Last week, Ethiopia and its bondholders came to an agreement on the restructuring of its sole $1 billion Eurobond, due to mature in 2024, and in default since end-2023. This does not quite mark the end of the saga – it still needs to be approved as comparable by the Official Creditor Committee (OCC), but its co-chairs, China and France, have indicated that they did not object[1], seemingly avoiding a collapse which occurred in January 2026 when the OCC rejected a deal between Ethiopia and its bondholders was rejected as too generous relative to the one with bilateral creditors.
Is it a good deal? “A good compromise should make everybody equally unhappy”. This quote, alternatively attributed to famous philosophers Larry David or Calvin & Hobbes, describes the recent Ethiopian bond restructuring particularly well. Bondholders have issued a scathing statement[2] and Debt Justice is equally critical[3]. Who is right?
These difficult negotiations also need to be seen in the context of a broader fight on debt restructuring frameworks. A $1 billion Eurobond is certainly small relative to the balance sheet of the funds holding it. The tooth-and-nail negotiations that lasted for 2.5 years indicate that the bondholders’ fight was a statement of principle. Even for Ethiopia, the bond itself was only 3% of its external debt, or 0.8% of its GDP. Ultimately, different proposals has a delta of $150mn, less than a tenth of a percent of GDP.
The bigger questions were architectural, with different actors trying to impose their views upon each other - and Ethiopia caught in the middle[4]. The IMF was perhaps too confident in its forecasts, and did not update them fast enough. Official Creditors wanted to show success after the quagmire of Zambia, and went fast, but then sought to impose outdated terms on private creditors. Bondholders sought to delay an agreement to benefit from improving economic dynamics, and when playing for time failed, they threatened to sue Ethiopia. All of this was hampered by slow communication of information[5].
There might be an upside: the process was indeed painful and frustrating for all parties involved, but the result is actually elegant, and the innovative warrant that emerged might inspire future deals. This blog – which will be followed by a longer analysis of the process and lessons for restructuring – seeks to present the main lessons of the deal.
A bit of background
When Ethiopia initially applied for the Common Framework in 2020, the assessment was relatively simple: its reserves had fallen below 1 months of imports and needed to be reconstituted. Exports were insufficient to restore the balance, and a rescheduling of claims was necessary as debt service was concentrated in 2021 to 2024. Debt stock was not that high (60% of GDP, half external) – and it was mostly concessional. The civil war in Tigray – in addition to its immense human toll, one of the deadliest in recent history, with over 600 000 casualties – also caused the economic situation to deteriorate considerably. Reconstruction was necessary, but was hampered by the debt burden.
After the conflict ended in November 2022, restoring economic stability was an important objective, and it required restarting an IMF program and restoring debt sustainability. It took close to two years to negotiate, around a set of major reforms, centred around a devaluation of about 100% and floating of the currency. In July 2024, an IMF program was approved, with debt restructuring as a precondition, in two regards:
- The country required $10.5 billion in financing from 2024 to 2028, of which the IMF would provide $3.5 billion, the World Bank and others $3.8 billion. Debt needed to be reworked to reduce new debt service by an additional $3.5 billion in that period.
- And more controversially, it judged that the value of external debt at the end of the program would be too high to restore sustainability. While exports were projected to rebound from $13 billion in 2024 to $19 billion in 2028, the present (discounted) value of debt was projected at $29 billion, or 154% of exports, above the 140% it considered as safe. While small, this breach was at the center of the 2 year controversy that ensued.
A controversial process
Official bilateral creditors had formed a committee (“OCC”) as early as 2021, co-chaired by China (which held $7.4 bn in loans, already rescheduled under a previous agreement in 2018) and France (the Paris Club collectively owned $2b in loans). Under pressure to show that the Common Framework was able to provide fast and sufficient debt relief, they accepted the IMF’s analysis and provided financing assurances in July 2024, followed by an Agreement-in-Principle in March 2025, setting out the terms of the restructuring, and putting in stone the necessity of a haircut in the value of debt.
Bondholders vehemently disagreed with the process. They held a minor claim on Ethiopia - $1 billion, or 3% of external debt. A smaller claim can provide leverage: if large creditors settle and debt becomes sustainable, a small creditor can receive its claim in full. The purpose of the comparability of treatment principle is precisely to avoid such free-riding, and became the center of the controversy.
In end-2023, Ethiopia had defaulted on its Eurobond due in 2024, under pressure of the OCC which had granted a standstill while the IMF program was negotiated, and did not want other creditors to be paid in the meantime. Bondholders were also convinced that the IMF program was low-balling exports figures, and that the value would quickly rebound to levels that would not show any need in reduction in the present value of debt: a simple rescheduling of claims would be sufficient.
