A white paper — “Rethinking AI Sovereignty” — released by World Economic Forum (WEF) in Davos a few days ago in collaboration with Bain & Co, says that India’s accumulated investments since 2010 accounted for 1.2-1.8 per cent of its 2024 GDP, compared to the US at 3.4-5.1 per cent, Singapore at 3.1-4.6 per cent, South Korea at 2.2-3.3 per cent, and China at 1.7-2.6 per cent.
Countries or regions which performed lower than India include Europe (excluding the UK), and Brazil. Countries above India include the UAE, Japan, Canada, and the UK. While the rest of the world collectively was nearly on a par with India, a bulk of the investments was in hardware, driven by just two companies — TSMC and UMC.
But the white paper clearly brings to the fore the fact that AI is a big bucks game that mega investments, even though there is no road map of adequate return on investments in the near future. Two countries, the US and China, are of course making big over-sized bets — they account for 65 per cent of the total AI investment since 2010, totalling a staggering $2,150 billion to $3,250 billion. Both countries have taken a full-stack approach. The report says that the projected additional investment annually till 2030 would be to the tune of $1.5 trillion.
One of the largest investments is going into building AI infrastructure (data centres) where the cumulative investment since 2010 has already hit $600 billion. The report estimates that the globe currently has over 1,136 hyperscalers, and the projection is that that number would go up to over 2,000 by 2030. India itself has investments of over $67 billion, which will be on board in the country in the next three years. These investments were announced by global big tech firms.
In hardware, investment of over $200 billion has already been made since 2010. This investment is expected to grow by 15-25 per cent annually, or $90 billion per annum, until 2030 as more high-end processors are required to power AI. Investment in foundational models, the report says, is projected to grow 25-35 per cent per annum, reaching at least $300 billion per year until 2030, driven by large language models, small language models, and classical machine learning.