Kampong AI and the Geography of Platform Leverage

Y Combinator runs in batches. The batch structure is not primarily an educational design choice. It is a density mechanism. You put a hundred founders in the same physical space for three months, you compress the timeline so the social stakes are high, and you let proximity do the work that no amount of Slack channels or Zoom calls can replicate. The conversations that shift a company’s direction happen in hallways, over meals, in the fifteen minutes after a talk when people are still processing what they heard. YC understood this early: the programme’s real product is not the curriculum or the capital. It is the engineered density.

San Francisco’s AI ecosystem operates on a similar logic, but with a different mechanism. The density is market-driven rather than designed. Venture capital concentrates in a geography, talent follows capital, companies follow talent, and the resulting proximity produces the network effects that justify the concentration. The mechanism works, but the cost structure is specific: you relocate or you miss the network. The density is available to anyone who can afford to live in the Bay Area, work the hours, and be present in the rooms where decisions are made. That is a meaningful filter. It selects for a particular demographic, a particular life stage, a particular tolerance for instability.

The question nobody asks often enough is whether that filtering is a feature or a cost. And whether there are other ways to engineer density that select differently.

The village as infrastructure

Singapore announced Kampong AI in Budget 2026: two repurposed blocks within LaunchPad @ One-North, 14,500 square metres of workspace for up to 70 AI companies alongside 200 residential units where founders and talent can live. Pilot phase from March 2026, full completion by 2028. It sits within a broader package: a National AI Council chaired by the Prime Minister, a Champions of AI programme providing tailored company-level support, expanded tax incentives for AI investment, and free premium AI tool access for Singaporeans taking approved training courses.

The name is the interesting part. “Kampong” is the Malay word for village, and in Singapore it carries specific cultural weight. The kampong is a form of social organisation where proximity produces obligation, where the boundaries between domestic life and productive life are porous, and where the community functions as an economic unit rather than a collection of individual actors. Singapore demolished its last kampongs decades ago to build public housing. Choosing the word now, for an AI infrastructure project, is a deliberate act of framing.

The framing signals a theory of innovation density that differs from the Silicon Valley model in structural ways. The Valley’s density is a byproduct of market forces: capital, talent, and companies cluster because each element attracts the others, and the clustering happens to produce innovation spillovers. The kampong model engineers density intentionally. The government builds the physical infrastructure, designs the live-work integration, selects the companies that participate through the Champions programme, and coordinates the supporting ecosystem across tax policy, skilling programmes, and institutional partnerships.

Neither model is neutral. Both create access filters. The difference is in who designs the filter and what it optimises for.

Who gets in, and what that tells you

San Francisco’s filter is economic and cultural. You need to afford the cost of living, you need to be willing to relocate (often from another country), and you need to operate within a professional culture that rewards constant availability and in-person presence. This selects heavily for young, unattached, well-capitalised founders and engineers. It works well for that demographic. It systematically excludes founders with families, founders without personal savings or investor backing, and founders whose professional networks are rooted in other geographies.

Kampong AI’s filter is different, but it is still a filter. Two hundred residential units at One-North will select for a specific profile: probably early-career, probably comfortable with high-density living, probably already in Singapore or willing to relocate there. The units will not accommodate large families. They will suit people at a particular life stage, building intensively, working long hours, willing to live where they work.

The question worth asking is what each filter produces and what it misses. The Valley’s market-driven density has generated extraordinary innovation in a narrow band: consumer technology, enterprise SaaS, and frontier AI research. It has been less effective at producing innovation that requires deep local market knowledge, regulatory navigation across multiple jurisdictions, or the kind of cross-cultural operational expertise that Southeast Asian markets demand.

Singapore’s designed density is explicitly optimised for a different kind of innovation: AI applications built for multilingual, compliance-diverse, mobile-first markets, with proximity to companies like Grab, Razer, and Sea that operate at scale across the region, and to research institutions like NUS and A*STAR that are producing AI talent calibrated to local conditions. The kampong structure extends that proximity beyond office hours. When your infrastructure engineer and your fintech compliance lead share a lift, the collaboration pattern is different from when they share a coworking floor.

The time horizon difference

The deepest structural difference between market-driven density and state-designed density is the time horizon.

San Francisco’s AI ecosystem operates on venture capital timescales. Capital flows to where returns are fastest, talent follows capital, and the resulting density is optimised for speed: rapid iteration, quick product-market fit, fast scaling. When the cycle turns, the density thins. Companies leave, talent disperses, and the ecosystem contracts until the next cycle of capital concentration rebuilds it. This is not a flaw in the model; it is the model. The density is a function of current capital allocation, and capital allocation shifts.

Singapore’s infrastructure investments operate on a different clock. JTC, the agency building Kampong AI, is the same entity that built Jurong Industrial Estate in the 1960s, which took two decades to reach its intended scale and is still operating today. One-North itself was masterplanned in 2001 and has been developing incrementally for twenty-five years. The kampong model assumes that the density compounds over time: the relationships built in the first cohort of residents become the foundation for the next, the institutional knowledge accumulates, the ecosystem matures rather than cycles.

This is Singapore’s consistent infrastructure logic applied to a new domain. The country has always understood that building the physical and institutional conditions for economic activity is itself a source of competitive advantage, and that the returns on that building are generational rather than quarterly. Whether that logic transfers to AI innovation, which moves faster and less predictably than container shipping or semiconductor manufacturing, is the open question.

What this means for practitioners

For anyone making career or business decisions about where to build in AI, the Kampong AI model is worth watching for what it reveals about the geography of platform leverage.

The conventional advice is to be where the frontier models are being built, which still means a handful of Western technology centres. That advice assumes the valuable layer is the foundational one, and that proximity to model development is the primary advantage. What Singapore’s approach suggests is that the application, deployment, and integration layers have their own geographic logic, and that the infrastructure for those layers, physical proximity, regulatory expertise, multilingual operational capability, sector-specific market knowledge, is being built most deliberately in places that the conventional map of AI development overlooks.

The kampong is one version of that infrastructure. It will not suit everyone. It selects for a specific kind of builder at a specific life stage, in a specific geography, working within a specific regulatory and cultural environment. So does Sand Hill Road. So does Y Combinator’s batch model. Every density mechanism is also a selection mechanism, and the selection shapes what gets built.

The question for practitioners is which density produces the kind of leverage that compounds for you, given your specific expertise, market knowledge, and professional network. The answer to that question is becoming more geographically distributed than most career advice currently acknowledges. And the places that are engineering that density most intentionally, rather than waiting for market forces to produce it, may turn out to have built something that the market-driven model cannot easily replicate: a community where the social infrastructure and the technical infrastructure are the same thing.