Private Capital Investment Opportunity in Vietnam Data Center




What is within a Data Center?

A data center (DC), in its simplest definition by IBM, is a physical room, building, or facility that houses IT infrastructure for building, running, and delivering applications and services. It also stores and manages the data associated with those applications and services. Beneath is a figure from CBRE showing where DC is used in our daily life.

Historically, using a computer, accessing a website, or running an app has always involved a process of storing and transferring data behind the scenes. In the age of AI, this same process hasn't become more frequent so much as far more computationally demanding AI workloads require drastically more processing power per request than a typical website or app ever did. Because of this, DC operators must ensure concurrent maintainability and fault tolerance at all times, in order to maintain the uptime (often 99.999%) that their clients require. DCs come in four main sizes: colocation, enterprise, hyperscale and edge.


Colocation: facilities are owned and operated by a third party, which rents out space to multiple tenants. The operator handles the building, power, cooling, and security, while each tenant brings and manages its own servers.


Enterprise: data centers are owned and operated by a single company purely for its own internal usage.

Hyperscale: facilities are massive data centers operated by a single large cloud or tech company such as AWS, Google, or Microsoft, designed to support enormous workloads, with the largest facilities.


Edge: facilities are small, decentralized data centers placed physically close to end users specifically to reduce latency for real-time applications like gaming, video calls, and IoT devices. Unlike colocation, enterprise, and hyperscale (which are categorized by who owns and uses the facility), edge is classified via proximity to their end user.


To understand the full scope of why a DC is needed, and what the buzz surrounding data centers is really about, the definition above needs to be broken down further. I've split a typical DC into four main functions: Energy & Power, Heat Removal Systems, White Space, and Network & Connectivity.


(Due to the nature of this writing talking about Vietnam, all technology listed here are only those that are applicable to the Vietnamese market, more advanced DC technology are used in the developed world)


White Space 


White space is the most important part of a data center, the only part of the facility that directly generates revenue for investors, as tenants pay based on the floor area (square meters or square feet) or power capacity (kW/MW) allocated within it. Contrasted with "gray space" (the non-revenue-generating mechanical and electrical support rooms), white space represents the secure, climate-controlled room where the core computing payload resides.


Within the white space, computer servers process, store, and route data requests; these servers act as the ‘gatekeeper’ to data. When a user or application requests information, the server authenticates the session, executes the necessary logic, and delivers the payload. If problems occur during this data access attempt, the response is often a 404 Not Found (the requested file does not exist on that server), a 403 Forbidden (access permissions are denied), or geographical blocking enforced by a state or IP filter. (If physical or hardware infrastructure fails entirely, the connection instead results in a timeout or total unreachability).


These servers are stacked side by side and vertically in standardized steel cabinets called racks (typically 42U to 48U in height). The individual servers within these racks are connected to one another and to the broader facility network via Ethernet and fiber-optic cables.


Network


Personal computers or devices first send a request to a server computer located in a DC.. Once access is approved, the data travels back via a transmission route, which works as follows: the personal device connects to a local network (Wi-Fi or cellular data) and sends the request to a local access point which is a router or the nearest cell tower.


From here, it enters the regional internet service provider's (ISP) network, in Vietnam's case, this would be Viettel, VNPT, FPT, and others. These ISPs, via underground fiber-optic cable running below the city, transfer this data to a DC with the appropriate server computers. There, after passing through the meet-me room, the data is routed to the designated server, which processes the request and sends the response back.


For international routes, ISPs send this data to cable landing stations, which then transfer it through underwater fiber-optic cables, carrying the data there and back. For example, a prompt entered into ChatGPT or Claude from Vietnam goes through this entire process in a matter of milliseconds.


Websites and apps, however, are not the only use of DCs. Things such as video games, video calls, and online payments, among many others, also rely on this process, which is why latency is critical. With AI prompting or a Google search, milliseconds of delay may not be critical; however, for a video game, this could be noticeable. 


Power & Energy


The biggest talking point within global digitalization and the rise of DCs is power. Given the immense power demand created by current AI workloads, DCs must be able to keep up with that demand at all costs to ensure uptime for clients.


Often, constructing a DC's physical building doesn't take nearly as long as securing approval and completing construction of the high-voltage transmission lines needed to bring adequate electricity from generation facilities to the DC.

Within power, the industry uses the term "N" to describe the exact capacity needed to run a facility with zero redundancy. The best-run DCs typically build in N+1 which means one additional unit of backup capacity beyond what's needed, or 2N acting like an entirely independent power system running in parallel, to protect against failure.