Their strategy was thus to wait to be proven right, and get to a deal that they would accept. In mid- to late-2025, it became clear that their analysis was correct: the IMF forecasts had indeed been too pessimistic, especially on exports (which was the most binding indicator of the capacity to repay). Led by a higher-than-expected volume of gold exports combined with a sharp rise in gold prices and a comparable boost in coffee exports, Ethiopia’s overall goods exports were twice as large as the IMF had expected just a year before.
A first settlement reached between Ethiopia and bondholders in January 2026, provided a reduction of the bondholders’ nominal claim of 15%, excluding past due interest. It also added a Value Recovery Instrument (VRI) that was tied to Ethiopia’s good exports, with an upside of $180m if they outperformed the IMF’s assessment. Overall, this would make the overall deal equivalent to no haircut. But by the time this deal was reached, Ethiopia was very much on track to outperform the IMF’s original assessment. In turn, this made the bondholders’ deal clearly in violation of the “comparability of treatment” with respect to the official creditors’ deal.
When the IMF subsequently published its review in January 2026, it acknowledged the reality of improved growth and external conditions and — technically — no need for haircut-based debt relief. Yet since the OCC had already agreed to debt relief, “comparability of treatment” required the same from private creditors. The OCC rejected the private creditors’ deal, sending Ethiopia and its bondholders back to the negotiating table. Bondholders chafed at the official creditors’ rejection of their deal and initiated legal proceedings — threatening to sue Ethiopia for full repayment of the defaulted Eurobond
The threat of litigation was always just a threat to put pressure on the official creditors. In the background, negotiations between bondholders and Ethiopia continued. In May 2026, another round of negotiation failed, this time rejected by bondholders as too stringent. The Ethiopian proposal entailed a lower nominal reduction of 12%, but payments would be more backloaded toward 2029, and did not include a recovery instrument. The bond would be amortized in 4 instalments, until end-2029. The rejection set off another round of legal threats.
The agreement
Only one month later, a final deal was reached. In terms of process, it was validated by the IMF as matching the parameters of its program and by the co-chairs of the OCC as comparable, before its publication. The road does not quite end there: other bondholders have to approve, other OCC members as well. Other official creditors such as the UAE still need to reach an agreement. Nevertheless, it does seem that Ethiopia is not far from a full settlement on its debt stack.
The final agreement is halfway between deals rejected by the OCC in January and by bondholders in May. It has two parts: a new bond, which is identical to the one proposed by Ethiopia in May, plus a warrant providing some potential upside to bondholders. The new bond has a nominal value of USD 880 million (a 12 percent nominal haircut), a 6.15 percent coupon, amortisation through July 2029, full payment of the USD 99.375 million in past-due interest plus a long-coupon covering 2024-2026, and a 0.5% consent fee. At a 5% discount rate, this represents about 10% in present value loss.
The main addition compared to the May proposal lies in the “New Money Warrant”, a separate, tradable security granting holders the right, but not the obligation, to subscribe at par to a future Ethiopian Eurobond (up to USD 1 billion, allocated one-for-one against existing holdings) carrying a coupon fixed at 450 basis points over six-year U.S. Treasuries at the time of issue. The future bond would run seven years with a six-year average life. Whether exercising the warrant - buying this bond at par - is valuable or worthless depends on Ethiopia’s borrowing conditions at the time of the strike. Its value is capped: it can be brought back at a maximal price of $90m (the actual price is set by a determination agent, but that part of the deal is not specified). In other words, if bondholders trigger the issuance of the New Money Warrant but Ethiopia does not wish to issue it — presumably because it can issue a vanilla Eurobond within 450 basis points — Ethiopia could redeem the bond by paying the bondholders $90m.
In terms of COT, we are typically in the grey area where there are trade-offs between criterions: bondholders get a lower haircut in exchange for more relief during the program. Usually, bondholders provide less cash relief (they like to be paid upfront), but in this case the bond was due very early in the period. Because the bond was maturing during the program, any extension reduces cash flow during the program period[5]. As a result, the key question was how much NPV loss they needed to provide to be comparable with the 12.5% provided by the OCC. In May 2026; bondholders had rejected the offer of a 10% loss. With the warrant, the OCC probably estimated that even in the upside of a cash payment of $90m in 2028, there was some (small) NPV loss, sufficient to match the fact that the debt service reduction during the program was larger for bondholders.
In June 2026, the deal was clinched thanks to a new instrument, a warrant with a right to lend to Ethiopia immediately after the program. There is no downside[6], and thus its minimal value is 0, and its maximum value is $90m in 2028-2029 as Ethiopia can offer to redeem it outright at a price to be determined (and whose formula was not disclosed), but capped at $90m.
