Asia's AI Talent Reckoning: Why Four Futures Actually Mean Two Hard Choices

The World Economic Forum just published four scenarios for how AI and workforce readiness might reshape jobs by 2030. Reading through it, one thing becomes clear: for Asia, this isn't really about four futures. It's about two hard choices happening right now.

The WEF frames this around two variables: how fast AI advances, and how ready workers are to use it. Cross those variables and you get four neat quadrants with names like "Supercharged Progress" and "The Age of Displacement."

But here's what the report doesn't say loudly enough: Asia is already living in all four scenarios simultaneously.

Singapore and parts of China are racing toward Supercharged Progress. Manufacturing hubs across Southeast Asia are sliding into displacement faster than anyone wants to admit. India's tech sector is building the Co-Pilot Economy in Bangalore while most of the country hasn't started. And plenty of markets are just... stalled.

The question isn't which future Asia gets. The question is whether regional fragmentation becomes permanent, or whether there's still time to pull the lagging markets forward before the gap becomes unbridgeable.

Let me break down what the WEF scenarios actually reveal about Asia, and why the next 18 months matter more than the next five years...

What the scenarios actually tell us

The WEF built these futures around measuring things like AI capability (using benchmarks like MMMU scores), labour productivity growth, unemployment rates, and something they call "scaling of agentic AI" (currently at 23% of businesses globally).

Their baseline data is sobering:

• AI adoption jumped from 55% of businesses in 2022 to 88% in 2025
• Demand for AI literacy skills is up 70% year-on-year
• Labour productivity growth globally sits at about 1.5% annually
• But there's no standardised measure of what "AI-ready" actually means

That last point matters enormously for Asia. When workforce readiness varies as wildly as it does across the region, aggregate statistics hide more than they reveal.

The infrastructure reality nobody wants to discuss

Here's what jumped out at me in the report: > Projected AI capital expenditure of $1.3 trillion through 2030.

For Asia, this isn't just a big number. It's an infrastructure crisis waiting to happen.

The WEF scenarios all assume that compute, energy, and data infrastructure can scale to meet demand. But across Southeast Asia, South Asia, and parts of China, that assumption is already breaking down.

Data centre projects are getting delayed because grids can't handle the load. Compute costs are rising faster in Jakarta and Manila than in Singapore or Seoul. Energy-intensive AI workloads are butting up against the reality that most Asian grids weren't built for this.

According to Ember's recent analysis, In the ASEAN region, power consumption by data centres is projected to account for 2% to 30% of national power demands and will reach 68 TWh by 2030 from 9 TWh in 2024^.

Malaysia recently had to pause new data centre approvals in several regions due to grid constraints. According to The Straits Times, "Johor rejects nearly 30% of data centre applications to protect local resources" as of November 2024 (https://www.straitstimes.com/asia/se-asia/johor-rejects-nearly-30-of-data-centre-applications-to-protect-local-resources). Vietnam's ambitious AI strategy is running headfirst into the reality that its national grid loses nearly 10% of generated power to transmission inefficiencies. Even China, despite massive infrastructure investment, is seeing provinces ration power to data centres during peak demand periods.

Wood Mackenzie projects that In Southeast Asia, power demand for data centres is set to quadruple from 2.6 GW to 10.7 GW between 2025 and 2035, accounting for 3-4% of peak demand by 2035, up from 1% in 2025.

The "Supercharged Progress" scenario projects exponential AI advancement. But if the power literally can't flow to where it's needed, exponential advancement hits a hard ceiling. And that ceiling arrives well before 2030.

Asia's AI future will be decided by energy policy as much as education policy.

Why "displacement" looks different here

The WEF's "Age of Displacement" scenario warns about mass unemployment, collapsing consumer confidence, and concentration of power in a handful of tech platforms.

For formal sector workers in developed economies, that's a plausible nightmare.

For Asia? The nightmare is already here, it just looks different.

Between 60-80% of workers across Southeast Asia and South Asia operate in informal economies. Street vendors, gig workers, small shop owners, informal manufacturing. These people don't show up in unemployment statistics because they were never formally employed.

According to the ILO's 2025 update, "Approximately 1.3bn workers, around 66% of the total workforce, are engaged in informal jobs" in Asia-Pacific, with South Asia has the highest rate, with nearly 87% of workers in informal employment.

When AI-driven automation hits their income streams, it won't trigger a policy response about displaced workers needing retraining. It'll trigger urban migration, collapsing household incomes, and political instability that governments have no fiscal capacity to address.

We're already seeing early signals across the region, though tracking displacement in informal economies remains nearly impossible when reliable data doesn't exist.

The WEF report talks about universal basic income and AI dividends as policy responses. That might work in Scandinavia. In markets where government budgets are already stretched thin and tax collection is patchy, it's fantasy.

Consider the fiscal reality: tax-to-GDP ratios in Southeast Asia average around 15%, compared to 34% in OECD countries. Indonesia collects roughly 10% of GDP in tax revenue. The Philippines manages about 14%. These governments don't have the fiscal headroom for ambitious social safety nets, regardless of political will.

The real displacement risk in Asia isn't unemployment. It's the collapse of livelihoods that were never counted in the first place.

Where Asia actually has an edge

The "Co-Pilot Economy" scenario is the one where I think Asia could genuinely lead, but not for the reasons the WEF emphasises.

The report talks about this future requiring "widespread AI readiness" and "gradual AI progress." It frames it as augmentation over automation, human-AI teams, and task-specific integration.

What it doesn't emphasise enough: this is actually how most Asian businesses already adopt technology.

