Southeast Asia may be winning the race to host the AI economy before it has answered a more uncomfortable question: how much power will the region actually gain from becoming indispensable to it?
The signs of success are increasingly difficult to miss. Data centers are rising across industrial estates, governments are competing for semiconductor and cloud investment, and electricity planning is becoming inseparable from digital policy. For countries that spent decades trying to move beyond commodity exports, low-cost manufacturing, or the downstream end of global technology chains, this looks like another opening for industrial upgrading.
Indonesia’s newest AI project captures both the promise and the ambiguity. On 6 August 2026, Indosat launched Zankore with Ooredoo Group, Nokia, and NVIDIA, targeting one gigawatt of NVIDIA DSX AI Factory capacity, with roughly 200 megawatts expected in the first half of 2027. Ooredoo is the lead investor, taking a 49 percent stake and committing approximately US$800 million, while NVIDIA provides the computing architecture and Nokia contributes the networking layer.
There is nothing inherently problematic about such an international configuration. No country builds an AI economy alone, and technological sovereignty does not require every component to be domestically produced. Zankore is revealing because it compresses Southeast Asia’s wider challenge into one project: global capital and technology are being assembled on Indonesian soil at enormous scale, while the longer-term distribution of capability and bargaining power remains open.
Stay ahead of the geopolitical week.
MD Briefing delivers expert analysis across five global fronts — the Indo-Pacific, energy, geoeconomics, European security, and the Middle East — every Monday morning. Free.
The language of “AI factories,” however, can encourage the wrong industrial analogy. A data center is not an electronics or automotive plant. A World Resources Institute review of more than 1,200 US data centers found that even the largest employ fewer than 150 permanent workers, with some employing as few as 25.
Employment alone is also the wrong test because compute functions as infrastructure. Like a port, telecommunications network, or electricity grid, much of its economic value comes from what other firms can do because it exists: process data locally, reduce latency, improve resilience, and gain easier access to computational capacity.
The harder test is whether domestic firms, universities, and researchers can convert that proximity into productivity, knowledge, and businesses of their own. A country can host enormous computing capacity and still gain relatively little strategic power if local users cannot run serious workloads on viable terms or if they lack the skills, data, and organizational capability to build on top of it.
Building links that survive the investment cycle
Malaysia and Indonesia are already experimenting with different versions of the same industrial-policy question: how can governments use investment and market access to make technological capability stay?
Malaysia’s National Semiconductor Strategy targets 60,000 highly skilled engineers by 2030, while new data center projects in Selangor face a 30 percent local-content requirement in areas including IC design and cooling systems.
Indonesia has pursued a similar logic through its domestic-content regime, known locally as TKDN (Tingkat Komponen Dalam Negeri), which sets domestic-value requirements for specified goods and services. Market access has repeatedly given Jakarta leverage over foreign technology firms, most visibly during the iPhone 16 dispute.
Neither approach guarantees industrial upgrading. A percentage can be satisfied through routine procurement or assembly without leaving much engineering knowledge behind. Local-content policies become meaningful when they pull domestic companies into technical activities whose skills, standards, and customer relationships remain useful beyond the original project; otherwise, governments may localize spending without localizing much capability.
That distinction is especially important for AI. Local compute creates potential value, but its wider economic impact depends on whether universities can run serious research workloads, domestic companies can obtain computing capacity on workable terms, and firms have enough technical and organizational sophistication to build products that other markets might eventually buy.
Southeast Asia has more cards than cheap power.
The region’s bargaining position begins well before an AI engineer enters a data center.
Southeast Asia sits across some of the Indo-Pacific’s most important production and maritime routes, connecting Northeast Asian manufacturing networks with the Indian Ocean and global markets. The physical AI economy still depends on moving chips, servers, transformers, cooling systems, and electrical equipment, which gives the region’s ports and industrial corridors renewed strategic relevance. The Philippines’ planned AI hub in New Clark City, for example, sits inside the Luzon Economic Corridor and is explicitly intended to connect the country more deeply with global advanced-technology supply chains.
ASEAN is also more than a location for foreign companies. Its banking, telecom, manufacturing, logistics, consumer, and public-sector markets represent large sources of future AI demand. Indonesia’s experience with global technology companies already demonstrates how access to a large domestic market can become negotiating leverage, and that logic becomes more important as AI shifts from experimentation towards enterprise deployment.