Depending on the scale of the DC, different strategies to get power apply. For a standard colocation or enterprise-sized DC, connecting directly to the regional electric utility via a substation or through Vietnam's grid-connected DPPA model under Decree 80/57, would often allow for the minimization of capital expenditure and maximized efficiency.


For hyperscalers and major DC campuses, full reliance on the public grid can introduce capacity constraints and demand curtailment risk during periods of grid stress. As a result, they may look toward private transmission lines or behind-the-meter power supply which mean having a transmission line running directly from a generation source to the DC, bypassing public utilities entirely. Understanding the bureaucracy behind building transmission lines, often DC would just locate themselves directly adjacent to, or within, the generation facility itself, sourcing power straight from a nuclear reactor, an LNG plant, or a battery storage facility, minimizing transmission distance to nearly zero.


Heat Removal System


With high energy usage, a large amount of heat is released as a byproduct. Electronic components have a maximum safe operating temperature; beyond that threshold, chips slow themselves down to avoid damage, which lowers the overall efficiency of the DC.


To resolve this, DCs must have a cooling system for their servers. Historically, with simpler CPU-based computers, air cooling via Computer Room Air Conditioners (CRAC) and Computer Room Air Handlers (CRAH) together with a chiller was enough. CRAC units work like any normal air conditioner, cooling air directly and blowing it into the room at a low temperature. CRAH units work slightly differently: the chiller acts as a refrigerator, cooling water, and the CRAH then uses that chilled water to cool air, which is blown into the white space to cool the servers.


However, with more advanced GPU/TPU servers drawing far more power and generating far more heat as a byproduct, traditional air conditioning no longer works, since air is a poor heat conductor compared to liquid.


For DCs to compete with one another and minimize costs, they must aim for the lowest possible power usage effectiveness (PUE) if they want to remain competitive in the long run (ensuring the most amount of energy being used for the white space as possible). With this in mind, some major DCs look at liquid cooling alternatives such as direct-to-chip cooling, which works by running liquid through tubes to a such as the GPU, removing heat and transferring it away. This, however, requires a large amount of capital if implemented at scale, so some operators still opt for CRAH.


The decision of which system to use depends on the specific market and operating conditions surrounding the DC. Oftentimes, a combination of both CRAH and liquid cooling is optimal. By establishing a DC close to a power source, an operator could create a power-access moat, and if combined with a carrier-neutral connectivity hub, this creates a further competitive moat. 


Revenue Model


Data centres typically operate under two lease structures: the triple-net (NNN) lease and the modified gross lease. In both, direct power and water costs are passed through to the tenant.

Under NNN leases, tenants also absorb property tax, insurance, and facility maintenance in exchange for a lower base rent. Under modified gross leases, tenants pay a higher base rent, and the operator retains responsibility for maintenance, staffing, security, and insurance out of that rental income.


Beyond base rent, interconnection is the highest-margin revenue stream. Because each server hosts distinct data, tenants require interconnecting ethernet or fibre optic cabling to transfer between systems at low latency. Operators facilitate this at the Meet-Me Room (MMR) for a recurring monthly fee, making it the most profitable line item within a data centre facility.


For non-NNN leases, operators may also generate managed service fees by providing on-site staffing for maintenance and operations, typically billed at an hourly rate.


That said, the core of the revenue model remains lease income. Sustained occupancy is therefore the single most important driver of returns for data centre investors.


Demand & Market


The AI boom and the growth of cloud services is driving the strong demand for DCs in Asia-Pacific as a whole, with CBRE believing that investor confidence in the asset class would continue to strengthen over the remainder of the asset class in the coming decade. 


Demand for DC in the age of AI stems from two main things: AI training and AI interference.


AI training is a process whereby AI runs through thousands of Graphics Processing Units (GPUs) which contain thousands of basic cores (processing units within a computer chip that read instructions and perform calculations) designed to perform thousands of basic calculations simultaneously or a Tensor Processing Unit (TPU) which is a chip highly specialized for AI. This then allows AI to learn from a vast amount of data at a rapid rate to be knowledgeable enough for usage. 


AI Interference refers to the process of prompting, where every time a prompt is made, Large Language Models (LLMs) run the query through billions of mathematical parameters to generate a token of response. With the wide adaptation of AI from industry to household, demand for DC is ever increasing. Though historically, AI training is the process that has been the core driver of DC demand, moving forward it is expected that AI interference will also be capturing a large portion of the demand too. 


