DistantNews
Support us
🇳🇵 Nepal /Technology

Nepal's AI budget: The right instinct, and the trap inside it

From Kathmandu Post · () English

Translated from English, summarized and contextualized by DistantNews.

At a glance

Analysis Sources not specified Context piece
  • Nepal's government plans its first Sovereign AI Compute Centre, allocating funds for AI processors to support local startups and leverage the country's hydropower resources.
  • The initiative aims to treat compute power as a national asset, a move welcomed by some as a step towards developing AI capabilities.
  • However, concerns exist that the focus on training infrastructure might overlook the more practical and widespread use of AI through inference, which requires less intensive resources.

Nepal's Finance Minister Swarnim Waglé announced plans for the country's first Sovereign AI Compute Centre, a move that has generated cautious optimism. The budget allocates funds to purchase AI processor units, aiming to make their use affordable for local startups and harness Nepal's abundant hydropower for compute power.

AI is not a simple software product. It looks much closer to an electricity grid, and the one who does not own the foundation must rent it forever.

— Article AuthorThe author explains the critical nature of owning the foundational infrastructure for AI development.

The initiative is seen as a significant step by the government to treat raw compute power as a national asset. The plan includes inviting Nepali experts working abroad to return as fellows and establishing a new fund, Matribhumi Kosh, to invest in "strategic assets" like an "AI factory."

The world’s data centres need a cheap and clean electricity source, and we happen to have a mountain of it.

— Article AuthorThe author highlights Nepal's potential advantage in leveraging its hydropower resources for AI compute needs.

However, the article raises a critical distinction between training AI models and using them. While training requires substantial investment in specialized chips and electricity, the practical application of AI, known as inference, is far cheaper, faster, and less power-intensive. This inference process is what most users experience, from analyzing medical scans to translating languages.

Despite my reservations, this makes me happy. This is the first time the government has treated raw compute as a national asset.

— Article AuthorThe author expresses a positive reaction to the government's recognition of compute power as a national asset.

The author expresses concern that the budget's emphasis on "AI factory" and "thousands of GPUs" leans towards the training aspect, potentially missing the mark on the more accessible and immediate applications of AI. The article suggests that Nepal's focus might be misaligned with the practical realities of AI deployment, which largely relies on inference rather than extensive model training.

‘AI factory’, ‘thousands of GPUs’ is the training language. It is a vocabulary we find in countries in

— Article AuthorThe author points out the language used in the budget, suggesting a focus on AI training rather than practical application.
DistantNews Editorial

Originally published by Kathmandu Post in English. Translated, summarized, and contextualized by our editorial team with added local perspective. Read our editorial standards.

More Stories

From Our Blog