Three very different stories from three very different parts of the world kept landing on the same idea this week. Africa’s four largest tech economies have formally admitted, in their own national AI strategies, that they depend on Google, Microsoft, NVIDIA, and Meta for the infrastructure beneath their digital economies. Nebius just bypassed the entire US grid interconnection queue by deploying 328 MW of behind-the-meter fuel cells from Bloom Energy, taking physical control of its own power supply rather than waiting for a system that was never designed for what it now needs to do. And researchers in Australia laid out the evidence that premature, politically driven infrastructure announcements cost taxpayers an average 35% more than projects that go through rigorous independent review first. The pattern is the same in all three stories: when you don’t control the foundational layer, someone else controls your future.
Africa Is Writing AI Strategies That Admit the AI Doesn’t Belong to Them
A landmark Rest of World investigation published this week documented what African policymakers have been saying quietly in conference rooms for years: Nigeria, Kenya, Egypt, and South Africa have each drafted national AI strategies that explicitly identify dependence on US technology companies as a threat to sovereignty and security. Africa holds less than 1% of global data centre capacity, according to the World Economic Forum. The top five African tech markets combined have less data centre capacity than France had in 2024, McKinsey found. Foreign servers, mostly in Europe, handle an estimated 80% of the continent’s internet traffic. Cassava Technologies, founded by Zimbabwean entrepreneur Strive Masiyiwa, is building AI infrastructure with Western technology behind it. A proposed $60 billion continental AI fund and a new pan-African AI council are meant to change the equation — but they are working against the grain of a system in which every layer of infrastructure, from the chips to the cloud to the cables, was built by someone else for someone else’s purposes.
The deeper problem is structural and it goes beyond technology. As TechCabal’s analysis put it plainly this week, Africa’s AI sovereignty challenge is “an energy problem wearing a technological disguise.” The Africa Data Centres Association estimates the continent needs at least 1,000 MW of new capacity across 700 new facilities just to meet current demand — and most of the countries with national AI strategies do not have the reliable grid infrastructure to power the data centres they would need to host their own compute. South Africa, Kenya, and Nigeria together account for 41% of the continent’s 223 data centres. The remaining 35 countries share fewer than 130 facilities. 71% of African executives, according to BCG’s AI Radar 2026, believe their job stability depends on executing an AI strategy this year — but the infrastructure to do that at scale, on African terms, does not yet exist. Rachel Adams of the Global Center on AI Governance offered the most honest framing available: Africa’s push for digital sovereignty cannot mean total independence from global AI supply chains, but it can mean stronger control over sensitive data, better procurement rules, investment in local infrastructure and skills, and clearer accountability for foreign providers. The continent that supplies the minerals powering the global AI economy cannot build its own AI future on data centres it doesn’t own, powered by grids it can barely keep running.
Read the Rest of World investigation into Africa’s hard road to AI sovereignty.
1. Nebius Just Proved You Don’t Have to Wait for the Grid
On 20 May 2026, Nebius Group and Bloom Energy announced a partnership to deploy 328 MW of Bloom’s solid oxide fuel cell technology as behind-the-meter power for Nebius’s AI infrastructure buildout in the United States, replacing previously planned combustion-based generation with clean fuel cells that can be sited and commissioned on accelerated timelines, with no new transmission build required and a deal estimated at approximately $2.6 billion by secondary reporting. The significance is not just environmental — it is strategic: every AI infrastructure operator facing a multi-year wait for grid interconnection is watching this deal to see whether behind-the-meter fuel cells offer a credible path to faster time-to-power at the scale that agentic AI workloads demand.
If the Nebius-Bloom model works at 328 MW, it becomes a template for every AI cloud operator, sovereign compute programme, and enterprise data centre operator that has a site and the capital but cannot get a grid connection in time to meet demand. The physical constraint that Goldman Sachs identified last week as the primary bottleneck of the agentic AI era — not chips, but power — is now being solved not by waiting for utility-scale transmission but by going around the grid entirely with modular, on-site generation. The company that cracks behind-the-meter power at AI factory scale is not just solving an energy problem — it is dissolving the queue that currently controls who gets to build AI infrastructure and when.
2. Australia’s Infrastructure Blowout Problem Has a Fix — and No One Has Applied It Yet
A major analysis published in The Conversation this week documented the systematic failure of Australian infrastructure project governance: premature political announcements made before independent business cases are complete cause an average 35% cost overrun, according to a 2020 Grattan Institute report, with Melbourne’s Suburban Rail Loop announced at $50 billion in 2018 with no business case and now estimated at $96 billion as the most visible example. The researchers propose three reforms modelled on Norway’s approach, which reduced the proportion of large road projects experiencing cost overruns from 72% before mandatory independent review to 27% afterwards: mandatory independent assurance before any announcement, an independent project authority with power to pause projects, and institutional separation of infrastructure planning from political timelines.