As being one of the fastest growing economies in the world, Vietnam positions itself at the forefront of the growth, though it's not the biggest player in Asia when it comes to DC, as CBRE put it, the ‘Rise of digitalization in Vietnam’. With growth being driven from 4 main drivers of:

Information & Communication Technology


Vietnam is among the countries with the highest number of internet users in the Asia Pacific region, with an estimated 101 million internet users in 2024, representing a 79% internet penetration rate. Internet usage is predominantly mobile-based due to high smartphone penetration, with 161.6 million active cellular mobile connections recorded in early 2023, equivalent to about 164% of the total population. 


E-Commerce


Many local businesses moved online during the global pandemic, with popular platforms like Shopee and TikTok offering cash-on-delivery convenience that suits local habits well. In 2020, Vietnam unveiled its Masterplan for National E-Commerce Development (2021–2025), which set a government target for the sector to reach USD 35 billion by 2025. 


Fintech & E-Payment


While Vietnam remains a cash-dominant economy, the digital payments space is getting increasingly crowded, with the State Bank of Vietnam having issued licenses to 27 intermediary payment service providers, most of which are e-wallet services. E-commerce and mobile payment users remain concentrated in Hanoi and Ho Chi Minh City, and the transaction value of digital payments was forecast to hit USD 32.2 billion in 2024. 


Cloud Services


Cloud services hold strong potential to provide economies of scale, enhance operational efficiency, and generate cost savings for adopters. Vietnam's cloud storage market was valued at USD 198.87 million in 2024 and is anticipated to reach USD 650.33 million by 2030, growing at a CAGR of 21.65%, driven by accelerating digital transformation across the country. 





With Vietnam's digital economy and internet population both growing at a fast pace and projections placing the country's digital economy Gross Merchandise Value at up to US$200 billion by 2030, Vietnam's DC market presents, more than ever, a strong case for investment. 



It is also projected that by 2030, Vietnam will see a 5.6x increase in DC capacity, growing from just over 100 MW in 2025 to 589 MW. The south is expected to capture the largest share of this increase, particularly in areas such as HCMC and Binh Duong, supporting the economic activity concentrated in this region. 


Overall the demand story of Vietnam shows us a story of a country in its rapidly developing stage, much of it is captured by the digital economy. From rising internet penetration and mobile connectivity, to the growth of e-commerce, fintech, and cloud adoption, each of these threads ultimately converges on the same requirement from computing and storage capacity, delivered from a data center. Looking from that side, the demand is almost certain, what remains is whether Vietnam's power infrastructure can actually keep pace with that demand, which is what would determine the construction of any DC at all.


Economic of DC Investment


Similar to commercial real estate, DC economics is driven mainly by rental prices and value appreciation, in which rent equates to:


Rent = Allocated Power (KW) x Price/KW


This means that rent is often a lock-up price for every contract (often with a CPI-linked increase per year), in which it only really changes if there is a clause in a contract either side could not keep up with (e.g. failure to keep up the uptime). 


The main driver of investor returns comes from the valuation of the DC, which, similar to commercial real estate, uses the real estate capitalization rate and net operating income. DCs often aim to compress cap rates by securing a power moat, signing long-term, creditworthy corporate tenants, maintaining a longer WALT, achieving high power density per rack, and benefiting from cheap debt.


Improving PUE lowers operating overhead and directly increases NOI, which in turn drives up valuation. Additional NOI and valuation drivers include rising occupancy, lease renewals, growing interconnection revenue, and, again, tenant creditworthiness.


Law, Legal and Regulatory


Similar to investing in any other real estate or infrastructure project in Vietnam, DCs do not have a specified foreign ownership cap (though if an investor expands into telecommunications, there is a cap of 49% for facilities-based services or 65% for non-facilities-based services). That said, in support of getting fast approval for the project, land use rights, and navigating bureaucracy, it is suggested that foreign capital hold 50% or less of a DC project, joint-venturing with a domestic player, as this allows the project to be classified as a "deemed domestic investor."


As a deemed domestic investor, the JV has full access to VND-denominated domestic loans while still being able to access foreign loans, on the condition that medium- and long-term loans must be registered with the State Bank of Vietnam (SBV). This allows the investor to arbitrage the two loan markets to secure the ideal rate.


Investing in a Vietnam DC, however, must be registered with the Vietnam Telecommunications Authority (VNTA), and must adhere to the Personal Data Protection Law, the Cybersecurity Law, and the requirement to disclose user information to the Ministry of Public Security upon request. If structured correctly, the JV is eligible for a preferential corporate income tax rate of 10% for up to 15 years.


