The National AI Roadmap 2021–2025 (AI-Rmap), published by MOSTI, was a genuinely ambitious document: six strategies, a national governance body, eleven flagship AI projects, and measurable KPIs across three two-year "horizons." Its vision — a high-tech, high-income Malaysia powered by a thriving AI ecosystem — was sound.
What it lacked was follow-through. By the time its 2025 end date arrived, the roadmap had been overtaken on almost every front that mattered: institutionally, technologically, and financially. This page sets out how, strategy by strategy, and why. It is not a claim that nothing happened — a great deal happened in Malaysian AI between 2021 and 2025. The point is narrower and more uncomfortable: very little of it happened because of this roadmap, and almost none of it was ever measured against the roadmap's own targets.
Evidence first — and silence counts
Each status below reflects the best publicly available evidence as of 2026: official announcements, budget documents, and reporting from credible outlets. Where the government published no data against a stated target, we record that as "no public data" rather than guessing a pass or fail. That absence is not a gap in our research — it is the finding. A roadmap with KPIs but no published scorecard is, in practice, unaccountable.
Four years, four hard truths
The six strategies, rated
A strategy-level summary. For the granular KPI-by-KPI breakdown, see the full targets scorecard →
| Strategy | Status | Why | |
|---|---|---|---|
| 1 · Establishing AI Governance | Superseded | The AI-CIU was never stood up as designed; governance passed to the Ministry of Digital and then NAIO. | Detail → |
| 2 · Advancing AI R&D | No public data | No consolidated reporting against the ring-fenced R&D funds, CoE targets, or GERD goal. Later Budget 2025 R&D money is NAIO-era, not roadmap delivery. | Strategy → |
| 3 · Escalating Digital Infrastructure | Delivered — private capital | The compute substrate arrived through a hyperscaler data-centre boom, not government programmes. The MSME cloud-adoption KPI was never reported. | Detail → |
| 4 · Fostering AI Talents | Partial | Large-scale skilling happened — but via later programmes like AIForMYFuture (400k+), not the roadmap's named AI-RUS platform. | Strategy → |
| 5 · Acculturating AI | Partial | Public AI literacy continued (e.g. AI untuk Rakyat), but the roadmap's specific mechanisms and KPIs were not tracked publicly. | Strategy → |
| 6 · National AI Innovation Ecosystem | Superseded | The AI-Catalyst / 11-consortia model did not become the operating structure; its role passed to NAIO working groups and the "AI Malaysia" platform. | Strategy → |
No strategy can be marked a clean, evidenced "delivered as planned." The closest — digital infrastructure — succeeded in spite of the roadmap rather than because of it.
Why the roadmap was overtaken
Not bad luck — five structural weaknesses, each of which a future plan would need to fix.
No one ever took the wheel
The roadmap's entire delivery model hinged on the AI Coordination & Implementation Unit — a standing body to own the strategies, coordinate ministries, and report progress. It was never established as specified. A plan with no accountable owner becomes a document, not a programme.
What happened to the AI-CIU →The mandate changed ministries twice
AI began under MOSTI. In 2023 a new Ministry of Digital was created and assumed the national digital and AI agenda; in December 2024 it launched the National AI Office. Each handoff reset priorities and left the AI-Rmap stranded with a ministry that no longer led on AI.
The institutional handoff → · NAIO vs the AI-CIU →It was written for the wrong AI
The AI-Rmap was drafted in 2020–2021 and framed around classical machine learning, analytics and IoT. ChatGPT's arrival in late 2022 redefined what "AI policy" had to address — generative models, foundation-model governance, compute access — none of which the roadmap anticipated. Officials themselves later said the landscape required a "refresh."
How GenAI dated the plan →The money came from somewhere else entirely
The roadmap imagined government-funded infrastructure and ring-fenced R&D in the low tens of millions of ringgit. What actually built Malaysia's AI capacity was private capital — Google, Microsoft, AWS, Oracle and Nvidia committing tens of billions to data centres. The state became a landlord and regulator of someone else's build-out, not the architect of its own.
The investment story →Nobody was keeping score
The roadmap defined KPIs for every strategic initiative and promised annual economic-impact assessments and a mid-term evaluation. We could find no published progress scorecard against those KPIs. Without public measurement, "living document" becomes a way of never being held to the last version — and that, more than any single failure, is why a four-year review like this one had to be assembled from the outside.
See the targets scorecard →Two questions worth their own page
This review is independent analysis by airmap.my. It assesses the National AI Roadmap 2021–2025 (a MOSTI publication; archived here) against publicly reported developments through 2026, including the establishment of the National AI Office, the creation of the Ministry of Digital, and reported data-centre investment.
Status assessments reflect available public evidence; where no progress data was published, this is stated rather than estimated. Full source list: airmap.my/sources. airmap.my is not affiliated with MOSTI, the Ministry of Digital, NAIO, or any Malaysian government agency.