We've seen this pattern play out with mobile payments, e-commerce platforms, digital logistics. Asian markets don't wait for top-down infrastructure to be perfect. They adopt pragmatically, learn by doing, and build hybrid systems that work around constraints.

Vietnam's tech sector offers a telling example. Despite ranking relatively low on formal AI readiness indices, Vietnamese developers have achieved some of the highest rates of GitHub Copilot adoption globally. Why? Because they learned by doing, sharing knowledge through informal networks rather than waiting for structured training programmes.

That same pattern could repeat with AI more broadly.

The WEF report shows that AI can reduce task completion times by up to 80% in some roles. That productivity gain doesn't require perfect training systems or government programmes. It requires workers who are willing to experiment, employers who don't over-formalise adoption, and enough digital infrastructure to let people try things.

Asia has all of that in abundance in its emerging markets. The question is whether formal institutions will get out of the way fast enough to let bottom-up adoption happen.

If the Co-Pilot Economy is Asia's best-case scenario, it won't be built top-down. It'll emerge from workers and small businesses figuring it out themselves.

The geopolitical problem the report doesn't address

The WEF scenarios assume relatively open flows of technology, talent, capital, and data.

That assumption doesn't hold for Asia in 2026, and it'll hold even less by 2030.

US-China tech decoupling is accelerating. Data localisation requirements are fragmenting digital infrastructure. Talent mobility is getting restricted. Export controls are limiting access to advanced chips and compute.

The semiconductor restrictions are particularly stark. According to the Council on Foreign Relations, US Commerce Secretary Howard Lutnick testified that Huawei will produce only 200,000 AI chips in 2025 while Nvidia would make four-to-five million AI chips in 2025, which he described as double its 2024 production.

Meanwhile, CNAS reports that ASML, the leading maker of DUVi lithography systems, sold 70 percent of its DUVi lithography systems to Chinese entities in 2024, highlighting how export controls are being circumvented even as they reshape Asia's AI ecosystem.

Data fragmentation is even more pronounced. Asia now has at least 17 distinct data localisation regimes, each with different requirements for where data can be stored, processed, and transferred. This creates what researchers call "data friction", where the cost and complexity of moving data across borders becomes a binding constraint on AI development.

India's Digital Personal Data Protection Act requires specific categories of data to remain within national boundaries. Indonesia's data centre localisation rules mandate that certain data must be processed domestically. Vietnam requires government approval for cross-border data transfers. Thailand is implementing similar restrictions. Each rule makes sense individually, but collectively they're creating a fragmented digital infrastructure that makes regional AI collaboration nearly impossible.

This creates a world where different parts of Asia end up on entirely different technology stacks, with incompatible data systems, divergent AI governance models, and competing standards.

The result isn't one of the WEF's four scenarios. The result is fragmentation where leading markets race ahead, middle-tier markets get stuck choosing sides, and smaller economies get locked out entirely.

This is already happening. It's just not showing up in neat 2x2 matrices.

The real risk isn't ending up in "Stalled Progress." It's ending up in a region where progress happens in islands that can't connect to each other.

What actually matters in the next 18 months

The WEF report offers nine "no-regret" strategies for businesses. They're sensible: start small, align tech and talent strategies, invest in human-AI collaboration, strengthen data governance, anticipate talent needs.

Fine. But here's what matters more right now for Asia:

First, energy and compute infrastructure.

If you're making decisions about where to build, expand, or invest, grid capacity and energy costs need to be in the model. This isn't a 2030 problem. It's showing up in project delays today. Singapore's strategy offers a blueprint. The country announced in 2024 that it would limit data centre capacity growth and instead focus on making existing infrastructure more energy-efficient. They're betting on quality over quantity, which may actually position them better than markets racing to build capacity without considering power constraints.

Second, informal economy integration.

If you're a government or large employer, you need a plan for what happens to informal workers whose livelihoods get disrupted. Pretending they'll transition into formal training programmes is delusional.

Thailand's "Upskill Thailand" programme provides an interesting model. Rather than formal classroom training, they're embedding AI literacy modules directly into the apps that gig workers already use, reaching hundreds of thousands of delivery drivers, ride-share drivers, and freelancers where they actually work.

Third, regional coordination.

If you're a policy maker, the fragmentation problem needs to be treated as urgently as AI adoption itself. Different technology stacks across neighbouring markets creates permanent friction that compounds over time.

ASEAN's Digital Economy Framework Agreement, currently being negotiated, could help. But it needs more ambition. As it stands, it's a lowest-common-denominator compromise. What's needed is genuine harmonisation of data standards and cross-border flows.

Fourth, speed over perfection.

If you're a business leader, the Co-Pilot Economy doesn't wait for perfect training systems. Let workers experiment, capture what works, and scale it. Formal programmes will be obsolete before they launch.

The WEF scenarios are useful for framing long-term possibilities. But the choices that determine which future Asia lands in are being made in budget meetings and infrastructure decisions happening right now.

The uncomfortable truth

By 2030, Asia won't neatly land in one of these four futures. It'll have markets living in each of them, with widening gaps between them.

The question isn't which scenario we should aim for. The question is whether we're okay with permanent fragmentation, or whether there's still time to build bridges between the markets racing ahead and the ones falling behind.

Because once that gap becomes unbridgeable, it doesn't matter how advanced AI gets in Singapore or Shenzhen. The region as a whole will have failed.

That's the real scenario the WEF report doesn't model. And it's the one we're currently heading toward.

Read the full report: Four Futures for Jobs in the New Economy: AI and Talent in 2030^.