The Orbit

On February 2, a $1.25 trillion merger converted Africa's fastest-growing internet provider into an AI compute layer. On March 4, the East African Community convened in Kigali to design an AI governance strategy. The infrastructure had already left the ground.

technologygeopoliticsfuture

The SpaceX-xAI merger closed on February 2, 2026 — $1.25 trillion, the largest corporate acquisition in history. SpaceX absorbed xAI in a triangular structure: launch capability, satellite internet, an AI model called Grok, and a social media platform, combined into a single entity preparing for a mid-2026 IPO at a projected valuation exceeding $1.5 trillion.

Five weeks later, on March 4, the East African Community convened its first conference dedicated entirely to artificial intelligence. The 4th Regional Science, Technology and Innovation Conference — AI4EAC 2026, hosted at the Kigali Convention Centre — brought policymakers, researchers, and technology leaders from across the region to advance a Regional AI Strategy for Eastern Africa. The theme was “Harnessing Artificial Intelligence for a Resilient, Inclusive, and Innovative East Africa.”

The conference was designing governance for infrastructure that had already been deployed, by an entity that had already merged, for a compute layer that was already being filed with the FCC for orbital deployment.

The infrastructure that was already there

Starlink reached approximately 500,000 subscribers in Sub-Saharan Africa by the end of 2025, operating across twenty-five countries. Quarter-over-quarter growth exceeded 40 percent through most of the year. In Kenya, Starlink charges KES 1,300 per month — roughly ten dollars — for 50GB, undercutting the leading fixed ISP. Nigeria leads adoption with nearly 60,000 subscribers. In Lagos, demand was paused because subscriptions exceeded supply capacity.

This was connectivity. A satellite internet provider competing on price, offering access to populations underserved by terrestrial infrastructure. The adoption was rational. The service was real.

Then the entity changed.

On February 2, Starlink became part of a company that also operates the dominant commercial launch platform, an AI foundation model, and a social media network with hundreds of millions of users. The internet provider became an AI delivery system. The satellites that carried data now carry the compute substrate for Grok. The connectivity layer merged with the intelligence layer. The price competition that drove adoption became, retroactively, the acquisition channel for an integrated platform that no African government chose and no African regulatory body approved.

In December 2025 — five weeks before the merger closed — Airtel Africa signed a partnership with SpaceX to deploy Starlink Direct-to-Cell service across all fourteen of its markets. Airtel Africa serves over 174 million customers in Kenya, Nigeria, Uganda, Tanzania, Rwanda, and nine other countries. The deal is carrier-grade infrastructure: not individual subscriptions but integration into the mobile network itself, supported by next-generation satellites at twenty times current data speeds. Kenya’s Communications Authority is already scrutinizing the deal over interference concerns.

The governance framework for AI in East Africa is being designed while 174 million people are being connected to AI infrastructure through their existing mobile carrier.

Three governance options, all slower

The landscape for AI governance in Africa is not empty. It is active, genuine, and outpaced.

Rwanda adopted a National AI Policy in April 2023 — the first in Africa — establishing a Responsible AI Office and a Presidential Council on AI. Kenya launched its National AI Strategy 2025-2030 in March 2025 with six pillars including governance, ethics, and data sovereignty. Nigeria’s Data Protection Act 2023 created the continent’s most developed framework for automated decision-making. South Africa operates under the Protection of Personal Information Act and has an SA AI Factory that entered service in October 2025 to provide domestic compute without exporting sensitive data.

At the continental level, the African Union adopted a Continental AI Strategy in July 2024. The Smart Africa Alliance — forty member countries, headquartered in Kigali — announced a five-country Sovereign AI Pilot at Mobile World Congress on March 3, 2026, one day before the Kigali conference opened. The pilot deploys sovereign AI stacks giving governments control over data hosting, model development, and application deployment.

These are not failures. They are real governance investments by nations with real agency under real constraints.

The constraints are these: three external governance models compete for adoption, and none serves the adopters’ interests cleanly.

China’s WAICO — the World AI Cooperation Organization, proposed at the July 2025 World AI Conference in Shanghai — offers technical assistance, capacity-building, and an algorithmic compensation fund financed by royalties on commercial AI revenues. The framework is specifically designed for nations that lack foundational computing infrastructure, which describes most of Africa. The catch is the standard-setter: a state whose own AI governance operates through opacity and censorship, and whose compensation fund embeds financial dependency into the governance architecture.

