Data centers are becoming the factories of the AI economy, but they're not creating factories' jobs. Between 2021 and June 2025, Malaysia approved 143 data center projects, with 25 receiving special digital status representing RM144.4 billion in investment. Yet these projects were expected to create only 1,429 permanent jobs. By comparison, Malaysia approved RM120.5 billion in manufacturing investment across 1,108 projects in 2024 alone, generating 87,695 job opportunities. That's roughly 74 times more capital per job in data centers compared to manufacturing.

The disparity reveals a fundamental tension facing developing economies worldwide. As artificial intelligence demand explodes, countries are racing to host the massive server farms that power AI systems. But the bargain they're striking may leave them with the infrastructure's environmental costs while capturing minimal economic benefits. The International Energy Agency expects global data center electricity consumption to rise from around 485 terawatt-hours in 2025 to approximately 950 terawatt-hours by 2030, with AI-focused facilities tripling their consumption during that period.

What Is an "Enclave AI Economy" and Why Should Countries Care?

An enclave AI economy describes a situation where a country hosts globally connected digital infrastructure and absorbs its resource costs while capturing limited employment, domestic capability, fiscal value, and control over the systems being built. Infrastructure and environmental burdens become local, while intellectual property, high-value workloads, and strategic control often remain elsewhere. Servers may sit in Johor, Visakhapatnam, or Santiago, yet universities can still lack computing access, startups remain customers rather than partners, and domestic workers enter mainly through construction and maintenance roles.

This creates what researchers describe as a "double asymmetry." The physical footprint and resource demands are concentrated locally, but the profits and control flow to global technology companies. Indonesia illustrates the challenge. In April 2024, Microsoft announced a $1.7 billion investment over four years in cloud and AI infrastructure alongside AI-skilling opportunities for 840,000 people. Yet headline figures reveal little about whether participants move into high-value technical roles, whether universities and startups gain meaningful access to computing resources, or whether domestic firms enter the infrastructure's supply chains.

How Are Developing Nations Responding to Data Center Demands?

• Strengthening Environmental Standards: Selangor, Malaysia began demanding stronger sustainability standards and requiring 30 percent local content in areas such as integrated-circuit design and cooling systems, signaling that governments are pushing back against one-sided deals.
• Community Resistance: In February 2026, residents in Gelang Patah, Johor staged Malaysia's first protest against a planned data center complex, citing water pressure, dust, and loss of green space, demonstrating that public opposition can force project redesigns.
• Legal Challenges: Chile's experience shows how environmental courts can reverse permits; Google redesigned its Santiago data center project, replacing water-based cooling with air cooling after an environmental court partially reversed the original approval due to concerns about the stressed Central Santiago Aquifer.

These responses reflect a growing recognition that political access alone cannot sustain data center projects over decades. What companies call "social license",genuine community acceptance and environmental stewardship,functions as operational risk insurance. Community resistance can delay construction, raise financing costs, and trigger expensive redesigns. Political intermediaries may secure entry under one administration, but a deeper social bargain can survive drought, scrutiny, and political turnover.

Why Are Nuclear Power and Data Centers Becoming Linked?

The energy demands of AI data centers are so massive that traditional power sources cannot keep pace. This has accelerated nuclear power's role in the infrastructure equation. In the United States, the nuclear industry has produced a rapid succession of reactor tests, regulatory actions, financing commitments, and power agreements, with artificial intelligence and data center demand running through much of the activity.

The clearest example is Pennsylvania's Three Mile Island Unit 1, now called the Christopher M. Crane Clean Energy Center. In June 2026, the Federal Energy Regulatory Commission granted a waiver allowing Constellation Energy to transfer 760 megawatts of capacity interconnection rights from a natural gas plant to Crane, putting the project back on track for a targeted 2027 restart. Constellation has a 20-year agreement to sell Microsoft all of Crane's electricity, capacity, and clean-energy attributes, providing the company with a creditworthy customer and predictable revenue to justify the approximately $1.6 billion restart program.

On the advanced reactor front, Kairos Power's Hermes 2 project in Oak Ridge, Tennessee represents another binding commitment. Targeted to begin operating in 2030, the plant is expected to provide as much as 50 megawatts to the Tennessee Valley Authority grid through what Kairos and TVA describe as the country's first utility power purchase agreement for electricity from an advanced reactor. The broader Google-Kairos arrangement establishes a pathway for as much as 500 megawatts of advanced nuclear capacity by 2035.

How Are Hyperscalers Addressing Their Carbon Footprint?

As data center energy demands surge, hyperscalers like Microsoft and Google have shifted from treating carbon removal as a speculative venture to making it a core operational strategy. Between 2021 and 2024, these companies primarily focused on signing Power Purchase Agreements for wind and solar projects. But the materialization of AI's energy footprint forced a strategic pivot. With U.S. data center energy demand projected to surge by 130 percent by 2030, it became evident that intermittent renewables and grid upgrades alone could not keep pace.

Microsoft now leads corporate carbon removal purchases with 626,000 tonnes purchased, far outpacing other buyers. According to a Carbon Direct analysis, the four largest hyperscalers require investments of $70 to $80 billion in projects capable of sequestering over 105 million metric tons of CO2 per year to meet their climate goals, a volume that exceeds the entire global operational capture capacity as of April 2026.

This demand has catalyzed institutional-grade investment mechanisms. The Frontier buyer's club, which includes major hyperscalers, pledged to invest over $900 million into carbon dioxide removal companies. This signals a market shift from small, pilot-scale purchases to large, structured procurement designed to build a functioning supply chain. The massive $725 billion capital expenditure cycle projected for hyperscalers in 2026 is creating a parallel, non-discretionary spending stream for carbon removal, transforming carbon removal project financing from venture-backed speculation to bankable infrastructure underwritten by corporate offtake agreements.

What Determines Whether Data Center Investment Becomes Real Development?

The critical question for developing nations is not whether data centers arrive, but what they convert that capital into. For investors, the easiest route into a market often runs through political intermediaries who accelerate permits, consolidate land, connect companies to utilities, and provide protection. That route is faster than negotiating with communities, universities, workers, local governments, and environmental groups, but it is more fragile.

Control matters most: who sets the indicators, owns the data, and audits the figures; which communities are recognized as affected; and what consequences follow when investors miss their commitments. Digital public infrastructure could provide shared rails for information, commitments, and accountability. Yet under captured institutions, the same system can obscure rather than reveal. A government can build a public dashboard that discloses little. Employment targets can count temporary construction while obscuring permanent technical roles. Related companies can appear as local suppliers. Consultation can be digitized yet remain performative, while environmental reporting arrives after decisions become difficult to reverse.

The ability to govern AI investment is therefore becoming part of state resilience. That task is harder because AI investment no longer sits inside a simple bargain between governments and companies. Data centers operate within a multiplex digital ecosystem shaped by utilities, hyperscalers, financiers, local governments, universities, communities, and global clients. Governments may approve a project, but grid operators determine whether it can be powered. Lenders price social and environmental risk. Communities can contest resource use, while universities and firms influence whether infrastructure produces local capability.