settlement relationship a bilateral negotiation with the central bank; the second turns an idle vostro balance into a yielding claim on the Government of India. This is precisely what Section 9 listed as an underused instrument — and the finding here is that the RBI had already done it eleven months before this report was written, and almost nobody noticed. The gap is not regulatory. It is promotional.

Take the unofficial estimate at face value for a moment, because it points somewhere uncomfortable. If Russian rupee balances were around $3.5 billion against annual Indian imports from Russia of $55.4 billion, the overhang is roughly 6% of one year's purchases — about twenty-three days of trade. That is a working balance, not a trapped fortune. And it fell by more than half in under a year.

Which means the rupee-overhang problem may be smaller and more mundane than the debate assumes, for the least satisfying of reasons: the mechanism is not operating at the scale that would create a large overhang in the first place. Most of the $55 billion is still being settled some other way. A balance of $3.5 billion is what you get when a settlement rail carries a modest fraction of the flow — not when it carries all of it and the money has nowhere to go.

13The $1.3 billion answer: Russian capital was never blocked by the rupee

Section 12 asked whether rupee balances come back to India as investment. There is a way to settle that question completely, and it has been sitting in a government spreadsheet the whole time. The Department for Promotion of Industry and Internal Trade publishes cumulative foreign direct investment equity inflows by source country. For Russia, the figure from April 2000 to March 2026 — twenty-six years — is US$1,320.65 million.

That is 0.18% of India's total FDI equity inflows, placing Russia around 31st among source countries. The implied inflow for FY2025-26 alone is roughly $18.5 million. For a relationship carrying $59.86 billion of annual trade, an eighty-year strategic partnership and the entire rupee-settlement apparatus, that is the whole recorded stock of direct investment.

All three numbers are defensible and they measure different things, but the gap between the first two is the finding. The largest Russian investment ever made in India does not appear in India's Russian FDI statistics at all, because it entered through Singapore. Capital that wants to reach India already has a route, and that route bypasses the rupee, the vostro account and the bilateral settlement mechanism entirely.

Which disposes of a premise this report has been testing since Section 4. The constraint on Russian investment into India was never that rupee balances were trapped or that a securitisation channel was missing. Russian capital has moved into India at scale — $12.9 billion in a single 2017 transaction, before rupee settlement existed — whenever it wanted to, in dollars, through third jurisdictions. When it does not come, the reason is that the investment case is unattractive or the investor is constrained, not that the plumbing is absent. Building better plumbing does not create water.

Every argument in this report for making Russia hold rupees depends on India selling Russia enough that the balance can be spent. Defence is the one sector where India has historically had a genuine two-way industrial relationship — and it is going the wrong way.

Russia supplied 68% of India's arms imports in 2012–16. By 2021–25 it was 40%, with SIPRI attributing the shift to France, Israel and the United States. The direction matters more than any single figure: the sector that could plausibly have absorbed rupee balances through offset and co-production is the sector where India is buying less from Russia every year, and where the flagship co-production project of the last decade — the Ka-226T light helicopter, agreed in 2015 — remains stalled.

Read together with Exhibit 18, the picture is consistent and it is not the one the de-dollarisation debate assumes. Russian capital reaches India when it has a reason to, using ordinary offshore routes. Russian FDI recorded against Russia is $1.32 billion in twenty-six years. The single largest investment became a sanctions liability. The defence relationship that might have balanced the trade is contracting by thirty percentage points a decade. Against all of that, the currency in which the oil invoice is denominated is a second-order question — which has been this report's argument from the beginning, and this is the strongest evidence for it.

Implications