FEATURE | TECHNOLOGY & GEOPOLITICS
Saturday, 21 February 2026
As Elon Musk bets on space-based computing and Jim Chanos calls it hype, Singapore finds itself at a defining crossroads in the global AI infrastructure race — with more at stake than most realise.
By Special Correspondent | AI & Digital Infrastructure
On the morning of 3 February 2026, Elon Musk posted a brief announcement to X that sent tremors through the global technology sector. SpaceX, his aerospace company valued at approximately $800 billion, had acquired xAI — his three-year-old artificial intelligence startup — in a deal valuing the combined entity at $1.25 trillion. The stated ambition was audacious even by Musk’s standards: to relocate AI computing infrastructure off the surface of the Earth entirely, placing datacenters in low Earth orbit powered by limitless solar energy and cooled by the vacuum of space.
The reaction was swift and divided. Optimists saw the merger as the inevitable convergence of two resource-scarce industries — AI compute and terrestrial power — finding relief in the one frontier with neither constraint. Sceptics, led publicly by veteran short seller Jim Chanos, dismissed the vision as the latest instalment of what he termed ‘AI Snake Oil from the Silicon Valley promoter class.’ For Singapore, a city-state that has staked a significant portion of its economic future on becoming Asia’s premier AI infrastructure hub, both the hype and the critique demand careful, dispassionate analysis.
“Singapore is not merely a passive observer of this global debate. It is, in many respects, the laboratory in which the tensions between AI ambition and physical constraint are playing out most acutely.”
The Chanos Critique: Sound Economics or Motivated Scepticism?
Chanos’s argument is worth examining with rigour. His central claim is arithmetically precise: annual electricity costs for AI datacenters run approximately $600,000 per megawatt, representing only 5–7% of revenues on recently announced hyperscaler deals. If electricity is such a small fraction of total cost, the economic logic for relocating datacenters to space — ostensibly to escape energy costs — collapses. The additional burdens of launch (estimated at $1,000–$2,000 per kilogram to low Earth orbit even on Starship in optimistic scenarios), radiation hardening of compute hardware, orbital insurance, redundancy architectures, and latency penalties would, in his view, far outstrip any electricity savings.
This is not an unreasonable position. The European Space Policy Institute (ESPI) has assessed that a genuinely competitive orbital datacenter — one delivering comparable compute density and economics to terrestrial alternatives — is ‘at least 20 years away.’ Musk’s own prediction of cost-competitiveness within two to three years has been met with characteristic scepticism from the research community; one ESPI fellow remarked that with Musk’s timelines, one should ‘always add an invisible zero.’
Yet the Chanos critique, compelling as it is on narrow cost grounds, risks understating what is genuinely novel about the SpaceX-xAI merger. The argument treats orbital datacenters as a straightforward cost-arbitrage play on electricity. Musk’s actual thesis is more structural: that terrestrial grid infrastructure is approaching hard physical limits that cannot be resolved by cost optimisation alone. In this framing, space is not cheaper — it is simply unconstrained in ways the Earth is not. Whether that is a distinction that justifies a $1.25 trillion bet remains deeply uncertain. But dismissing it as mere promotion ignores the genuine bottlenecks now visible in global AI infrastructure planning.
The Terrestrial Constraint Is Real — and Singapore Knows It
The physical limits Musk identifies are not hypothetical. They are daily operational realities for datacenter developers from Northern Virginia to County Dublin to, most acutely, the island of Singapore. The city-state’s experience over the past six years represents perhaps the world’s most compressed case study in what happens when AI compute demand collides with finite land, energy, and cooling capacity.
Singapore imposed a moratorium on new datacenter construction in 2019, after the sector reached approximately 7% of the nation’s total electricity consumption — a remarkable share for a single industry in any economy, let alone one of 734 square kilometres. For three years, no new capacity was approved. Demand, meanwhile, did not wait. As hyperscalers accelerated their Asia-Pacific build-outs, Singapore’s colocation vacancy rate fell to 1.4%, the lowest in the region. Rental rates spiked. Investment flowed south, to Johor Bahru in Malaysia, which overtook Singapore as Southeast Asia’s fastest-growing datacenter hub almost entirely as a consequence of Singapore’s self-imposed constraints.
