Thesis01How We Invest02Portfolio03About04Research05Pitch06Contact07
LAVA · Africa & Global
← Research
cross-border paymentsstablecoinslocal stablecoinsFXpaymentstokenizationAfricaemerging markets

Clearing Without The Politics

How Africa can clear cross-border payments without a common settlement unit. Respect open graphs that route across local currency stablecoins, tokenized fund units, and mutual credit, with legal finality anchored in regulated financial market infrastructure.

Some friends recently asked us what we meant when saying we look for genuinely new financial technologies to invest in. Not just "cheaper, faster, better". Not just orchestration with stablecoins and a bit of programmatic treasury management bolted on top. Not just solving fragmented liquidity in underserved African corridors with slightly better routing algorithms.

Genuinely new.

What follows is an attempt to describe some of the design space we see, and to shift your Overton window such that you, too, begin to think bigger than dollar-denominated stablecoins.

Key Ingredients

The major point we want you to take home from this is that the feature we’re after is “oneness of money”, not one kind of money. That is, my money must have the same quality and standing as the money used by the central bank. It must participate on equal footing in the same technological medium.

What we do not need is one kind of money, like the eco or the East African Shilling or the dollar: each participant ought to be able to transact in the unit they prefer for whatever reason they like. Those units just need to clear in the same graph, because they each have the same standing and quality, but not necessarily the same price.

The power and the problem of clearing

At Bretton Woods in 1944, Keynes arrived with a proposal for the International Clearing Union. Its unit of account was the bancor. No nation would hold bancor as reserves; no citizen would ever touch one. It existed purely on the books of the Clearing Union, as the unit in which trade surpluses and deficits between member central banks were recorded and offset. Every country got an overdraft facility sized to its share of world trade. Radically, both excessive creditors and excessive debtors paid charges on their positions. Germany running a persistent surplus would be penalised just as Greece running a persistent deficit would be.

Keynes' insight was that international trade does not require a global reserve asset at all. It requires clearing. Most flows offset each other; only residuals need settling; and if the system is designed well, the residuals are actively discouraged from accumulating. Money, at the international level, could be a pure balance-sheet phenomenon: a unit for recording and extinguishing mutual obligations, rather than a thing to be hoarded.

But Keynes lost. The White plan won, the dollar became the world's reserve, and every peripheral economy on earth has been paying the toll ever since. Nowhere pays it more visibly than Africa: a continent of roughly forty currencies where the majority of intra-continental trade is still invoiced, routed, and settled through dollars in correspondent accounts in New York and London, at spreads that would be considered a scandal anywhere else.

One reason Keynes lost is that his plan required everyone to agree on a settlement unit and, therefore, on who governs it: who sets the parities, who sizes the overdrafts, who levies the charges, whose exchange rate moves when the index says it must. These are not just technical questions, they are political.

The politics has never worked out: not at Bretton Woods, not in the Eurozone (which is what a hard-currency clearing union without symmetric adjustment looks like in practice), and not in the decades of African monetary-union roadmaps that remain perpetually a decade away. ECOWAS has been announcing the eco since 2003. The East African shilling has been imminent since 2013.

Clearing is the powerful part. The agreed-upon unit is what consistently fails.

The interesting question is whether we can have the first without the second.

It turns out that we can. It just took eighty years, two obscure European experiments, and some graph theory to show how.

Liquidity lives in the cycles

In 1991 Slovenia declared independence, lost the Yugoslav market overnight, fought a ten-day war, and watched inflation hit nearly 250%. GDP fell 8.9% in 1991 and another 5.4% in 1992. Liquidity vanished.

So, Slovenia’s payments agency began running a monthly multilateral set-off of invoices between companies: software now known as TETRIS Core Technologies. Firms submit their outstanding obligations; the algorithm assembles them into a single directed graph; and it then finds the largest structure of overlapping cycles that can be extinguished simultaneously. In 1992, the first full year of operation, obligations equal to 7.58% of GDP were cleared this way, with no money changing hands. The system still runs today, quietly and counter-cyclically: in the 2012 liquidity crisis it cleared €683 million (1.89% of GDP) across 14,000 companies; in the good year of 2019, €209 million. Demand for it rises exactly when liquidity disappears.

