Decolonizing Research — Insights
Source: Decolonizing Research: Data Colonialism and the Path to Sovereign Digital Research Ecosystems, Carlos Vargas, Societās Partnerships S.A., September 2025.
The more connected a Global South research system becomes to the global digital infrastructure, the less control it may have over its own knowledge. That is the paradox at the centre of this report.
Digital infrastructure is the engine of modern science — high-performance computing, shared datasets, real-time sensor networks, global collaboration. But for much of the Global South, access to that engine arrives almost entirely through centralised, foreign-owned platforms, and the result is not autonomy but a new form of dependency. The data a nation's researchers generate is stored, processed, and monetised under someone else's jurisdiction and on someone else's terms. There is a name for this dynamic — data colonialism — and the response to it is not retreat from global science but the deliberate construction of sovereign digital infrastructure.
What follows are seven observations drawn from the report. The full report develops each in depth, including the technological pathways and the multi-pillar framework for action.
Read the full report
The complete analysis, the three comparative case models, the technological pathways, and the four-pillar framework, fully referenced.
Go to ReportRead the companion report, The Sovereign Leap
The firm's briefing on why Global South universities must lead the transition from technological dependence to digital sovereignty.
Go to BriefingDigital infrastructure is never a neutral tool. It carries power with it
Cloud and connectivity are not neutral utilities. Infrastructure is inseparable from the policy, economic, and geopolitical structures it sits inside, and the Global South's reliance on centralised, foreign-owned cloud reveals a power asymmetry that echoes older patterns of domination.1 The choice of where data lives and who processes it is not a procurement decision; it is a question of who holds power over the knowledge produced.
That reframing changes the question a research system faces. It is no longer "which platform is cheapest and fastest" but "what does this platform choice mean for our long-term capacity to govern our own science." These are different questions with different answers, and conflating them is how dependency becomes invisible.
¹ World Economic Forum, "What is digital sovereignty," 2025; Couldry & Mejias, "Data colonialism: rethinking big data," LSE, 2019.
"Data colonialism" is not a metaphor. It is a described mechanism
Data colonialism is the systematic appropriation of personal, social, and communal data by global corporations and states — a capitalist accumulation by dispossession that treats social life itself as a raw resource simply there for the taking, much as historical colonialism treated land and bodies.2 The mechanism runs through ordinary instruments: an End-User License Agreement quietly dispossesses data producers of the right to own and control what they generate, and the value flows elsewhere.
The examples are concrete and uncomfortable. DNA data collected from Indigenous peoples has been critiqued as an extension of colonial extraction, where researchers may advance their own careers rather than serve the communities the data came from; predictive-policing systems have reinforced existing racial biases in ways that mirror historical patterns of segregation.3 Naming the mechanism accurately is the precondition for designing infrastructure that does not reproduce it.
² Couldry & Mejias, LSE, 2019; Purdue Critical Data Studies, "Data Colonialism." ³ Decolonizing Research (2025), §"The Geopolitics of Research Infrastructure."
The "free connectivity" offer is the part that most deserves scrutiny
The narrative of generous tech companies bringing "free" connectivity in the name of progress and "connecting people" functions as a modern-day civilizing mission — a framing that masks the underlying purpose: building infrastructures with monopolistic tendencies that create deep societal dependency on a handful of platforms.4 The generosity is real at the surface and structural at the root: each "free" service deepens a lock-in that makes local, sovereign alternatives progressively harder to start and sustain.
The choice is not a foolish one, though, which is exactly what makes it hard to undo. Developing nations opt for foreign providers for rational reasons — the imperative of rapid development, lower prices, and technical assistance that local alternatives cannot yet match. An appeal to security or sovereignty alone cannot defeat a powerful economic logic. The honest path is not to shame the choice but to build alternatives credible enough to make a different choice possible.
⁴ Decolonizing Research (2025), §"The Geopolitics of Research Infrastructure"; CSIS, "An Overview of Global Cloud Competition," 2025.
Three routes to the same destination
Three distinct, real models of building sovereign research infrastructure show how autonomy is actually achieved — and together they make a single counterintuitive point: sovereignty does not require self-sufficiency.
The first is regional collaboration. The BELLA Programme secured an Indefeasible Right of Use over 40 optical channels on a direct submarine cable between Latin America and Europe, giving RedCLARA a dedicated research-and-education link that no longer routes through the United States — built, notably, through partnership with the EU rather than in isolation from it.5
The second is state-led public infrastructure. India's "India Stack" built a foundational layer of open-source digital public goods — identity, payments, data exchange — that lowered the barrier to entry for domestic innovators and reduced market concentration without the state centralising control of the services themselves.6
The third is navigating constraint. African NRENs, often unable to fund their own infrastructure and treated as competitors by commercial ISPs, have found inventive routes anyway: Zambia's ZAMREN partnered with power and water utilities to access their existing fibre backbones.7
What unites the three is the finding beneath them: the most successful models work through strategic interdependence and a phased approach, not economic self-sufficiency. Sovereignty here is not isolation — it is building the infrastructure and governance capacity to engage with the world on equitable terms. Autonomy is available even without abundance, provided the strategy is deliberate about ownership, governance, and whom it partners with.