The region also enters this cycle with industrial capabilities that many emerging economies would have to build from a thinner base. Malaysia and Singapore occupy important positions in semiconductors and electronics; Vietnam and Thailand sit inside deep manufacturing networks, while Indonesia combines market scale with minerals, telecom infrastructure, and expanding compute.
Geography, demand, and industrial depth, therefore, give Southeast Asian governments something valuable to negotiate with before they master the full AI stack. The strategic mistake would be to use those assets mainly to attract the next facility rather than exchange that leverage for capabilities that strengthen the next negotiation.
The domestic bargain is getting heavier.
This matters because the physical cost of hosting compute becomes visible long before its strategic benefits are clear.
The International Energy Agency notes that an AI-focused data center can be ten times more capital-intensive than an aluminum smelter. In Malaysia, Ember projects electricity consumption from data centers to rise from 8.5 TWh in 2024 to 68 TWh in 2030.
Governments make real choices to accommodate that growth. They approve industrial land, expand grids, manage water allocations, and may provide fiscal support because they expect wider economic benefits to follow. Political tension grows when those commitments are obvious while the gains remain difficult for communities to identify.
Malaysia is already encountering that tension. Reuters reported in July 2026 that residents in Johor protested data center developments over water, electricity, and environmental concerns. Communities do not experience another gigawatt as an abstract improvement in national computing capacity; they experience construction, land conversion, pressure on infrastructure, and promises about employment.
An enclave-like outcome can emerge even when the facilities themselves succeed commercially, particularly if their physical demands remain local while much of the higher-value capability stays elsewhere. Yet the deeper strategic issue is whether the host country emerges with more room to negotiate when the next investor, chip architecture, or technology restriction arrives.
The technology bargain is already geopolitical.
That room for maneuver matters because access to AI’s most strategic layers is increasingly shaped by geopolitics rather than markets alone.
The Philippines provides a clear example. Its planned 4,000-acre AI industrial hub in New Clark City is explicitly linked to the US-led Pax Silica initiative. The U.S. Mission to ASEAN describes Pax Silica as an effort to strengthen secure supply chains stretching from critical minerals and semiconductors to AI-related infrastructure.
For Manila, that architecture can create valuable industrial opportunities while also showing how access to strategic technology increasingly arrives through economic-security relationships.
Advanced computing follows the same logic. On 10 July 2026, the US Bureau of Industry and Security authorized the UAE government and selected companies to receive advanced computing items, including AI chips and servers, license-free, explicitly tying that access to the broader US–UAE AI cooperation framework.
Southeast Asian countries do not all have to enter the same geopolitical architecture. Indonesia’s Zankore points to another configuration: Indonesian infrastructure and market access combined with Qatari capital, American compute, and Finnish networking technology. Different ASEAN economies will continue to assemble different combinations of external partners.
The strategic question is whether those partnerships expand national options over time. Countries with sophisticated suppliers, engineers who understand architecture rather than only operations, research institutions capable of using advanced compute, and domestic firms selling higher-value products can adapt when technology rules change. Countries whose principal contribution remains land, electricity, and market access have much less room to maneuver.
ASEAN itself is unlikely to become a single investment authority negotiating with hyperscalers for ten different economies. A more realistic regional role is to reduce avoidable fragmentation through common expectations on resource efficiency, transparency, and project standards, limiting the ability of highly mobile investors to arbitrage dramatically different national requirements.
The harder work still falls to member states: turning access, investment, and market scale into capabilities that widen their options in the next technology cycle.
Indonesia already has an early signal of what that can look like. On 13 August 2026, UGM, Indosat, and NVIDIA formally launched the UGM Indosat NVIDIA AI Technology Center, giving researchers access to advanced GPU infrastructure and creating a pathway for research, training and industry collaboration.
For Zankore, the larger test is whether that pattern spreads: whether Indonesian engineers move into higher-value layers of the stack, domestic researchers and firms gain usable access to its computing capacity, and Indonesian companies build products on top of it that can compete beyond the domestic market.
Southeast Asia could become indispensable to the global AI economy without becoming substantially more powerful within it.