Why does this work? Because an economy's payment obligations are not a list; they are a graph, and a strongly connected one. Fleischman, Dini and Littera analysed a full year of B2B invoice data from the Sardex network in Sardinia (3,199 firms, 138,378 transactions) and found that the entire network formed one strongly connected component: there is a path from any firm to any other. There are no clean supply chains in real economies; there is, in their phrase, a Gordian knot of mutual indebtedness. And in a strongly connected graph, every firm sits on cycles: A owes B owes C owes A, extended and intertwined across thousands of participants.

A cycle of obligations requires zero external liquidity to discharge. If A owes B 20, B owes C 20, and C owes A 20, then sixty units of debt can be extinguished by a simultaneous set-off in which nobody pays anybody anything. However, no individual firm can see the cycle, and no sequence of bilateral payments can execute it without someone fronting cash first. Cycles are invisible from inside the graph and undischargeable one edge at a time. That is what a liquidity crisis is at the micro level: gridlock, payments that cannot be settled individually but can all be settled simultaneously. It takes as little as 3% of payment obligations sitting in cyclic structures to start propagating instability through a financial system.

Contrast this with Sardex, where an average of 23% of the total value of invoices (touching over 40% of all invoices) could be cleared with no liquidity whatsoever. A quarter of the economy's payment traffic was, in effect, already paid. It's just that nobody could see it.

Hence: “liquidity is in the graph”. It is not something you import from a bank, a donor, a Eurobond, or a stablecoin float. A very large fraction of it is latent in the structure of who already owes whom, retrievable if you can see the whole network and execute set-offs atomically.

For centuries, the only institutions able to do this were closed clearing clubs: banks clearing with banks, behind membership walls, precisely because netting requires exposing your positions and nobody exposes positions to competitors without rules. The general public, and every African SME, has always been locked out. That is a technology problem now, not a fact of nature. We will come back to it.

Mutual credit: or how to stop worrying and double the clearing

Set-off clears the cycles. Credit clears the chains (the acyclic residue left over). This is the second ingredient, with a lineage running from the Swiss WIR (founded 1934, in the teeth of the Depression, still operating) through to Sardex (founded 2010, in the teeth of another one).

Mutual credit is disarmingly simple. Members of a circuit start at zero. When A buys from B for 100 credits, A goes to −100 and B to +100. The sum of all balances is always exactly zero; the "money" is created by the act of trade itself and destroyed when negative balances are repaid through sales. Credit lines are modest (Sardex assigns roughly 2% of a member's annual turnover) and balances bear no interest in either direction, so hoarding is pointless and velocity is ferocious: around 5, against roughly 1 for the euro. Backing is not an asset in a vault; it is each member's contractual commitment to sell goods and services into the circuit, at around 10% of turnover. This is a 5:1 ratio of committed productive capacity to extendable debt, which is the inverse of a speculative bubble.

Fleischman, Dini and Littera re-ran the Sardex obligation graph with a simulated mutual-credit facility attached, modelled as just another node in the graph, whose credit lines create new edges and therefore new cycles. The result: obligation-clearing alone discharged ~25% of the network's debt; obligation-clearing plus mutual credit discharged close to 50%. Half the payment obligations of a real 3,000-firm economy, settled with no fiat money, using credit lines worth 2% of turnover.

Sit with that for a moment: you can clear roughly half the payments in a graph, with no money.

Importantly, the liquidity source was not privileged, not central, not sovereign. It was a node. It participated in cycles like everyone else. Which raises the question: if a liquidity source is just a node, why should there be only one that we all have to agree on?

Clearing without a unit

So, what can we learn from designs like Cycles?

