JAKARTA - Governments across Southeast Asia have accelerated the rollout of AI training initiatives in recent years, but as more participants complete these programmes, questions remain on how they are actually absorbed into local labour markets and to which extent the employers have the capacity to make the most out of these skills.
Thailand’s National AI Committee, chaired by Prime Minister Paetongtarn Shinawatra, held its first meeting on May 1, 2025, and set a two-year plan to train 10 million general AI users, 90,000 AI professionals, and 50,000 AI developers.
Malaysia opened its National Training Week to all ASEAN citizens in June 2025, offering 65,000 free courses spanning AI, digitalisation, green technology, and leadership, with a target of 5 million participants. Singapore's National AI Strategy 2.0, announced in December 2023, wants to triple the country's AI practitioner pool to 15,000 over five years.
The Philippines, this year's ASEAN chair, launched its National AI Strategy Roadmap 2.0 through the Department of Trade and Industry in July 2024.
But while these targets outline government ambitions and training volume across the region, publicly available data on whether post-training outcomes such as job placement rates, timelines to employment remain limited.
Training Numbers vs. Placement Numbers
Government-reported numbers on AI workforce development largely focus on counts of courses completed, certificates issues, or training seats opened. But external analyses, including those by LinkedIn and Access Partnership, say that this focus is driven by rising demand for AI skills, which pushed policymakers to prioritise training numbers.
Their January 2025 analysis found that “57% of Southeast Asia’s workforce (164 million workers) may be impacted by GAI (generative AI).” It also found that the skills needed for jobs across Southeast Asian countries have changed by as much as 40% since 2016 and could change by as much as 72% by 2030.
Still, skills are changing faster than most public training programmes can be designed, funded, launched, and evaluated. The report also describes workers as being “impacted” by generative AI which does not necessarily indicate transitions into higher-paying roles or defined AI career pathways.
A June 2025 World Bank report, Future Jobs: Robots, Artificial Intelligence, and Digital Platforms in East Asia and Pacific, found that “only about 10% of jobs in the EAP (East Asia & Pacific) region involve tasks complementary to AI, much lower than the 30% share in advanced economies.” That makes placement data just as (if not more) relevant than training totals.
Governments, platforms, and education providers can report participation and course completion, but these figures do not show whether workers transition into paid roles that use these skills.
Employer data in the region also points in the same direction. The Digital Education Council’s AI in the Workplace 2025 report, produced with the Global Finance & Technology Network, found that only 3% of employers believe higher education is adequately preparing graduates for an AI-driven future.
Pearson, drawing on its Lost in Transition work on education-to-employment gaps, has also warned that in some parts of ASEAN, 75% of employers say recent graduates are “not job-ready.”
The ASEAN Business Advisory Council, in a June 2025 paper on talent mobility and the digital economy, also reported that Southeast Asia will need nearly 9 million additional ICT professionals by 2030, against a backdrop where, in some parts of ASEAN, fewer than half of vocational graduates already work in their trained field.
Country-level evidence makes that placement question even harder to avoid. In the Philippines, the World Economic Forum’s Future of Jobs Report 2025 found that 68% of workers are expected to require training to meet changing skill demands with only 38% having completed training at the time of reporting.
In Malaysia, official labour data showed around 1.95 million employed persons with tertiary education working in semi-skilled or low-skilled jobs in the third quarter of 2024, with the skill-related underemployment rate at 36.8%.
Singapore, however, is a different case. AI Singapore says the programme has a graduate placement rate of more than 90% with alumni entering roles such as AI engineer, MLOps engineer, and data scientist. That model is small, employer-linked, and outcome-tracked, and is closer to a placement pipeline than a mass training campaign
For the rest of the region, the number that matters comes after the course ends. Six months later, where did the trainee work? What was the job title? Did pay improve? Did their quality of life increase? Did the role use AI every week? Was the employer a local firm, a foreign vendor, a cloud provider, or an outsourced services company?
Such outcome-level data is not yet available.
The Missing Work
The same report that noted the AI ambition also mapped what is holding it back.
STT GDC and Ecosystm's Mind the Gap report, drawing on a survey of more than 600 enterprise and digital-native leaders across nine Asian markets, found that 49% of organisations report insufficient compute capacity, 53% cite inadequate storage, and 82% face network bottlenecks or latency issues. Then, about 83% say fewer than half of their IT locations are equipped to support AI workloads.
Chris Street, Group Chief Revenue Officer at STT GDC, said in the report that organisations across Asia are “discovering that AI success now depends less on training models and more on foundations,” and that without “scalable infrastructure and operational readiness in place, it becomes difficult to convert early AI ambition into consistent business value.”
The report found that 72% of organisations allocate less than 5% of their IT budgets to AI.
Of the 62 Philippine organisations surveyed in the same study, none had reached the “Leader” classification as 79% of them were still at Builder stage where pilots do exist but haven’t *really *scaled. 76% of them also reported critical AI talent gaps, 53% lacked in-house expertise to manage complex AI infrastructure, and 94% described their organisational culture toward AI as sceptical, cautious, or ambivalent.
