Horizon

The World Bank Says AI Hardware Now Drives Most of Malaysia's Export Growth

Malaysia's 2026 forecast rises to 5.1%, but the same report warns that a pullback in AI capital spending would remove a key support.

Marcus Tan

3 min read ·

2

The World Bank raised its 2026 growth forecast for Malaysia by 0.7 percentage points to 5.1% in its East Asia and Pacific Economic Update, released on 6 October. The main reason is AI. Demand for AI-related electrical and electronic products and ICT services pushed Malaysia's GDP growth up to 6% in the second quarter from 5.4% in the first, Xinhua reported. For the wider region, the Bank now expects 4.5% growth this year, up 0.3 points, and among the larger economies Vietnam gets the biggest upgrade, 1.1 points to 7.4%.

The figure that matters most for Malaysia is about concentration, not growth. According to the report, AI-related goods accounted for more than half of total export growth in most of the region's economies by April, and more than 70% in Malaysia, the Philippines, Thailand and Vietnam. China, Indonesia, Malaysia, the Philippines, Thailand and Vietnam together exported US$1.4tn of AI-related goods in the twelve months to April.

Picks, shovels and one customer base

Malaysia's role in this boom is clear. It packages and tests chips, assembles servers and components, and hosts the data centres in Johor and the Klang Valley. All of that is real income and real jobs. But the report spells out what the dependence means: the regional upswing rests largely on capital spending decisions made by a handful of US technology companies and the suppliers around them.

The World Bank's risk section is worth reading slowly. It cites estimates that gross AI-related capital spending has reached 6% of US GDP, similar to the share IT investment reached at its peak in 2000. Of the US$2.9tn of AI capital expenditure planned for 2025 to 2028, about US$800bn is expected to be financed by private credit, a market that is less visible and has not been tested by a severe downturn. A reversal, the Bank says, would remove a key pillar holding up growth in many countries. AMRO, the ASEAN+3 Macroeconomic Research Office, made a similar point in July: if global technology investment slowed to its 2024 pace, regional growth could fall to 2.5% in 2027, The Star reported, the weakest since the Asian financial crisis outside the pandemic years.

Exporting AI is not the same as using it

The report also makes a point that gets less attention: making AI hardware and adopting AI are different things. The Bank's press release notes that businesses in the region face high costs, thin expertise, and security and privacy worries when they try to use the technology. The Bank sees wider adoption as the route to productivity gains and better jobs. In other words, the export boom is not automatically turning into a productivity boom at home.

That fits what I hear from Malaysian founders. The country is world-class at building the physical layer of AI and only average at using it. Most of the value from a GPU that leaves a Penang test floor goes to whoever runs software on it, and that is rarely a Malaysian company.

Why the timing matters

The report lands three days before Budget 2027 is tabled on 9 October. Deputy Finance Minister Liew Chin Tong has called AI a once-in-a-generation boom and said the government will look at how semiconductors and data centres can deliver more direct benefit to the country, The Star reported last month. The World Bank's numbers make the case for that, but they also point to where the money is most needed.

A budget that adds more incentives for data centre construction and chip assembly would increase a concentration the World Bank has just warned about. A budget that funds local adoption would hedge it: compute credits for SMEs, procurement that buys from Malaysian software companies, and support for the design and IP work that the Arm programme is meant to seed.

I expect the budget to do some of both. The test is the balance. If more than 70% of your export growth comes from one technology cycle, the best time to diversify is while that cycle is still going up.

Sources

Responses (2)

Sign in to leave a response.

  • Sofia Marchetti

    The private credit figure is the one I would underline. US$800bn of opaque financing behind one technology cycle is exactly where surprises tend to come from.

  • Fakhrul

    Compute credits only help if they can be spent on capacity priced for small teams. Otherwise they become a subsidy to hyperscalers by another route.

More from Marcus Tan

Recommended from Horizon