The relevance of Australia’s blowout problem extends well beyond a single country’s transport projects. The same dynamic — political announcements driving premature commitment before rigorous appraisal — is visible in infrastructure programmes across the UK, the United States, and dozens of emerging markets simultaneously deploying large capital programmes under political pressure to show results. Norway’s quality assurance reform is one of the most evidence-based infrastructure governance improvements ever implemented, yet only a handful of countries have adopted it. If Australia’s researchers can persuade their government to act, the case study will matter globally. The world does not have a shortage of infrastructure ambition or capital — it has a shortage of the institutional discipline to spend that capital on projects that actually work.
3. Africa’s Private Capital Gap Is the Real Barrier to Energy Infrastructure Delivery
ESI-Africa published a major analysis this week on the private capital mobilisation challenge facing Africa’s energy infrastructure, grounding the argument in a straightforward problem: African Review’s Aaron Chehab, head of strategy at Arabian Construction Company, articulated the “bankability” challenge directly — before a single dollar of private capital can be committed, projects must demonstrate the contractual, regulatory, and revenue certainty that turns an investment thesis into a financeable asset, and most African energy projects currently fail that test before investors even reach due diligence. Unlocking grid capacity, accelerating infrastructure financing, and crowding in private capital are the three levers African governments must pull simultaneously, but they are pulling them sequentially at best and selectively at worst.
The practical implication is that the energy infrastructure gap constraining Africa’s AI ambitions is not primarily a problem of investor appetite — BlackRock, KKR, Brookfield, and African Development Bank affiliates have all signalled interest in African energy infrastructure. It is a problem of project preparation, regulatory predictability, and the institutional capacity to structure deals that survive the scrutiny of international capital. South Africa’s Ifisa, Tanzania’s infrastructure bonds, and Kenya’s geothermal-anchored data centre pipeline are all early examples of what bankability looks like when it is achieved. The countries that systemise that model will attract private capital at scale. Every African energy project that fails to reach financial close is not a financing problem — it is a governance problem dressed up as a financing problem.
1. SpaceX’s IPO Has a Wild Infrastructure Idea Buried Inside It
Business Insider this week flagged the most underreported element of the SpaceX IPO preparation: a paper cited in the prospectus discussion argued that AI’s future energy demands may require moving compute infrastructure into orbit, citing Google’s radiation-hardening tests for space-based computing as early validation. SpaceX’s Starlink constellation and its planned Starshield government satellite network position it as the only organisation currently capable of operating at the intersection of orbital infrastructure and AI compute at scale. This is speculative infrastructure planning at its most extreme, but the people who dismissed offshore oil platforms, underground data centres, and nuclear-powered AI factories as impossible six years ago are less confident now.
Read the Business Insider analysis on the infrastructure implications buried inside SpaceX’s IPO.
2. Vancouver Is Asking Residents to Shape Its Infrastructure Investment Priorities
The City of Vancouver has opened a public consultation process asking residents and businesses to share their priorities for infrastructure and amenity investment through the 2027 to 2030 capital planning cycle. The process reflects a growing trend of cities using participatory planning to legitimise infrastructure prioritisation decisions that will inevitably involve trade-offs between transport, housing, digital connectivity, energy, and climate resilience investments in constrained fiscal environments. For urban infrastructure planners, Vancouver’s process is worth studying as a model for how democratic legitimacy and technical rigour can be combined at the city scale — particularly as AI-driven demand signals begin to compete with community priorities for the same grid capacity and development sites.
Read about Vancouver’s public consultation on priorities for future infrastructure investment.
3. Capacity Global’s Digital Infrastructure Platform Is Connecting the Next Billion
Capacity Global published its May 2026 analysis on the digital infrastructure platform powering regional connectivity across underserved markets, framing the commercial case for wholesale connectivity infrastructure as the foundational layer beneath every national digital economy ambition. The piece argues that digital infrastructure investment in emerging markets is most capital-efficient when it operates as a shared platform connecting multiple operators and services, rather than as a series of competing national networks with duplicated physical assets and incompatible standards. For investors watching the regional connectivity space, shared platform models in markets like Southeast Asia, West Africa, and the Middle East are producing returns that challenge the assumption that emerging market digital infrastructure is inherently higher risk than developed market equivalents.
35%
That is the average cost overrun suffered by infrastructure projects that were announced politically before an independent business case was completed, according to the Grattan Institute, a number that explains why the world’s largest infrastructure investment programmes consistently deliver less than they promise and cost more than they should, and why governance reform is at least as important as capital mobilisation in closing the global infrastructure gap.
Watch for the first operational results from Nebius’s 328 MW fuel cell deployment with Bloom Energy, expected later in 2026, because if Bloom’s solid oxide technology performs as contracted at AI factory density — delivering reliable, clean, behind-the-meter power without transmission build — it will trigger a wave of similar partnerships across the AI infrastructure sector and permanently change the site selection calculus for every data centre developer currently stuck in a grid interconnection queue.
Publisher, Benjamin Yaw Manu, Founder, nimdier.com | Author of Thriving in Uncertainty — get your copy on AmazonEmail: hello@nimdier.com
Have feedback or a topic you’d like to see covered? Reply to this message and let us know.
Warm regards
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.