What If 120 Nations Built Their Own AI… and Didn't Need Silicon Valley or Beijing?

Prologue: A Room in Bandung

In April 1955, twenty-nine nations — most of them newly independent, many of them scarred by colonialism — gathered in a modest conference hall in Bandung, Indonesia. They had no nuclear weapons. They controlled no global financial institutions. They held no permanent seats on the UN Security Council. And yet, over six extraordinary days, they rewrote the rules of geopolitics.

The Bandung Conference gave birth to the Non-Aligned Movement — a collective refusal, by more than half the world's population, to be absorbed into the binary logic of the Cold War. The message was deceptively simple: We will not choose your side. We will build our own.

Seventy-one years later, a strikingly similar dynamic is unfolding — not over nuclear warheads and proxy wars, but over something arguably more consequential: artificial intelligence. And once again, the nations being told to pick a side are beginning to ask whether the question itself is the problem.

I. The Duopoly Nobody Voted For

To understand what is emerging, you first need to understand what it is emerging against.

The global AI landscape in 2026 is dominated by three paradigms, though only two of them wield genuine infrastructural power. The United States operates a market-driven model, where the largest AI systems are built by private corporations — OpenAI, Google DeepMind, Anthropic, Meta — funded by venture capital and governed primarily by shareholder value. China operates a state-coordinated model, where AI development is embedded within a national strategy for economic modernisation and social governance, with companies like Baidu, Alibaba, and ByteDance operating under close party oversight.

The European Union, often cited as a third pole, has carved out a regulatory identity through the AI Act — but it remains largely a consumer and regulator of AI, not a foundational builder. Brussels writes the rules. It does not train the models.

All three paradigms share an unspoken assumption: the rest of the world — the 85% of humanity living outside the US, China, and the EU — will be a recipient, a customer, a data source. Not a creator.

This assumption is so deeply embedded in Western strategic thinking that it rarely gets examined. When policymakers in Washington discuss "AI competition," they mean the US versus China. When European regulators draft AI governance frameworks, they model them on European values and European markets. When Silicon Valley executives talk about "democratising AI," they mean making their products cheaper and more accessible — not surrendering architectural control.

But something is shifting. And it is shifting fast.

II. The Signals Are Already Here

In February 2026, the Carnegie Endowment for International Peace published a landmark analysis on "South-South AI Collaboration," documenting how governments across the Global South are forging direct partnerships that bypass the traditional US and China technology pipelines entirely. This was not a theoretical paper. It was a map of what is already happening.

India has committed $1.2 billion to develop and democratise AI through an open-source-first strategy. The Indian government's vision is not to replicate Silicon Valley but to build AI infrastructure tailored to Indian languages, Indian agricultural cycles, Indian healthcare challenges, and Indian governance structures. Red Hat's Asia-Pacific CTO described it as "a new paradigm for AI sovereignty" — a model where openness and national interest are not in tension but in alignment.

Brazil has taken a different but complementary path. As the driving force behind UNESCO's AI Ethics Guidelines, Brasília has positioned itself as a global advocate for data sovereignty and equitable AI access. The Brazilian approach is less about building foundational models and more about ensuring that wherever AI is deployed, the communities it affects retain meaningful control over their data and their futures.

The African Union has developed a continental AI strategy that begins not with the technology but with the problem — asking what African societies actually need AI to do, rather than adapting Western solutions to African contexts. ASEAN is doing the same. So are coalitions across Latin America and the Middle East.

These are not isolated initiatives. They are the early tremors of a tectonic shift.

III. The Fourth Path

Dr Cornelia Walther, a scholar at the Wharton School, has given this movement a name: the "Fourth Path."

Where the American model optimises for profit, the Chinese model for control, and the European model for regulation, the Fourth Path optimises for collective flourishing. It is built on four principles, each beginning with "T" — an acronym that feels almost too neat, but captures something essential:

Tailored to local contexts — not imported wholesale from Menlo Park or Shenzhen, but designed from the ground up for the communities it will serve.
Trained on representative data — meaning the datasets reflect the linguistic, cultural, and demographic realities of the populations using the technology, not the internet browsing habits of English-speaking Westerners.
Tested for equity impacts — with rigorous assessment of who benefits, who is harmed, and who is excluded before deployment, not after.
Targeted toward collective benefit — where the measure of success is not quarterly earnings or social stability scores, but tangible improvements in health, education, agriculture, and governance for ordinary people.

This is not anti-technology. It is not anti-Western. It is, at its core, a demand for relevance — for AI that speaks Yoruba, understands monsoon patterns, diagnoses tropical diseases, and respects indigenous knowledge systems.

And it is a demand that is becoming impossible to ignore.

IV. The Uncomfortable Question

Here is where the article asks you to sit with something uncomfortable.

What if this movement succeeds — and the West doesn't like the answer?

Consider a concrete scenario. Nigeria develops an AI diagnostic system optimised for malaria — a disease that kills over 600,000 people annually, almost exclusively in sub-Saharan Africa and South Asia. The training data comes from Nigerian hospitals, annotated by Nigerian clinicians, and validated against Nigerian patient outcomes. The system outperforms every Western-built alternative because it was designed for this context from the start.

Nigeria then shares its methodology — openly, without licensing fees — with Indonesia, where malaria is also endemic. Indonesia adapts the model, improves it, and shares the improved version with Bangladesh. Within three years, a network of Global South nations has built a malaria diagnostic capability that no single Western company could match, because no single Western company ever had the incentive to try.

Now scale that logic across agriculture, education, climate adaptation, and governance. The compounding effect is extraordinary.