⁵ BELLA / RedCLARA, European Commission International Partnerships. ⁶ India Stack, indiastack.org; ORF, "Digital Public Infrastructure as a Catalyst for Private Sector Innovation." ⁷ UbuntuNet, "ZAMREN demonstrates NREN synergies," 2024.
Sovereignty does not mean self-sufficiency, and that distinction is liberating
Digital sovereignty does not require economic self-sufficiency.8 The distinction matters because the self-sufficiency framing is paralysing: a research system that believes sovereignty means building everything itself, in head-to-head competition with the US and China, will conclude correctly that it cannot afford sovereignty and give up. For some countries, resources will simply never be enough to compete head-on with established players.
The reframing dissolves the trap. If sovereignty is the capacity to govern one's digital destiny and engage the world on equitable terms — rather than the capacity to produce everything domestically — then it becomes achievable through strategy rather than scale. Each of the three case models secured a sovereign asset through partnership, public-goods design, or creative use of existing infrastructure, none of them through self-sufficiency. The question for an institution or a nation is therefore not "can we build our own?" but "can we structure our engagement so that we govern what matters?"
⁸ Decolonizing Research (2025), §"Pathways to Sovereignty."
The technology to decentralise already exists; the binding constraint is governance
There is a genuine technological pathway away from the centralised model. Decentralised data architectures distribute storage and processing across nodes, letting each domain manage its own data while keeping it accessible — federated learning keeps sensitive data local and shares only model updates, which is decisive for biomedical and genomic research; edge computing processes data near its source, cutting the latency and cost of shipping it to distant clouds.9 Next-generation connectivity makes these viable at scale: LEO satellites reach deserts, oceans, and disaster zones terrestrial networks cannot, and 6G is being designed around distributed data management.
Adopting the technology, though, is not enough. Decentralisation without a matching governance model produces fragmentation and integration chaos; what makes it work is a federated governance model — shared, clearly defined data policies and standards applied across domains while preserving each domain's autonomy.10 The technological shift requires a simultaneous, deliberate build-up of political and institutional capacity. The hardware pathway is real, but it delivers sovereignty only when paired with the governance capacity to direct it — which makes this a leadership problem, not a procurement one.
⁹ Decolonizing Research (2025), §"The Technological Shift"; Milvus, "Federated vs centralized learning." ¹⁰ Decolonizing Research (2025), §"Emerging Data Architectures."
The sovereign infrastructure exists, but it has barely begun
The working models can give a misleading impression that the problem is being solved. Of the roughly 12,000 institutions that RedCLARA's sovereign infrastructure could ultimately serve, only about 3,000 are currently connected — a quarter of the potential, with three-quarters still unconnected.11 The sovereign alternative is not hypothetical, but neither is it mature; it has started, and it has a very long way to go.
The existence of working models — BELLA, India Stack, the African data-centre boom, open-source platforms like MOSIP and OpenStack — proves the path is real, but the gap between what the infrastructure could serve and what it currently reaches is the actual scale of the work ahead. For an institution weighing whether to invest in or connect to sovereign infrastructure, the number cuts both ways: the foundation is there to build on, and the early-mover advantage is still available because the build-out has barely started.
¹¹ Decolonizing Research (2025), §"The Sovereignty Gap"; GÉANT, "World Regions — Latin America."
Decolonising research is a long-term act of will, not a technical upgrade
There is no quick fix. Decolonising the digital research ecosystem is a long-term undertaking requiring sustained political will and collaboration, built on four pillars working together: proactive policy and governance (data residency, open standards, interoperability to prevent lock-in); diversified and sustainable funding (public investment, regional development banks, NRENs acting as demand aggregators); human and social capital (local expertise, and ethical frameworks built on reciprocity and self-determination rather than thin notions of individual consent); and South-South collaboration.12
No single pillar suffices — technology without governance fragments, governance without funding stalls, funding without local capacity reproduces dependency in a new form. The sharpest ethical demand sits inside the human-capital pillar: building this capacity in ways that respect and empower marginalised and Indigenous communities, prioritising reciprocity and mutual benefit, so that the construction of sovereign infrastructure does not quietly repeat the extractive patterns it was meant to end. Decolonising research is, in the end, a question of sustained intention as much as of technology or money.
¹² Decolonizing Research (2025), §"A Strategic Framework"; ARDC, "Indigenous Data"; AfricaConnect3.
Decolonizing Research is a critical analysis of data colonialism in global research infrastructure and a strategic framework for building sovereign, equitable digital ecosystems across the Global South. These seven insights are its entry points.