Five Things In Vietnam's $1B AI Centre Plan

Vietnam's $1B Da Nang AI Centre Is Reshaping ASEAN's Compute Map

The USD 1 billion 100 MW AI data centre planned for Da Nang's High-Tech Park by Create Capital Vietnam, Haimaker.ai, and Vietnam DataGen has moved from term sheet to ground in the past month, and the project is now meaningfully changing the ASEAN compute map. The first 50 MW tranche is scheduled to commission in Q3 2027, with permitted expansion to 250 MW by 2029.

For a country that had no significant sovereign AI infrastructure 24 months ago, this is a step change. Here are five things the project actually changes for ASEAN's AI build-out.

1. The ASEAN Compute Map Is No Longer Singapore-Plus-Malaysia

For most of the past five years, ASEAN's sovereign compute story has been dominated by Singapore, with Malaysia as a meaningful second tier through Cyberjaya and Johor. That structure is now changing. Vietnam's USD 1 billion bet, alongside Thailand's Microsoft USD 1 billion commitment and Indonesia's smaller buildout in Batam, means ASEAN now has four serious sovereign compute markets.

The Da Nang project is the largest single AI-specific data centre announcement in Vietnam to date, and it represents roughly 21% of ASEAN's 2026 sovereign compute capex outside Singapore. The financing structure is also distinctive, with the Vietnamese state holding a minority stake through Vietnam DataGen rather than purely guaranteeing private capital.

2. Da Nang Is The Right Location For The Wrong Reasons

The Vietnamese government chose Da Nang over Ho Chi Minh City and Hanoi for explicit policy reasons. Da Nang has lower land cost, faster permitting, and a target central-Vietnam digital cluster strategy. Those reasons are sound on paper.

The practical issue is that Vietnam's AI demand sits in HCMC and Hanoi. Da Nang is a six-hour drive or two-hour flight from each, and the country's interior network capacity has historically struggled with high-volume AI workloads. The campus design includes a dedicated 200 Gbps fibre link to both endpoints, which addresses the bandwidth gap, but does not change the latency for real-time inference workloads.

This is not a fatal problem. Most AI training and inference is asynchronous, and the cost savings on land and power justify the location. But it does mean Da Nang is unlikely to attract latency-sensitive financial services workloads, which will continue to anchor in Singapore.

3. The Power Story Is The Most Interesting Part

The campus is designed to draw heavily from Vietnam Electricity renewable sources, including a 60 MW solar array commissioned alongside the campus and offshore wind feeding into the central grid. That is a real environmental positioning, and it differentiates Da Nang from hyperscaler builds in Singapore and Malaysia that are increasingly grid-power dependent.

The risk is grid stability. Vietnam has had public outages in the past 24 months, and the central grid has historically been less reliable than the southern industrial corridor. Create Capital and Haimaker have specified backup capacity at 130% of nameplate load, which is unusual and reflects how seriously the operators take the risk.

4. Haimaker Is The Most Interesting Operator

The most underappreciated detail is the operating partner. Haimaker.ai is a Hanoi-based AI infrastructure firm spun out of Viettel in 2024, and it has quietly become Vietnam's most credible AI infrastructure operator.

Haimaker is also the local partner of choice for Nvidia's Vietnam strategy, and it has been awarded the operating rights for two smaller prior data centre builds in Hanoi and HCMC. The Da Nang project is its largest mandate to date.

The firm's founder, Nguyen Quang Hieu, was previously head of Viettel's data centre group, and the team includes senior engineers poached from Singapore-based hyperscalers and from Korean SI firms. Haimaker is positioning itself as the operator that links Vietnamese sovereign compute to international hyperscaler customers, and the Da Nang project is the proof point.

5. The Customer Mix Will Be Mostly Regional, Not Vietnamese

The planned customer mix is structurally different from a Singapore data centre, which is dominated by global hyperscaler workloads. Da Nang's mix is:

• 40 percent domestic Vietnamese demand, anchored by VinAI, FPT Smart Cloud, and a handful of healthcare and logistics startups.
• 35 percent ASEAN regional customers, the strategically important slice that positions Da Nang as a cost-efficient alternative to Singapore for non-latency-sensitive workloads.
• 25 percent non-ASEAN, mostly Korean and Japanese.

The domestic Vietnamese demand reflects the rise of Vietnamese AI startups, including VinAI, FPT Smart Cloud, and a handful of younger firms in healthcare and logistics. The regional ASEAN demand is the most strategically important, because it positions Da Nang as a cost-efficient alternative to Singapore for non-latency-sensitive workloads.

What This Means For The Region

For readers tracking the wider ASEAN tourism AI push and Vietnam's AI Act, the Da Nang campus is the infrastructure layer of a broader Vietnamese AI strategy that is now coherent across regulation, talent, and compute. Vietnam is positioning itself as the most credible second-tier ASEAN AI market, behind Singapore but ahead of Thailand and Indonesia on infrastructure.

The wider read for the region is that ASEAN sovereign compute is no longer a single-city story. The cost-efficiency advantage that Da Nang offers will pull workloads that would otherwise have gone to Singapore by default, and the competition between ASEAN cities for AI workloads is now genuinely live.

Da Nang is not trying to compete with Singapore on latency. It is trying to compete on cost-per-token for training workloads, and on that measure it will be very competitive in ASEAN by 2028.

What Could Go Wrong

The Da Nang project is not without risk. The most acute is power, where the grid stability question is real. The second is talent, where Da Nang has a much smaller AI engineering pool than Hanoi or HCMC, and the campus operating staff will need to relocate. The third is timing, where the Q3 2027 commissioning is ambitious and depends on supply chain conditions for transformers, switchgear, and cooling that have been tight throughout 2025 and 2026.

For outside investors, the watch items in 2026 and early 2027 are:

1. Q3 2026 - power purchase agreement disclosure with Vietnam Electricity.
2. Q4 2026 - transformer delivery confirmation given the tight 2025-2026 supply chain.
3. Q1 2027 - customer pre-signing disclosures, especially the Korean and Japanese hyperscaler slice.

Each of these will signal whether the project is on track or slipping.