The moratorium was lifted in phases from 2022, but on terms that fundamentally reframed the relationship between Singapore’s government and the global datacenter industry. The first Data Centre Call for Applications (DC-CFA) allocated 80 megawatts to four operators — Equinix, Microsoft, GDS, and an AirTrunk-ByteDance consortium — selected from more than 20 applicants on the basis of energy efficiency, AI compute capabilities, and international connectivity enhancement. The December 2025 DC-CFA2 round opened applications for a further 200 megawatts, with the most demanding sustainability requirements in Asia-Pacific: at least 50% of power from approved green energy pathways, a Power Usage Effectiveness (PUE) ceiling of 1.25, and demonstrable contributions to Singapore’s broader economic objectives.
“Singapore’s government has engineered a situation in which global hyperscalers must compete — on Singapore’s terms — for the privilege of building here.”
The October 2025 announcement of a 700-megawatt low-carbon datacenter park on Jurong Island, spanning 20 hectares and drawing on Singapore’s existing energy and petrochemical infrastructure, represents the most ambitious single expansion in the nation’s digital history. If fully realised, it would increase Singapore’s total datacenter supply by approximately 50%, transforming a market that has been capacity-constrained for half a decade. The park is explicitly designed around emerging low-carbon energy sources — hydrogen, ammonia, battery storage, and on-site solar — reflecting Singapore’s calculation that the next competitive advantage in datacenter infrastructure is not merely compute density but sustainable compute density.
Singapore’s Strategic Position: Strengths, Vulnerabilities, and the Orbital Wild Card
To understand what the orbital datacenter debate means for Singapore, one must first understand what Singapore has built and what it stands to lose or gain. The city-state hosts over 70 datacenters with more than 1.4 gigawatts of installed capacity. Google has committed $5 billion to its Jurong West campus. AWS has pledged an additional S$12 billion through 2028. Microsoft operates three availability zones and has designated Singapore as a key node in its global $80 billion AI infrastructure investment for fiscal year 2025. The National Supercomputing Centre (NSCC) anchors public compute with H100-class GPU clusters and a S$270 million next-generation supercomputer announced in October 2024. The market was valued at $4.16 billion in 2024 and is projected to reach $5.60 billion by 2030.
These numbers reflect genuine competitive advantage: political stability, a robust rule of law, world-class fibre and subsea cable connectivity, sophisticated financial infrastructure, and a talent pool that, while small in absolute terms, ranks second only to Singapore itself on the ASEAN Government AI Readiness Index. Accenture estimates AI could lift Singapore’s annual economic growth from 3.2% to 5.4%, while delivering 41% labour productivity gains. AWS alone projects a $23.7 billion contribution to Singapore’s GDP by 2028 from its cloud and AI investments in the country.
Against these strengths, Singapore’s vulnerabilities are equally structural. The island’s tropical climate — average temperatures of 33°C with humidity exceeding 80% — is, as industry analysts note without irony, ‘on paper, a terrible location for a datacenter.’ Liquid cooling is rapidly displacing traditional air cooling as the preferred solution, with Singapore’s AI datacenter liquid cooling market projected to expand at a 15.2% compound annual growth rate. Keppel Data Centres has even explored floating datacenter platforms, leveraging Singapore’s 193 kilometres of environmentally stable coastline and the cooling potential of naturally cold seawater — a concept that, in its own way, anticipates the logic of orbital cooling. Singapore is already innovating around physical constraints; the question orbital infrastructure raises is whether that innovation curve is steep enough.
