Machinepower

Machinepower Index / Dossiers / Malaysia

Rank 31 of 50 · Tier 3

Rank 24–40 across the scoring rules tested. How this is tested

Malaysia

Compute Rentier — For nations whose main asset is the ground itself, meaning cheap clean power, land and stability: host other nations' compute and turn geography into rent, jobs and relevance.

Score4.45
Potential6.10
Bottleneck cost−1.65
Weakest cell: Sovereign Compute at 2.0 Binding layer: Watts Since the October fix: No change logged Edition: Q4 2026 · updated October 2026

How the Score, Potential and bottleneck cost are worked out: methodology.

The three layers

WattsMaterial power · holds the weakest cell
5.4 / 10

Energy, grids, data centres and silicon — the physical capacity to run models at scale.

Compute tracked by Epoch AI: Malaysia has 1.06 million H100-equivalents on its soil with the owner not reported. Hosted compute is shown, not scored.

WeightsIntelligence power
5.1 / 10

Models, alignment capability, standards influence and the talent that produces them.

WillPolitical power
7.8 / 10

The capacity to wire AI into the state and the economy while holding public consent.

Where the score comes from

LayerSub-dimensionBasisScore
WattsPlanning & PermittingAssessed8.2
Firm Power CapacityDataset5.0
Sovereign ComputeDataset2.0
Hardware & Silicon AccessDataset6.2
WeightsFrontier Model OutputAssessed3.2
Alignment CapabilityAssessed5.2
Regulatory StandingDataset7.8
Talent DensityAssessed4.3
WillPolicy CapacityDataset8.1
Public Sector AdoptionDataset6.5
Workforce TransitionDataset8.4
Public TrustDataset8.0

Assessed marks a measurement we judged rather than took from a dataset: 4 of the 12 here.

Sources

Ipsos AI Monitor 2026April 2026
Public Trust in AI
70% say AI products have more benefits than drawbacks

Share of adults agreeing that products and services using AI have more benefits than drawbacks; the Index scores public trust as 1 + share ÷ 10. Survey of 23,532 adults across 32 countries, March to April 2026.

View source →
YTL AI Labs2025
Frontier models (assessed)
3.2/10, assessed

Malaysia's main domestic model is ILMU, a Malay-language LLM from YTL AI Labs; it serves local needs but is far from the frontier.

View source →
YTL Power / NVIDIANovember 2025
Hosted AI Compute Capacity
NVIDIA GB200 campus operational (Johor)

Malaysia's first NVIDIA-powered AI data centre — a 1,640-acre campus in Kulai, Johor with a dedicated 500MW solar plant — went live in late 2025 with global hyperscalers among its tenants.

View source →
Malaysia Budget 2026October 2025
State Backing for the Hosting Play
RM5.9B (~$1.4B) AI allocation

Budget 2026 funds Malaysia's positioning as the regional hub for AI infrastructure and data-centre capacity — rent-earning strategy backed with public money.

View source →
DIGITIMESMay 2026
Foreign Tenant Concentration
Hyperscalers & Chinese cloud giants

Johor has become one of the world's fastest-growing data-centre corridors, hosting US hyperscalers and Chinese cloud providers side by side — compute on Malaysian soil, commanded from abroad.

View source →
IMF2023 data, published June 2024
AI Preparedness Index: human capital and labour market policies
0.169 of a possible 0.25 · 8.4/10 in the Index

Used for the workforce measurement. The IMF combines schooling, pay, labour mobility and active labour-market policies; the Index rescales it so the weakest of 173 economies scores 0 and the strongest 10.

View source →
World BankDecember 2025
Business Ready 2025: business location and utility services
Business location 61.5/100 · utility services 85.9/100

A cross-check for planning and permitting, which stays assessed: it scores building permits and electricity connections for ordinary firms, not grid-scale connections for data centres.

View source →
Ember2025
Firm generating capacity
39 GW (2025)

Fossil, nuclear, hydro and other dispatchable plant; averaged with generation for firm power.

View source →
Oxford InsightsMay 2026
Policy capacity
80/100

Feeds policy capacity.

View source →

Bottlenecks

These hold Malaysia below its Potential. Its two lowest scores cost it the most: sovereign compute at 2.0 (part of Watts) and frontier model output at 3.2 (part of Weights).

Watts (Material)Sovereign Compute

Assessment

The clearest live example of the Compute Rentier playbook. Johor has become one of the fastest-growing data-centre corridors on Earth — hosting Singapore's overspill, global hyperscalers, and Chinese cloud giants on Malaysian land, power, and tax terms. The state earns rent and leverage from compute it hosts but does not command: domestic model capability remains minimal.

The art of the possible

Malaysia's rentier position is a foundation, not a ceiling: the skills, power infrastructure, and capital flowing through Johor could bootstrap sovereign capability — regional models, domestic cloud champions — the way hosting economies have historically climbed the value chain.

Strategic play — Compute Rentier

Bank the rent Johor earns and invest it in depth: Penang already holds about 13% of the world's chip assembly, test and packaging, and concentrating that share is the most plausible route to a real chokepoint, while hosting revenue funds regional models and domestic cloud champions.

Leverage — the nearest position. Roughly 13% of the world's chip assembly, test and packaging: the largest share of the chain held by any nation outside the register. Many firms rather than one lever, which is why it is not a chokepoint. Depth in Penang is the route to one.

Open Malaysia in the live Index → — every source behind every number, and the comparison tool.