First: there is no settlement unit. Obligations are denominated in whatever unit the two parties chose: naira, rand, cedi, shillings, dollars, tokenized fund units. Acceptances and tenders express, per participant, which liquidity sources may be used to discharge them. We can separate the two functions that modern currency conflates: the unit of account (what debts are denominated in) and the medium of exchange (what discharges them, here and now). Nobody votes on a unit. Nobody governs a parity grid. There is nothing to fight over at a summit, because all any participant ever expresses is their own balance sheet and preferences.

Second: liquidity sources are just nodes. In the graph, holding an asset is being owed by its issuer: your bank balance is the bank's obligation to you; a stablecoin balance is the issuer's obligation to you. So banks, mutual-credit circuits, stablecoin issuers, DeFi lending pools, and money-market funds all enter the graph the same way any firm does: as nodes with obligations radiating out and acceptances radiating in. The idea that firms are liquidity sources and liquidity sources are firms sounds like wordplay, but it really matters. It means there is no privileged settlement institution whose balance sheet everyone else must route through, and therefore no argument about who gets to be it, or how to govern it.

Third: multiple currencies clear side by side without any exchange. A single settlement flow can traverse several liquidity sources at once: a USDC cycle here, a shilling cycle there, coupled through firms that sit on both. Each currency circuit operates independently with no FX conversion anywhere. This sounds unbelievable to Africans, but it is possible. A firm at the intersection of two cycles benefits from set-off in both without handling either currency. Where a genuine exchange is useful, an exchange is just one more node offering tenders and acceptances in two currencies.

The beauty of thinking through cycles is this: so long as a small number of participants are willing to hold a currency, a much larger group can benefit from it.

MTCS shows the clearing works at national scale on real data. Mutual credit shows the liquidity source can be endogenous and node-shaped. Cycles shows the whole thing can run as an open protocol with no unit at all: every participant declaring their own acceptances, every debt denominated in what its parties can actually pay, and the graph, not a treaty, deciding which currencies matter.

The African Graph

How does all of this apply to Africa?

Hard-to-source currencies stop being blockers. The reason a Ghanaian importer settling a Nigerian invoice routes through dollars is not that either party loves dollars; it is that the bilateral naira–cedi market is too thin to price. In an obligation graph, that trade does not need a naira–cedi market. It needs a path (through any chain or cycle of firms, banks, and liquidity sources whose acceptances connect the two) and paths are exactly what a continent-scale graph has in abundance and bilateral FX markets lack. Recall: the denser and more strongly connected the graph, the more of it clears with no liquidity at all. Intra-African trade is strongly connected, even if the payments systems aren’t.

Widely accepted currencies become busy nodes, not anointed ones. Take the Kenyan shilling. Many companies are currently willing to hold KES. At the time of writing, it ranks among the most stable currencies in the world: Bloomberg measured its volatility over the past year at roughly 1.5% (among the top five globally) supported by $4bn+ in annual diaspora remittances and reserves above five months of imports. In graph terms, that willingness is a dense mesh of acceptances pointing at one node, and the solver will naturally route large volumes of settlement flow through it. The shilling would be one of the busiest nodes in the African graph.

But, it is not historically the least volatile currency: it fell 21% against the dollar in 2023, its worst year in three decades, before central-bank action and a Eurobond restructuring produced the current calm. Any architecture that had anointed the shilling as its settlement unit in 2022 would have imported that whole drama into its core. In an open graph, nothing is anointed, so nothing breaks. Acceptances for a wobbling currency shrink, flows re-route through other nodes, and the currency's centrality is repriced continuously by the only mechanism that has ever worked: thousands of participants updating what they will accept. A busy node is an emergent fact, revisable every clearing epoch. A settlement unit is a constitutional commitment, revisable only by crisis. The distinction is the fulcrum of this entire essay: please take a moment and reflect on this.

