Aolani and Nvidia Plan 22,000-GPU AI Expansion Across Malaysia and the Philippines

Southeast Asia is set for another major expansion of artificial-intelligence infrastructure. Aolani Cloud plans to deploy more than 22,000 Nvidia Blackwell Ultra GPUs across Malaysia and the Philippines, with the rollout expected to begin in 2027.

Aolani’s announcement describes the project as an expansion of sovereign AI infrastructure designed to serve governments and enterprises in the region. The deployment is notable not only for its scale but for its geography: Malaysia and the Philippines are emerging as participants in an AI infrastructure market historically concentrated in the United States, China and a small number of global cloud hubs.

For the Philippines in particular, the project could expand local access to high-end accelerated computing. But the economic value will depend on much more than the number of GPUs installed.

Why 22,000 GPUs is significant

Modern AI training and inference workloads depend on large clusters of accelerators connected through high-speed networks. Nvidia’s Blackwell Ultra platform is designed for demanding generative-AI and reasoning workloads, including large-scale inference.

A deployment exceeding 22,000 GPUs represents substantial compute capacity. It can support model training, fine-tuning and inference for customers that might otherwise need to obtain capacity from data centers outside the region.

Keeping workloads closer to users can reduce latency and may help organizations meet data-residency or sovereignty requirements.

What sovereign AI means in Southeast Asia

“Sovereign AI” generally refers to a country or organization maintaining greater control over the infrastructure, data and models used for artificial intelligence. The concept has gained importance as governments consider whether strategically important AI workloads should depend entirely on foreign cloud regions.

For Southeast Asian countries, local infrastructure can support models trained on regional languages and datasets while giving regulated industries more options for where sensitive data is processed.

Infrastructure alone does not create sovereignty, however. Chip supply, cloud software, networking technology and model dependencies can still cross borders. A locally located GPU cluster is one component of a larger technology stack.

Why Malaysia has become a data-center hub

Malaysia has attracted substantial data-center investment because of its proximity to Singapore, available land and growing connectivity. Johor in particular has developed into a major data-center market as operators look for alternatives to Singapore’s constrained land and power supply.

The growth also creates pressure on electricity and water systems. Large AI clusters require significant power, and the economics of a facility depend on reliable energy at competitive prices.

That means Malaysia’s ability to convert data-center construction into long-term economic value will depend partly on energy planning and local supply-chain development.

The Philippines has a different opportunity

The Philippines has a large services economy, a young digital workforce and strong demand for cloud connectivity, but it has historically had less hyperscale computing capacity than larger Asian hubs.

Local access to advanced GPUs could help universities, startups, business-process outsourcing companies and large enterprises experiment with more demanding AI workloads without relying exclusively on overseas infrastructure.

The country also has an opportunity to connect AI infrastructure with its existing strength in customer service and business-process operations. As AI agents become more capable, companies may increasingly combine human service teams with automated systems. Local compute can support that transition.

Power will determine how much capacity becomes usable

Announcing thousands of GPUs is only the beginning. AI accelerators require data centers capable of delivering dense, reliable electricity and removing enormous amounts of heat.

Blackwell-class systems can require sophisticated liquid cooling and high-capacity electrical infrastructure. A project’s real deployment schedule therefore depends on substations, transformers, cooling systems, networking and construction—not just chip delivery.

This is increasingly the central constraint across the AI industry. Compute capacity is becoming inseparable from energy capacity.

Local economic benefits are not automatic

Large data centers involve substantial capital investment but do not necessarily employ huge numbers of people once construction is complete. The larger economic benefit comes when infrastructure attracts businesses that build products and services on top of the compute.

For Malaysia and the Philippines, the key question is whether regional companies can use the new capacity to create models, applications and services rather than simply hosting workloads for global customers.

Training programs, university access and startup-friendly pricing can influence whether the benefits spread beyond a small number of large enterprises.

Nvidia gains from geographic expansion

For Nvidia, projects like Aolani’s extend demand into markets outside the traditional hyperscale cloud providers. Governments and regional cloud companies increasingly want their own AI infrastructure, creating a new customer category for advanced accelerators.

The company benefits not only from GPU sales but from the broader ecosystem around its networking and software platforms. Large deployments can reinforce that ecosystem as developers optimize applications around Nvidia hardware.

What to watch before 2027

The project’s most important milestones will be specific site announcements, power commitments, construction progress and confirmed delivery schedules. Large infrastructure projects can change substantially between announcement and operation.

It will also be worth watching which customers use the capacity. Government workloads, enterprise AI, startup access and academic research would each create different economic impacts.

Pricing will matter as well. Local infrastructure is most transformative when regional organizations can afford to use it.

Bottom line

Aolani’s planned 22,000-plus-GPU expansion is another sign that the geography of AI infrastructure is widening. Malaysia is strengthening its position as a data-center hub, while the Philippines could gain access to significantly more advanced local compute.

The headline GPU count is impressive, but the long-term value will be determined by electricity, cooling, connectivity and who actually gets to use the capacity. If those pieces come together, Southeast Asia could become more than a location for data centers—it could become a larger participant in building and operating the AI systems those data centers support.