In the hills of Ituri, in the conflict-scarred northeast of the Democratic Republic of the Congo (DRC), Ebola is once again killing people, and this time the system built to stop it is visibly failing. The outbreak confirmed in mid-May 2026 is the DRC’s seventeenth since 1976, but it is the first of the post-USAID era. By […]
In the hills of Ituri, in the conflict-scarred northeast of the Democratic Republic of the Congo (DRC), Ebola is once again killing people, and this time the system built to stop it is visibly failing. The outbreak confirmed in mid-May 2026 is the DRC’s seventeenth since 1976, but it is the first of the post-USAID era. By early June, the official count stood at well over 500 confirmed cases and around 90 deaths, with the true figure widely believed to be far higher. It is caused by the Bundibugyo strain, for which there is no licensed vaccine and no specific treatment, in a province where close to a million people are already displaced by fighting.
The outbreak was not just predictable, but predicted. Not because of the nature of the virus itself, but because the systems and organisations that could have seen it coming and responded effectively had been substantially weakened. By the time the first patient fell ill, the community health workers, contact tracers and laboratory logistics networks that sustained every previous Congolese Ebola response had been defunded for months. US foreign assistance to the DRC fell from $1.4 billion in 2024 to a small fraction of that in 2026, and humanitarian funding for the affected areas collapsed from over $900 million to under $200 million in a single year. USAID itself closed in 2025, and the United States withdrew from the World Health Organisation (WHO) at the start of 2026, leaving Washington largely outside the information flows it once helped lead. Aid workers and former officials have been blunt: the cuts very probably delayed detection, and the cost of that delay is measured in lives lost.
This is what the breakdown of an imperfect but functioning system looks like. The global health architecture that had evolved over decades was never elegant. It was duplicative, donor-driven and politically awkward. But it worked often enough. That settlement is now fundamentally gone, and Ituri is a sign that progress is being reversed.
The end of the world the SDGs were built for
When the Seventy-ninth World Health Assembly (WHA) closed in Geneva in late May, it did something that would have been unthinkable a decade ago. It launched a formal, Member State-led review of the global health architecture itself; an admission that the system built to deliver global healthcare is no longer fit for the world in which it now operates. Like the UN80 reform agenda, to which the Assembly tied the process, it is a tacit acknowledgement of a new global settlement.
The Sustainable Development Goals (SDGs) were a product of a particular moment, a broadly liberal, broadly multilateral consensus in which the principal question was how to coordinate willing actors around shared targets. That world is gone. The withdrawals and funding cuts made visible in Ituri are the markers of a wider retreat. The UK abolished DFID in 2020 and has since announced that aid will fall to 0.3% of national income by 2027, its lowest level since 1999 and a long way below the 0.7% commitment that still sits, unhonoured, on the statute book. The WHO has trimmed its 2026 to 2027 budget from $5.3 billion to $4.2 billion, and even that leaner figure carries a financing gap which is reported to be close to half of what’s needed. It is notable that the major economies of Brazil, China, India and Indonesia all declined to sign the April Joint Political Declaration on architecture reform. Bilateral health deals are proliferating in place of multilateral commitments. The centre, such as it was, is not holding.
The comparative advantage of health institutions
Agulhas recently had the privilege of working with MOPAN, the multilateral performance network, on a study of comparative advantage in the multilateral health ecosystem, contributing to the global debate surrounding it. We mapped the mandates and practices of nine multilateral organisations, namely Unitaid, the Global Fund, Gavi, WHO, UNICEF, UNFPA, the GFF, the Pandemic Fund and the World Bank, against the six priority functions proposed for a reformed architecture. The study was presented at WHA79 and now feeds directly into the WHO-hosted Joint Process. Two findings from that work seem to be especially relevant to the challenge ahead.
The first concerns how organisational mandates grow. Again and again, we found that mandate expansion across these organisations has tracked donor priorities rather than system needs. In a generously funded world, that produces overlap and a degree of healthy redundancy. In a world facing a projected 19 to 33 per cent fall in global health financing, dominated by earmarked contributions and shifting toward domestic financing in lower- and middle-income countries, the same pattern produces something far more dangerous: a set of institutions optimised for the preferences of shrinking donors, rather than for the coherence of the system as a whole.
The second concerns how functions connect, or fail to. Our analysis of product research and development (R&D), market shaping and access, traced the division of labour in which Unitaid creates market innovations and the Global Fund deploys them at scale. It is a genuine strength of the current system. But a drug-resistant TB case studyii showed how that same chain of sequential dependencies can fail when there is no joint planning mechanism holding the actors together. The innovation arrives, the scaling does not follow, and the patient waits. When financing was abundant, the architecture was forgiving. But no more.