Politics are minimised. Every prior attempt at African settlement integration has had to answer: whose currency wins, whose central bank anchors it, whose banks get the franchise? Even PAPSS (the most serious effort to date, and a real advance) is an arrangement between central banks and commercial banks in which net residual balances are still backstopped and settled in hard currency through Afreximbank. It is a better-managed version of the hub; but the hub remains, and so does the queue of governance questions that come with it.

In an obligation-graph architecture, a central bank, a tier-one Nigerian bank, a Kenyan fintech, and a Sardinian-style SME circuit can all participate through the same primitives. Banks do not lose their role (they are superb liquidity nodes and credit underwriters) but they hold no gate.

Nothing about such open clearing protocols requires forty governors to agree on anything. The reason this class of system has never gained traction is political, and this design sidesteps the political question rather than answering it.

Finality, and where it comes from

One real world problem we know too much about in Africa is that a settled payment is only valuable if it finally and irrevocably extinguishes the debt: enforceable in court, opposable in insolvency, recognised by an auditor and a tax authority. The clearing literature has partial answers that rest on private obligation law (the UNIDROIT principles), and TETRIS worked because a state agency ran it. Neither travels well: private set-off binds only the parties who signed up, and a state agency binds only its own jurisdiction.

The standard that does travel is the BIS Principles for Financial Market Infrastructures: a well-founded legal basis in every relevant jurisdiction. It is the bar every corporate treasurer and central bank will hold a clearing system to. Blockchains supply atomicity: everything in a cycle settles, or nothing does. But atomicity is a property of computation, and finality is a property of law. An open clearing protocol that stops at atomicity will clear invoices between consenting SMEs forever and never touch a corporate treasury.

So the genuinely new thing is not the graph alone. It is the graph settling in assets that already live inside the regulated perimeter. Recent South African work shows a pattern that could be replicated across the continent. Units in a large, regulated money-market fund are tokenized through a stack of regulated vehicles, such that the resulting cryptoasset is a direct, ring-fenced claim on the fund itself, and the ledger it lives on is operated by licensed financial institutions.

Local, jurisdiction-specific instruments can emerge from regulated financial market infrastructure in each country. Such instruments can be especially busy nodes in the pan-African graph when blessed by local regulators, without needing other regulators or central banks to agree on their construction or aim, which is critical in a place as diverse as Africa.

A stablecoin, by contrast, is a corporate liability made bearer-transferable; its finality is only as good as its issuer, its reserves, and its redemption queue. An instrument that is a direct claim on fund units, validated by the regulated institutions that hold them, imports the finality, insolvency treatment, and supervisory perimeter of the existing system, while remaining a programmable node in an open graph.

This is what resolves the apparent contradiction in "open, regulated infrastructure". The network is open to anyone with obligations to clear: any firm, not just banks, down to the SME circuits whose mutual credit doubles clearing capacity for exactly the businesses no correspondent bank will ever serve. What is permissioned is not participation but validation. It is open at the edges, where the obligations live; regulated at the core, where finality is guaranteed by accountable parties.

Genuinely new

Anything new takes some effort to understand. This is our latest submission:

A pan-African settlement graph in which every participant (bank, telco, fund, SME circuit, trader) can get its payments settled without recourse to the Dollar, or some pan-African settlement unit. Clearing events discharge the maximum debt with the minimum liquidity, routing through whichever currencies and tokenized fund units the graph itself has made busy. No dollar in the loop unless two parties want one there. No summit, no anchor currency, no continental central bank as a precondition. Finality supplied not by cryptography alone but by instruments that are emergent properties of locally regulated FMI, validated by the accountable parties.

Keynes was right that the world needs clearing, not a hegemon's IOU. He was wrong about needing bancor to do it. Slovenia proved clearing at a national scale. Sardinia proved mutual credit improves clearing. Cycles shows how to run it as an open protocol with no unit at all. And the South African structure supplies the last ingredient: legal finality.

The clearing is the point. The unit is a trap. The design space for inter-African settlement opens up the moment you stop trying to agree on the unit, and start respecting the graph.

Bibliography