Its principle is elegant, as it is based on market indicators rather than more ambiguous or political triggers. The warrant holders will exercise it if they estimate that 450 bps spreads is a better-than-reasonable compensation for their risk. In mid-2028, there will not be a market signal to guide this assessment, so it will come from ratings or market views. In a way, it is akin to a VRI: if Ethiopia is doing well, its (virtual) spread will be low, and the warrant gains in value. Back of the envelope calculations indicate that the value grows linearly as the risk premium falls: it reaches $90m around 280 bps (which would be 6.8% with current UST yield). The elegance comes from the fact that its upside is not based on manipulable values, or with poor correlation with actual ability to pay.
It avoids the pitfalls of regular VRIs. In the recent past, the Zambia deal was passed on the World Bank/IMF composite indicator, an opaque, manipulable and non-continuous assessment, perhaps the worst imaginable. The deal with Sri Lanka includes a bond whose value varies with GDP expressed in USD, which could be poorly correlated with actual capacity to pay. The initial VRI proposed in Ethiopia was linked with goods exports value in USD. This was technically sounder than the Zambian and Sri Lankan VRIs as it is a more natural metric of capacity to repay, but it was still imperfect. Bondholders were tying Ethiopia’s debt service to the one line item that was clearly rising. Moreover, Ethiopia would forfeit the highs of a commodity cycle over the coming years, as rising gold and coffee prices would trigger higher debt service, and subsequently suffer the lows of a possible commodity price fall.
In the case of the New Money Warrant, the structure is natural. The warrants avoid such difficulties: markets will look through the risk at a 6 year horizon, and as imperfect as they are, they assess risks in a way that will reflect Ethiopian probability of default. If exercised, those warrants will provide expensive, but useful liquidity to the government, after the end of the IMF program. These qualities could make such an innovation durable for other restructuring cases.
Yet, it is not without problems, either: the spread where they gets activated is too high, the instruments will be illiquid and might lead to litigation. A spread of 450 bps is probably too high: currently, it would price above Nigeria (7.5%, or 350 bps above UST) and close to Kenya (8.5%, 450 bps) and just below Bolivia (8.8%, 480 bps). This corresponds to about B to B- rating, which is where Ghana emerged from its restructuring[7]. It will also be a complex instrument, only traded among a small group of specialist funds, leading to distortions rather than a useful indicator of country risk it would ideally be. Finally, some terms are still to be determined, including, importantly the pricing mechanism. When and how the warrant can be bought back, and when and how it can be exercised could lead to tensions.
Conclusion
This deal allows the Ethiopian ministry of finance to remove the Damocles sword of a lawsuit (even though how actually tranchant that sword would be is another question, since a judgement against a sovereign is notoriously hard and expensive to enforce). It can now redirect its focus to economic growth and stability. Does it reduce the fear of bond bogeymen in future debt restructuring processes? In part only: the process needs to be reformed to be able to reach faster deals.
Some marginal fixes, such as a faster disclosure of the COT parameters of agreements with the official committee (recommended by the Global Sovereign Debt Roundtable at its April 2026 meeting) whould have helped. But the lessons are that broader changes are needed: in particular, a much more parallel process, à la Hagan-Setser would have been helpful. Finally, it has strengthened the argument in favor of statutory protection, even if in very limited and defined aims, to avoid using lawsuits as a negotiating tactic.
[1] Appreciate the language: “[They] provided their non-objection, subject to approval by the wider Official Creditor Committee.”
[2] Among choice quotes: “[The] process to have exposed broad flaws in the architecture of sovereign debt restructuring “
[3] “Bondholders have successfully used the threat of legal action in the UK to wring more money out of the Ethiopian people.”
[4] During #DebtCon7 in Paris, the head of the Zambian Debt Management Office quoted the African proverb: “when Elephants fight, it is the grass that suffers"
[5] For instance, an Agreement in Principle of official creditors in March 2025, setting out the main parameters of the deal. But it took 6 months for private creditors to know what those parameters were, and their publication occurred in October 2025.
[6] The FDL team is confused as how to reach 48% - which seems to exclude the $300m payment on July 15th, 2028. Yet, the IMF program comes to a close at the end of July.
[7] Unlike the VRI agreed in January 2026, which was “symmetrical”, but the downside scenario was so unlikely that de facto it could be considered as irrelevant.
[8] Using OAS arrives at slightly lower values – in this case, 450 bps would be closer to Congo, Ecuador and Argentina, all rated B-