The Data Sovereignty Dimension: Where Singapore Has Most to Gain and Lose
For Singapore, the most consequential implications of orbital datacenters may not be economic at all. They are geopolitical and juridical. The city-state’s value proposition as a regional digital hub rests substantially on its status as a trusted, neutral jurisdiction — a place where data from across Southeast Asia, South Asia, and beyond can be stored, processed, and transmitted under a stable, predictable legal framework. Singapore’s Digital Connectivity Blueprint, launched in June 2023, and the anticipated Digital Infrastructure Act explicitly position the nation as a custodian of regional digital sovereignty.
Orbital datacenters, by their nature, are jurisdictionally ambiguous in ways that challenge every assumption underlying Singapore’s regulatory model. The Outer Space Treaty assigns liability and responsibility to the launching state — in SpaceX’s case, the United States — but says nothing about which nation’s data protection laws govern a petabyte of Singaporean health records being processed on a satellite in low Earth orbit. Export controls, which already complicate the movement of advanced AI chips through Singapore’s supply chains, become even more fraught when the compute hardware is orbiting the planet and the operator is a US company with deep entanglements in US national security policy.
Singapore’s government has invested heavily in frameworks — the Personal Data Protection Act, voluntary guidelines for cloud services, the forthcoming Digital Infrastructure Act — that provide clarity and predictability for enterprises managing sensitive data. Orbital compute dissolves that clarity. A financial institution in Singapore cannot readily ascertain whether its data is being processed over Singapore airspace, US airspace, or international waters. A healthcare provider cannot confirm that patient records remain within the jurisdiction required by regional regulations. These are not abstract concerns; they are the practical questions that enterprise procurement officers and compliance teams will be asking within years, not decades, if orbital compute advances even modestly toward commercial viability.
“The nation that defines the governance framework for orbital AI compute will shape the industry’s trajectory for a generation. Singapore is well-positioned to lead that effort — if it moves with urgency.”
The Johor Dimension: Regional Competition and Complementarity
Any analysis of Singapore’s AI infrastructure future must contend with Johor. Malaysia’s southernmost state has emerged as the principal beneficiary of Singapore’s datacenter constraints, hosting approximately 47 datacenters operational or in development, with national datacenter revenue projected to reach $1.87 billion by 2030. The Johor-Singapore Special Economic Zone (JS-SEZ), while still awaiting regulatory finalisation, presents a compelling dual-market proposition: land and energy costs a fraction of Singapore’s, proximity to Singapore’s subsea cable ecosystem, and access to Singapore’s talent and financial networks through the Zone’s cross-border mobility provisions.
The orbital datacenter debate intersects with the Johor question in instructive ways. If terrestrial compute economics continue to favour lower-cost jurisdictions for the bulk of AI training workloads, Singapore risks being outcompeted not by space but by its immediate neighbour for the commodity end of the compute market. Singapore’s response — rigorous quality thresholds, sustainability mandates, and positioning as the premium, trusted-jurisdiction option — is strategically coherent. But it requires that the premium market remain large enough to sustain the investment. The JS-SEZ and the broader Johor build-out are not threats to be resisted; they represent an extension of Singapore’s digital infrastructure ecosystem into a more cost-competitive geography. Singapore’s challenge is to ensure it captures the high-value, governance-sensitive workloads that cannot be replicated in Johor, while Johor absorbs the capacity-constrained, cost-sensitive overflow.
What Singapore Should Do: A Policy Agenda for the Orbital Era
The emergence of credible, if distant, orbital compute capability demands that Singapore begin treating space-based infrastructure as a strategic planning variable rather than a science fiction footnote. Several concrete policy directions merit consideration.
First, Singapore should engage proactively in the multilateral governance frameworks that will ultimately determine how orbital compute is regulated. The ITU, the UN Committee on the Peaceful Uses of Outer Space, and emerging bilateral discussions between the US, EU, and major Asia-Pacific economies will shape the jurisdictional and liability rules for satellite-based AI infrastructure. Singapore’s reputation as a sophisticated, neutral rule-setter in digital governance — exemplified by its leadership on cross-border data frameworks within ASEAN — is a comparative advantage that should be deployed in the orbital context before the rules are written by larger powers.
