Data centres have recently emerged as a widely-discussed infrastructure theme in Nepal’s political and business circles. Announcements covering global corporations such as Google and Meta looking to invest were all over the news. Regardless of how much of that will materialise, the enthusiasm is understandable. Countries are racing to position themselves in the digital economy, with data centres at the heart of this transformation.

While this sector deserves serious attention, it also demands a clear understanding of its complexities. Data centres are not a one-size-fits-all opportunity. Depending on the use case, technical and capital requirements can vary substantially. In particular, AI-scale data centres are sophisticated assets whose success depends on a country’s underlying strengths in power, infrastructure quality, capital and governance frameworks as well as regulatory stability.

Treating these different segments as one risks oversimplifying both the opportunity and the challenge.

Power: Availability, scale and reliability

Power today has become the biggest bottleneck to AI capacity expansion globally. While supply is expected to catch up over the next 3–5 years, investors are currently prioritising immediate power access. Another important dimension is availability: These facilities require an ultra-high uptime of more than 99.99 percent, equivalent to just 52 minutes of downtime per year, with minimal voltage and frequency disturbances.

While Nepal’s grid has recently achieved commendable improvements, it unfortunately cannot meet the stringent technical requirements of an AI facility today. Nepal’s continued seasonal dependence on Indian imports, weak transmission and distribution infrastructure and its inability to support multi-hundred-MW concentrated loads (now considered typical for AI training) add further risks to the project profile.

Capital

AI data centres are among the most capital-intensive assets globally, with an all-inclusive cost of $30–50 million/MW. At this rate, a 100 MW data centre facility would cost $3–5 billion. Nepal’s GDP in 2024, per the World Bank, was $42 billion. Even a single facility could approach a material share of Nepal’s GDP. This begs the question of who will fund these investments.

Such projects also depend on a broader ecosystem: insurance, contracting frameworks, legal and financial systems, and regulatory certainty. Without a track record at this scale, investors will approach the sector cautiously. Higher risk will make financing even more expensive.

Local and regulatory dynamics

Water has been a major contentious community issue for data centres. Depending on usage and design, a single kilowatt-hour of power consumption could demand over one litre of water for cooling. Now imagine having a few such facilities in the water-constrained Kathmandu Valley—they could easily consume a significant share of the city’s water demand.

Site selection is another crucial topic, especially in densely populated urban areas. Locating an AI infrastructure in/near cities like Kathmandu requires clarity on land use prioritisation, safety buffers and long-term urban planning. Communities may be less amenable to facilities that operate 24/7, produce loud noise and vibrations from large cooling systems or generators, or consume sizable local power and network infrastructure capacity. In the absence of clear zoning guidelines, siting decisions could lead to long-term conflicts with locals.

Specialised labour is another constraint. Securing this labour will either mean building local capacity through extensive training, a time-consuming process or importing, which will put pressure on cost. AI facilities also tend to generate relatively limited direct employment compared to their capital intensity, raising questions around their impact and opportunity cost for a developing economy like Nepal.

The broader regulatory landscape in Nepal will also need to evolve. Current environmental and infrastructure approval processes do not explicitly account for the unique characteristics of data centres—including their energy intensity, water use and heat and electronic waste generation. Given the resource-intensive nature of this infrastructure’s operations, appropriate regulations will be needed beyond the initial approvals to cover its entire lifecycle.

The question is not whether Nepal should participate in the sector, but whether there are segments, other than AI, that are a better fit. A nuanced understanding of this sector is therefore critical.

What should Nepal focus on?

While AI-scale data centres present significant challenges, the opportunity within the broader data centre ecosystem remains compelling. Below are some suggestions:

Crypto mining or hosting represents a near-term opportunity. These facilities are less sensitive to uptime constraints, can be deployed at a sub-MW scale and are significantly less capital-intensive (roughly $1 million/MW)—all these qualities fit Nepal’s situation better. Moreover, Nepal’s cheap hydropower at Rs4.80/kWh ($0.035/kWh), albeit only in the wet season, is competitive globally. Expanding winter supply through domestic solar, battery and storage-based hydro could unlock further low-cost power. Note that while Nepal Rastra Bank does not recognise cryptocurrency currently, investment in crypto data centres may still offer possibilities for global hosting and mining services.

Enterprise and domestic data centres represent another important avenue. As Nepal’s digital economy continues to expand—with rising internet use, fintech adoption and digital services—demand for local data storage and processing will grow. Developing domestic capacity also has strategic importance for data sovereignty and national security.

A related opportunity lies within the AI sector, not in large-scale training but in inference (AI applications). As the industry evolves, a growing share of computing demand is expected to shift toward inference and localised applications. Nepal’s emerging technology ecosystem could participate meaningfully in this segment. Even if Nepal cannot achieve global AI leadership, any AI-based applications deployed locally, including those by corporations, will benefit from having local specialised capacity through lower latency and reduced reliance on cross-border infrastructure.

Nepal’s long-term strategic advantage

One area where Nepal holds a clear structural advantage is clean energy. With a hydropower-dominated power system, complemented by high-quality solar in the dry season, the country can offer low-carbon electricity at globally competitive prices.

While sustainability considerations have taken a back seat in the current global AI race, this is unlikely to remain the case indefinitely. As regulatory pressures increase and climate commitments tighten, the carbon footprint of digital infrastructure will once again become central to investment decisions. With the right investments in storage and system balancing, this could position Nepal as an attractive destination for environmentally conscious data centre operators in the long term.

The global momentum behind data centres is undeniable, and Nepal is right to explore its role in this transformation. Nepal possesses real advantages—clean energy, competitive power costs, a sizable population and a growing digital economy. However, it also faces a multitude of structural constraints, plus an important policy question of how to balance Nepal’s scarce resources between domestic consumption and industrial development.