by Tan Aik Keong (AK)
Malaysia has decades in semiconductors but remains stronger in process engineering than in IP and advanced chip design. How did we get here — and is it too late to move up the value chain?
Malaysia's current position is a byproduct of our own success; for 50 years, we optimised for the "how" (process excellence) while outsourcing the "what" (IP and product design) to MNCs. That created a service-provider mindset, where our ecosystem was built for zero-defect manufacturing rather than the high-risk, high-reward nature of chip architecture. We didn't necessarily fail to innovate — our incentives were locked into a middle-income trap where funding models favoured physical assets over intangible IP. But it's far from too late. As the industry pivots toward AI accelerators and GPUs, the frontier of innovation has shifted toward advanced packaging — an area where Malaysia already holds significant global real estate. By using the National Semiconductor Strategy (NSS) to bridge our backend strength with frontend design, and building a venture-backed risk appetite for IC design, we can move from being a global factory to a strategic architect for the region's AI-driven future.
Do Malaysian children get enough exposure to robotics, sensors and coding in school — and can any long-term AI strategy work without fixing this at the foundation?
Exposure to robotics, sensors and coding in our schools remains fragmented, creating a digital divide where foundational AI literacy is often treated as an elective rather than a core competency. Without fixing this at the grassroots, any long-term national AI strategy risks building on hollow ground — we can't produce world-class chip designers or AI architects if students only experience technology as passive consumers. That's why initiatives like MyMahir National AI Council for Industry (MyMahir–NAICI), a joint effort between TalentCorp and MyDIGITAL, matter. As a council member of TalentCorp's MyMAHIR Future Skills Talent Council for the digital sector, I see firsthand that closing the talent gap needs a unified, industry-led roadmap aligning classroom learning with the actual hardware and software demands of the AI era. Centralising these resources through the MyMahir portal is finally moving us toward a whole-of-nation approach — making sure the next generation isn't just digitally literate, but industry-ready.
Are Malaysian OSAT firms ready to move into advanced packaging for AI chips?
I can't comment on that one.
The US and China dominate frontier AI. What kind of AI companies should Malaysia actually be building — and are there lessons from a company like Nokia?
Malaysia may never birth a global frontier-model giant, but our relevance lies in achieving AI sovereignty — the ability to control our own digital destiny. We should nurture a specialised ecosystem of local system integrators and vertical AI firms that build localised models on open-source foundations, tailored with Malaysian data and cultural context. Following the "Nokia lesson," we need to pivot toward becoming the critical infrastructure and connectivity backbone for the region — just as Nokia shifted from handsets to 5G and data-centre networking, Malaysia must evolve from a passive host of data centres into an active manager of a sovereign AI cloud. That means a local-first strategy: using local infrastructure (like the RM2 billion sovereign cloud announced in Budget 2026), managed by local talent, to solve high-value local use cases in healthcare, energy and government services. By focusing on this supporting "AI pick-and-shovel" industry — secure, audited, ethically-aligned AI systems — Malaysia can become ASEAN's strategic control tower for AI.
Malaysia's AI exposure is mostly indirect — OSAT, data centres, servers, networking. Is that a weakness, or a strength to build on?
Calling Malaysia's indirect AI exposure a weakness is too narrow a view — it's more accurately a strategic entry point. We lack listed AI pure-plays, but we hold the indispensable picks and shovels of the AI gold rush: our strength in OSAT, data centres and precision engineering. That's a pragmatic foundation, not just "better than nothing." Following the Nokia model of infrastructure reinvention, Malaysia should aim to be a system-architect hub, not just a process-engineering one. By using the National Semiconductor Strategy to move into high-value advanced packaging (2.5D/3D integration) and connecting that to our growing data-centre footprint in Johor and Cyberjaya, we can capture the hardware-software intersection — evolving from a passive host of servers into an active provider of sovereign AI cloud infrastructure. We don't need to be Silicon Valley to matter; mastering the physical layer where AI actually runs secures Malaysia's role as the control tower for AI deployment across Southeast Asia.
What has Agmo been doing on AI and talent development, and what's the ambition for the next few years?
At Agmo, we're not just talking about the AI future — we're building it, through two tracks: sovereign AI infrastructure, and grassroots AI talent development.
On infrastructure, we've built a suite of Made-in-Malaysia AI products designed to solve local problems with local data. SMEHero.AI is the first localised agentic AI platform for Malaysian SMEs, grounded in a trusted dataset of Malaysian law, labour practices and business culture, with a human-in-the-loop expert layer. Glem.AI is our enterprise-grade sovereign AI platform that lets government and corporates orchestrate multi-model, multi-modality AI (text, audio, vision) while keeping data securely within Malaysia's borders. AI Lens and DahReply automate the drudge work of business — vision AI for document processing, and conversational AI that handles customer service with natural, localised nuance.
On talent, we believe AI literacy has to be a lifelong journey. Through Agmo Academy we're running a kindergarten-to-career roadmap: Agmo Junior introduces robotics, coding and AI concepts to children, moving them from passive consumers to active creators; and in partnership with TalentCorp (MyMahir) and Yayasan Peneraju, we're training the next generation of agentic AI builders, with a curriculum focused on high-value skills like AI vibe coding and agentic AI building.
Our ambition is for Agmo to be the strategic architect of Malaysia's AI landscape — moving the country from using foreign AI to owning sovereign AI, and proving a homegrown company can build world-class, audited, ethical AI systems purpose-built for the complexity of the Southeast Asian economy, not just a copy of global models.
As a founding member of Konsortium AI Negara (KAIN), what has the consortium achieved, and what's the long-term goal?
Since launching, KAIN has become a strategic partner to government and a catalyst for the private sector. We published the KAIN Malaysia AI Report 2025, the nation's first comprehensive independent review of the AI ecosystem — a reference point for policymakers and investors to separate hype from actual industrial readiness. We've also mapped the Malaysia AI Ecosystem (v3), covering over 150 key players and growing across infrastructure, LLMs, vision AI and robotics, giving visibility to local startups that were previously working in silos. KAIN members have contributed to the NAIO working groups and Budget 2026 deliberations, advocating for a "government as lead customer" model to build local AI credibility.
Right now we're focused on strengthening the community — awareness initiatives, specialised training, industry meetups — while staying in ongoing dialogue with government to make sure national AI policy, investment frameworks and programmes are designed to actually benefit local players, so the AI dividend gets reinvested into local innovators.
Our long-term vision is the "200 & 100k" goal: over 200 high-growth AI companies and 100,000 AI professionals by 2030, built on four pillars — talent, collaboration, infrastructure and policy — with KAIN working to make Malaysia ASEAN's strategic control tower for AI.
Anything to add?
To achieve real AI sovereignty, government needs to move beyond being a lawmaker and become the lead customer for Malaysian AI products. Being the first buyer of local solutions gives startups the flywheel they need: real-world data access, institutional credibility, and a fast feedback loop for product iteration. That turns Malaysia into a genuine proving ground for world-class AI built by Malaysians, for Malaysians — anchoring our top talent at home, and producing export-ready products that have been battle-tested at national scale, able to compete globally with a "proven in Malaysia" seal.
Part of the AK AI Corner column. Originally published in The Edge on Feb 26, 2026.