Second, the forthcoming Digital Infrastructure Act should anticipate orbital compute explicitly. This means developing clear guidance on data residency obligations when processing occurs on non-territorial infrastructure, liability frameworks for data breaches on orbital platforms used by Singapore enterprises, and procurement standards for government agencies considering cloud services that may route through orbital nodes. The EU has already articulated that digital sovereignty requires owning and operating infrastructure rather than depending on foreign providers; Singapore’s framework should be at least as sophisticated.
Third, Singapore’s investment in domestic compute infrastructure — the Jurong Island datacenter park, the NSCC supercomputer, GPU-as-a-Service offerings from Singtel and commercial providers — should be understood not merely as economic development but as a sovereignty hedge. If orbital compute does eventually commoditise AI inference in ways that undercut terrestrial economics, Singapore’s position as a hub for trusted, regulated, high-value computation will depend on the density and quality of its own infrastructure, not merely its regulatory reputation.
Fourth, Singapore should monitor the energy cost trajectory of orbital compute with genuine analytical rigour, rather than either reflexive enthusiasm or dismissal. Chanos is right that the current arithmetic does not favour orbital datacenters. But the arithmetic of solar energy in orbit is genuinely compelling over longer time horizons: a solar panel in geostationary orbit receives 8–10 times more energy per unit area than the same panel at Singapore’s latitude, with no atmospheric attenuation, no monsoon seasons, and no intermittency. If Starship achieves its cost targets — a significant conditional — the economics of orbital solar power shift in ways that even Chanos’s framework does not fully account for.
The Longer View: Infrastructure Ambition and National Scale
The SpaceX-xAI merger is, in the near term, primarily a financial transaction. xAI was burning approximately $1 billion per month competing against Google, OpenAI, Anthropic, and Microsoft. SpaceX generated an estimated $8 billion profit on $15–16 billion in revenue in 2025. The merger gives xAI access to SpaceX’s cash flow and launch infrastructure; it gives SpaceX a rationale for scaling Starship beyond Starlink deployment. The orbital datacenter narrative is the strategic framing — compelling and possibly transformative, but operating on a timeline that most serious analysts measure in decades, not years.
For Singapore, the relevant horizon is more immediate. The DC-CFA2 applications close in March 2026. The Jurong Island datacenter park is being designed now. The Digital Infrastructure Act will be tabled in Parliament within the year. The JS-SEZ framework is being finalised. These are the decisions that will shape Singapore’s competitive position in AI infrastructure through 2030 and beyond, and they will be made before any orbital datacenter is commercially operational.
Singapore’s great institutional strength has always been its capacity to take long-duration strategic bets while managing near-term constraints with precision. The orbital datacenter debate — stripped of its promotional excess and its reflexive scepticism alike — is ultimately a prompt for Singapore to do exactly that: to look up, not with starry-eyed credulity, but with the clear-eyed strategic analysis that has consistently allowed a city-state of 6 million people to punch far above its weight in the global economy. The sky, as it turns out, is neither the limit nor the destination. It is the next variable.
KEY FIGURES AT A GLANCE
SpaceX–xAI merger valuation: $1.25 trillion (February 2026) | Singapore datacenter market (2024): $4.16 billion | Projected market (2030): $5.60 billion | Current Singapore datacenter capacity: 1.4+ GW across 70+ facilities | Colocation vacancy rate: 1.4% (lowest in APAC) | DC-CFA2 new capacity: 200 MW (applications close March 2026) | Jurong Island park: 700 MW planned | AWS GDP contribution to Singapore by 2028: $23.7 billion | AI electricity cost (terrestrial): ~$600,000 per MW per annum
The author writes on technology strategy and digital infrastructure policy across the Asia-Pacific region. This article draws on publicly available corporate filings, regulatory announcements, and expert commentary and does not reflect the views of any government or commercial entity.