Carlo Malana, President and CEO of STT GDC Philippines, said the data showed organisations hitting “an infrastructure and capability ceiling,” with compute, storage, and connectivity constraints “making it difficult to move from pilots to reliable, scaled deployment.”
Cisco’s AI Readiness Index 2025, published in October last year from a survey of 8,000 senior IT and business leaders across 30 markets, found that only 13% of organisations across Asia Pacific, Japan, and Greater China qualified as “Pacesetters,” meaning that they’re firms that can deploy AI at scale and show measurable returns. Only 15% had networks fully ready for AI and only 19% had fully centralised data infrastructure.
Ben Dawson, Cisco's Senior Vice President for Asia Pacific, Japan, and Greater China, noted that Pacesetters “are four times more likely to move pilots into production.” That does not mean the remaining organisations have no AI in production, but rather, suggest most are less ready to move pilots into scalable and measurable deployment.
Infrastructures That Are Still Uneven
While training can be done easily and spread through smartphones and online platforms, data centres, grid upgrades, and cooling systems are an entirely different kind of geography (and politics).
The International Energy Agency put global data centre electricity consumption at around 415 terawatt-hours in 2024 with projections reaching roughly 945 terawatt-hours by 2030.
For ASEAN specifically, Ember’s May 2025 report found that six major economies (Indonesia, Malaysia, the Philippines, Singapore, Thailand, and Vietnam) have 2.9 gigawatts of new data centre capacity in the pipeline. In Malaysia alone, data centre power demand is projected to rise from 9 terawatt-hours in 2024 to 68 terawatt-hours by 2030.
Still, that growth isn’t even spread evenly. Singapore, a tiny city-state country which Microsoft championed as one of the global leaders in AI adoption, held 1 gigawatt of operational data centre capacity as of February 2025 and ranked as the world’s fifth-largest data centre market, but after tightening approvals through its Data Centre Call for Application process in 2022, investment shifted.
Malaysia, primarily through Johor, held 507 megawatts of operational capacity as of early 2025, with 2.4 gigawatts of new capacity under development. Ember projects Malaysia’s data centre electricity demand to rise from 9 terawatt-hours in 2024 to 68 terawatt-hours by 2030, which would account for 30% of the country's national power consumption and exceed Singapore's total electricity use in 2023.
Indonesia’s data centre electricity use, on the other hand, is projected to quadruple from 6.7 to 26 terawatt-hours over the same period. Wood Mackenzie's December 2025 analysis projected the region’s total data centre power demand will reach 10.7 gigawatts by 2035, up from 2.6 gigawatts in 2025.
The ASEAN Guide for Sustainable Data Centre Development, published in December 2025, noted that per-rack energy use has risen from around 8 kilowatts in 2021 to as high as 50 kilowatts expected by 2027, driven by AI workloads, but the grid infrastructure in most of the region was not built for those densities.
Shabrina Nadhila, Energy Analyst at Ember, wrote in the May 2025 report that “ASEAN's booming data centre industry risks derailing energy transition goals without urgent action.”
So that means, Johor, greater Jakarta, Metro Manila, Bangkok, and Ho Chi Minh City are where the computer is being built, but a worker in Davao, Lampung, or Chiang Rai who finishes an AI course swims into a local market that may have no operational AI infrastructure for years.
The (Actually Very) Useful Benchmark
Southeast Asian governments are also, in most countries, major employers. If public sectors are not building AI into their own operations, a great share of potential demand for AI-trained workers simply does not translate into real jobs.
The OECD and Asian Development Bank’s Government at a Glance: Southeast Asia 2025, published in December 2025, found that “most surveyed SEA countries — six out of eight, or 75% — have no AI initiatives in their public sectors and are yet to develop use cases.” Only Indonesia and Singapore reported any operational use of AI in government at the time of the survey, which reflects conditions as of December 2023.
That means, six of the eight countries also had no instruments in place to govern the responsible use of AI algorithms in the public sector at all. The region scored 0.37 on the OECD Digital Government Index against a member-country average of 0.61, with only 23% of high-value government datasets easily accessible to the public, compared with 47% across OECD countries.
The OECD notes that the Philippines has implemented digital government initiatives since the survey's reference period. Governments do move. The broader picture, though, is very much cautious and slow-moving because a government whose high-value datasets are largely inaccessible to the public is some distance from building AI systems that put trained workers to use.
What they do count, however, is enrollment, and often the conversion rate into actual jobs stays off the scoreboard. The World Bank’s Digital Progress and Trends Report 2025 describes connectivity, compute, context/data, and competency/skills as foundations that need to be tracked together to understand a country’s real AI capacity.
Across Southeast Asia, though, the public numbers cited above are much stronger on training targets than on comparable tracking of placement, compute, connectivity, or usable data.
So can Southeast Asia support the AI workforce it is building? Not really, not yet at least. Governments can fund training programmes faster than power grids, data centres, and employers can absorb the workers coming out of them. The part that often goes missing is what happens after those workers finish the courses.
Without consistent tracking of placement, infrastructure, and employer demand, training numbers alone offer only a partial picture of the region’s AI readiness.
Reader Conversation