Ituri is the live demonstration. Surveillance, R&D, supply logistics and frontline response are distinct functions performed by distinct actors, and the system only saves lives when they connect. Pull the funding from the first link – detection – and even the most sophisticated downstream capacities never get the chance to act. A reformed architecture that does not maintain the connective tissue between functions will reproduce exactly this failure.
Why architectural reform will be so difficult
The objective of reform is easy to state and brutally hard to achieve: a system that delivers the priority functions of global health coherently, equitably and affordably, with clear accountability for who does what, under the conditions of declining resources and contested leadership.
The difficulty will not primarily be technical. We broadly know which functions matter, and roughly who is best placed to perform them. The difficulty will be structural, and one that political theory has described long before global health existed as a field.
Health security is a global public good; non-excludable and non-rival. Once early warning of an outbreak like that in Ituri exists, every country benefits whether or not it paid for surveillance, and one country’s protection does not diminish another’s. This combination produces the classic problem of collective action. When everyone benefits regardless of who funds it, each state can reason that frontline capacity is costly to provide, so the individually rational move might be to let someone else pay. The trouble is that when the largest contributors withdraw, and the countries left exposed cannot fund the capacity themselves, the good does not simply go under-provided. It collapses, and the risk falls on everyone.
This is the prisoner’s dilemma, which assumes two rational actors who can, in principle, be brought to cooperate once they see that walking away leaves them all worse off. The harder truth of a multipolar world is that this assumption no longer holds.
For many states, foreign policy has become more transactional, conducted in the language of the deal: bilateral, short-term, and judged by what can be extracted now rather than what compounds over a decade. Realpolitik is messier than any game, and it is increasingly being played by states that are not acting in ways the theory expects, nor trying to. When the players stop behaving rationally, the reassuring logic that suggests cooperation can always be recovered, stops being a reliable guide.
That makes the standard escape route narrower than it looks. The classic answer to the dilemma is repetition: when actors expect to meet again, cooperative strategies built on reciprocity can outperform pure self-interest, because reputation and the prospect of future dealings change the present calculation. International institutions are precisely the machinery that keeps the longer-term future in view, as they were built to do, as a result of the global failures of the past.
The underlying problem is older than global health itself: how do sovereign actors cooperate when there is no higher authority to enforce a bargain? The durable answer has been that such institutions, by turning one-off dealings into repeated ones, can hold the line without a world government to compel it. The WHO, the treaty regimes and the pooled financing mechanisms make interactions repeated and visible, lower the cost of monitoring who is contributing, and let reciprocity and reputation accumulate. The transactional deal approach does the opposite. It shortens that horizon deliberately, trading the slow returns of membership and reputation for an immediate, legible win. Where a state has decided that the institution itself is the constraint, repetition loses its grip.
Seen this way, the danger of the present moment is not only the loss of a large funder. It is that the preference for bilateral deals each power can control collapses the repeated-game machinery that made global cooperation make sense in the first place, pushing actors toward a deal-based logic in which leaving everyone to fend for themselves dominates, and in some cases beyond even that, toward a politics that does not treat the collective payoff as a goal at all.
A deal can be struck and unstruck in a news cycle; an architecture has to hold for decades. Reform therefore requires agreement among actors who no longer share a single theory of why the system should exist, who increasingly appear to prefer arrangements they can hold in their own hands, and who are being asked to consolidate and cede ground precisely as their own fiscal space contracts. Rationalising an architecture means someone, somewhere, doing less, and performing less functions, and at the moment no institution and no donor is volunteering to be that someone.
That is the gap between the architecture we have and the architecture we need. It will not be closed by goodwill, or by another declaration. It will in part be closed, if at all, by patient, evidence-based work that shows precisely where the functions sit, where they overlap, where they fail to connect, and what a more coherent configuration would actually require of each actor. It will also need the competitive organisations that have expanded their objectives to win resources, to be honest about their role in the wider ecosystem, and to change their behaviour for the benefit of the whole. And it will be a messy business of trade-offs, gaps and contestation.
The MOPAN comparator study has given us a vital view of what we have. The work ahead must help tell us how to build what we need, and to do it in time.
End Notes
[1] The Guardian, ‘US is ‘simply choosing not to stop’ Ebola outbreak after massive public health cuts, experts
say’, May 2026
[2] MOPAN, ‘MOPAN PERFORMANCE INSIGHTS Comparative Advantage in the Multilateral Health Ecosystem Output 1: Mapping of Organisational Mandates Against Future Priority Health Functions, Case study 2, p31