The Chatham House assessment of this dynamic is characteristically blunt: the US-China AI race is forcing countries worldwide to fundamentally reconsider who owns their digital infrastructure. Not who hosts it. Not who licenses it. Who owns it? The distinction matters enormously because ownership determines who sets the terms, who captures the value, and who gets left behind when the next upgrade ships.

V. The Vacuum

If the Fourth Path is a pull factor — an aspirational vision of AI sovereignty — then the behaviour of existing powers is the push factor.

The United States, under the current administration, has deprioritised multilateral AI diplomacy. It abstained from the Paris AI Summit statement and halted USAID funding that had supported technology capacity-building in the Global South. The message, intended or not, was clear: America's AI strategy is about American competitiveness, full stop.

China, meanwhile, has been far more engaged — but the engagement comes with strings. The Digital Silk Road initiative has embedded Chinese AI infrastructure into governance and commerce across Africa, Southeast Asia, and Central Asia. The short-term benefits are real: affordable surveillance systems, smart city platforms, and digital payment networks. But so are the long-term dependencies. When your national AI infrastructure runs on Huawei hardware, processes data through Alibaba Cloud, and operates within a Chinese-designed governance framework, sovereignty becomes a polite fiction.

The EU, for all its regulatory ambition, has struggled to export its model. The AI Act is impressive on paper, but it was designed for European market conditions, European institutional capacity, and European legal traditions. Asking Kenya or Cambodia to adopt Brussels' regulatory framework is a bit like asking someone to wear a bespoke suit tailored for a different body.

Into this vacuum — where America is disengaged, China is entangling, and Europe is prescribing — the Fourth Path is advancing with remarkable speed.

VI. Digital Colonialism and the Case for Refusal

There is a harder edge to this story, and it deserves to be named plainly.

Academic researchers have begun using the term "digital colonialism" to describe what happens when AI systems built in the Global North are deployed in the Global South without meaningful local input. The term is deliberately provocative, and it is deliberately accurate.

When a facial recognition system trained predominantly on Caucasian faces is deployed in sub-Saharan Africa, it fails disproportionately — and the consequences fall on the people least able to challenge them. When a language model trained on English-language internet text is used to generate policy recommendations for a Hindi-speaking population, the cultural assumptions embedded in the training data are invisible but consequential. When a Western edtech company deploys an AI tutor in Southeast Asia that cannot distinguish between Bahasa Indonesia and Bahasa Malay, the system is not neutral. It is negligent.

The Fourth Path is, among other things, a refusal to accept this negligence as an acceptable cost of progress. And that refusal is not coming from activists or academics alone. It is coming from finance ministers, technology secretaries, and heads of state.

VII. The Scenario — What Happens Next

If you've read this far, you've earned the speculative part.

Late 2027: A coalition led by India, Brazil, Indonesia, and Nigeria announces the "Bandung Protocol for Artificial Intelligence" — a framework for shared, decentralised compute infrastructure funded by pooled sovereign wealth. The initial investment is modest — $8 billion, a fraction of what any single US tech company spends annually on AI. But the structure is revolutionary: no single nation controls the network, contributions are proportional to GDP, and all foundational models trained on the shared infrastructure are open-source by default.

2028: The coalition's combined addressable population — 3.5 billion people — becomes the world's largest AI user base overnight. Open-source models trained on Hindi, Portuguese, Bahasa, Hausa, and Swahili begin outperforming proprietary Western models in healthcare diagnostics, crop yield prediction, and primary education across partner nations. Silicon Valley companies, accustomed to setting global standards, are suddenly scrambling to comply with Bandung Protocol data standards or lose access to the fastest-growing digital markets on Earth.

Early 2029: The first "AI Non-Aligned Summit" convenes in Jakarta — deliberately echoing the 1955 conference. Sixty-seven nations send delegates. The invitation reads: "We are not anti-West or anti-China. We are pro-humanity." The summit produces the Jakarta Declaration on Digital Self-Determination, which establishes three foundational principles: data generated by a nation's citizens remains sovereign; AI systems deployed within a nation must be auditable by local institutions; and no nation shall be compelled to adopt another's AI governance framework as a condition of trade.

Late 2029: The World Trade Organisation begins formal consultations on whether the Jakarta Declaration's data sovereignty provisions are compatible with existing trade agreements. The answer is far from clear — and the ambiguity itself is the point. The Global South has created a new negotiating position, not by asking permission, but by building something the world cannot afford to ignore.

VIII. The Real Question

The scenario above is speculative. The trajectory beneath it is not.

Every data point cited in this article is real. The $1.2 billion Indian commitment is real. The Carnegie research is real. The Chatham House warnings are real. The World Economic Forum's 2026 report on sovereign AI pathways is real. The academic literature on digital colonialism is real and growing.

The question is not whether the Global South will assert itself in AI. It already is. The question is whether the Global North will interpret that assertion as a threat to be countered — or an invitation to be accepted.

History suggests the answer depends on who is paying attention.

In 1955, the Western press largely dismissed Bandung as a sideshow — a gathering of irrelevant nations with grandiose ambitions. Within fifteen years, the Non-Aligned Movement had fundamentally reshaped the United Nations, decolonised a continent, and created the geopolitical architecture that still frames North-South relations today.

The delegates at Bandung understood something that strategists in Washington and Moscow did not: when 85% of humanity decides to move in the same direction, the remaining 15% do not get to call it a fringe movement.

The Fourth Path is forming. The protocols are being drafted. The compute is being pooled. The models are being trained.

The only question left is the one this publication was built to ask:

What if they succeed?