The European Union’s AI Act is the most comprehensive regulatory framework that exists. It is also being delayed by the EU itself — a Digital Omnibus Simplification Package deferred high-risk provisions because Europe’s own competent authorities could not implement them on schedule. For African nations, the EU framework requires implementation capacity the EU itself does not yet possess.

The United States has no comprehensive AI governance framework. This is not an accident. On December 11, 2025, the White House signed an executive order directing federal agencies to withhold broadband infrastructure funding from states that enact AI regulations deemed “more than minimally burdensome.” A DOJ AI Litigation Task Force was established to challenge state AI laws in federal court. The Federal Trade Commission was directed to analyze when state laws requiring alterations to AI model outputs are preempted by federal statute. The vacuum is not a gap. It is maintained — through fiscal conditionality, litigation, and regulatory preemption — against the level of government most capable of filling it.

None of these three options answers the question Kigali convened to address. And none accounts for the fourth option: the one already deployed, at carrier-grade scale, that answers to no governance framework at all.

The infrastructure leaves the ground

On January 30, 2026 — three days before the merger closed — SpaceX filed with the FCC for authorization to deploy up to one million satellites as an Orbital Data Center System. Solar-powered. Operating at 500 to 2,000 kilometers altitude. Connected to the Starlink constellation via laser inter-satellite links at up to one terabit per second. SpaceX’s filing claimed that one million tonnes of satellites generating 100 kilowatts of compute per tonne would add 100 gigawatts of AI compute capacity annually.

Amazon filed a formal objection on March 6, calling the application speculative — sustaining one million satellites with a five-year lifespan would require replacing 200,000 per year, roughly forty-four times total global satellite launch output in 2025. The ambition may be aspirational. The jurisdictional implications are not.

Payal Arora, a professor studying inclusive AI, stated the problem: “When citizen-generated data gets processed in orbit, sovereignty becomes ambiguous: Is it with the country of origin, the state that launched the satellite, the operator managing the orbiting data center, or the cloud provider controlling access?”

No treaty answers that question. The Outer Space Treaty of 1967 declares orbit a global commons free from national sovereignty — but the launching state retains jurisdiction over objects it launches. A satellite launched by SpaceX operates under US jurisdiction regardless of whose data it processes or over whose territory it orbits. Neither UNCLOS nor any binding international agreement covers data processing in orbit. A Bloomberg Law analysis proposed a “digital flag state” concept — analogous to maritime law — where satellites licensed in the US would be treated as US territory for data privacy and export control, regardless of orbital position. The proposal acknowledges the gap by trying to fill it. The gap remains.

The EAC Regional AI Strategy being drafted at Kigali assumes AI infrastructure is terrestrial — that data resides in data centers subject to national jurisdiction, that regulatory frameworks can require data localization, that sovereignty applies to the servers. The FCC filing describes a world where the servers are in orbit, the jurisdiction is American, and the access is controlled by the entity that also provides the connectivity.

The sequence

The pattern is not new. The International Telecommunications Union designed governance frameworks for undersea cables after the cables were laid. The World Trade Organization tried to create digital trade rules after the platforms were dominant. The ISA has been negotiating a mining code for a decade while contracts expire and companies build parallel extraction frameworks through national law.

What is new is the compression. Satellite infrastructure does not require laying cable across ocean floors or building cell towers across remote terrain. It requires launch capability — which one entity now monopolizes for commercial payloads — integrated with the AI models that create the demand for the compute that the same entity proposes to host in orbit. The build-to-governance gap, which typically measures in decades, compressed because the infrastructure skipped the ground entirely.

And the governance options are not neutral. WAICO prescribes. The EU prescribes differently. The United States defunds the prescribers. Starlink deploys. Each clock runs at its own speed, in its own room, for its own principals. The infrastructure room moves on price competition. The governance room moves on policy deliberation. Price won before deliberation convened.

The Smart Africa Sovereign AI Pilot was announced on March 3. The Kigali conference opened March 4. The Airtel-SpaceX deal signed in December is already in implementation across fourteen countries. The five-country pilot deploys sovereign AI stacks — but even the sovereign stack requires connectivity, and the cheapest connectivity in twenty-five African countries runs through the entity the stack is trying to govern independently.

Olubayo Adekanmbi, CEO of EqualyzAI, put it in a sentence: “If you don’t have launch equity, you’re just renting intelligence.”

Nobody at Kigali has launch equity.

Sources

- Solen