Vietnam DC Market


Vietnam's DC market scene has been established for a long time, yet is dominated by only a few players (Viettel, VNPT, CMC Telecom, FPT Telecom, and VNG), most of which run their own telecommunications services, which does not offer a carrier-neutral DC, as previously discussed.



As seen in the graph above, the current players in the Vietnamese market are dominated by State-Owned Enterprises (SOEs); for others, like FPT or VNG, they have been long-term, major domestic players in the telecom space with a strong relationship with EVN. This gives them more reliable access to power, and thus higher uptime. Viettel, for example, currently operates the largest DC (30MW) in Vietnam, and is currently building Vietnam’s largest DC at 140 MW, by leveraging their grid relationship, they are able to have strong opportunities like this and not fearing power shortages. 


According to Mr. Louis Nguyen, Saigon Asset Management's (SAM) CEO, the key obstacles deterring hyperscaler clients from entering Vietnam have been licensing, land, and power. By partnering with the Vietnam-Singapore Industrial Park (VSIP), a joint venture between Sembcorp and Becamex IDC,SAM is on its way to building a US$1.5 billion DC campus, gaining access to existing substation infrastructure and an existing EVN connection rather than having to negotiate interconnection from the start. 


Overall, DC investment in Vietnam should mainly come through developed industrial zones, such as the various VSIP parks, to build on existing infrastructure, or through DC campuses similar to the one SAM is constructing, in order to mitigate risk.


Challenges


As with any emerging market, the challenges of investing in and building infrastructure do not come easily; it is therefore suggested that foreign capital partner with domestic players in a joint venture to mitigate risk by leveraging their domestic knowledge.


One key obstacle to DC development in Vietnam is permit acquisition and navigating the bureaucracy stemming from land-use rights compliance, electrical safety, and environmental protection in which these areas require local expertise for faster approvals and know-how, so that approvals for the project or its transmission lines don't waste significant time.


After clearing approvals, finding a reliable energy source becomes the next challenge. Sourcing from EVN does carry a risk of blackouts and power shortages. N+1 or 2N redundancy does not fully solve this problem as its only main function is to solve brief local outages and not region or nationwide shortages like one in June 2023 which persisted for weeks. For this, mitigation strategy from diversifying power sources such as DPPA-backed renewable is suggested.


Vietnam's energy strategy as a whole has historically depended heavily on coal and fossil fuels, constituting over half of the national energy mix. Power shortages do occasionally happen when there are unplanned outages at old coal plants and low hydropower generation with the most severe being June 2023. After this incident, the Ministry of Industry and Trade concluded that EVN has been slow in its investment in power sources and grid infrastructure. After this, although the government have tried addressing it by completing the 500 KV Circuit-3 transmission line, which double North-South transmission capacity, and also expanding DPPA availability, in 2026, EVN still continued to ward the rising supply pressure from AI and digital transformation, meaning this still bear a great risk looking forward.


Natural disasters, especially in the central region of Vietnam, are another major risk; however, since major demand for DCs is concentrated toward the North and the South, this remains a consideration but not as critical.


Another risk with investing into Vietnam DC comes from the cooling aspect with the hot and humid environment of Vietnam. Cooling towers are not as effective due to the climate, and the country also faces a water-stress risk particularly in the Mekong Delta with saltwater intrusion and seasonal drought, which could potentially create social dissatisfaction amongst the locals as it would compete directly with local agriculture water sources.


For investors targeting Vietnamese firms and Vietnamese clients, this is not a major risk. However, international hyperscalers and firms entering Vietnam also bear the risk of data sovereignty and data governance requirements. As of January 1, 2026, Vietnam’s Personal Data Protection Law was in effect, yet the law is still untested and new.  If the regulatory framework changes as enforcement matures, this may affect operations hence why some hyperscalers have not yet fully committed to the market.


Conclusion


Vietnam's DC market is investable, and it's becoming an increasingly attractive opportunity for foreign capital to tap into and capture the upside. With a rapidly growing digital economy, a young and cost-competitive workforce, and a strategic geographic location, Vietnam's DC market deserves serious exploration now more than ever. Risk remains present, and partnering with a local company under a JV is therefore suggested to navigate complex regulatory requirements and mitigate localized risk. With the new legal framework now in place, regulatory risk is becoming more transparent; however, we do suggest that international investors entering Vietnam's DC market also look at diversifying their energy sources through DPPAs with renewable generators, rather than relying fully on EVN. 



Comments

Popular Posts