The Global Nature of AI – It Can Never Be local

Published: July 19, 2025. Source: Image by Gerd Altmann from Pixabay

The Artificial Intelligence (AI) revolution is a geopolitical race redefining alliances and power. If previously we used gold or the US dollar, now the proxy is AI (even though we are not sure how it works across borders). The pandemic helped pushed AI to the forefront and the unregulated launch of Chat GPT caught global imaginations. Policy makers are floundering trying to understand the impact of AI, the economic advantages versus the cost of investment and regulating AI that seems to be encroaching in every industry.

But what drives AI? First, money – Today’s AI systems are built on decades of research costing billions of dollars. Much of this early investment was government money, some of it in defense, some in the space race. Today, we see an increasing amount of investment from the private sector, but it comes with its own strings attached (the need for commercialization versus basic research across non-STEM fields, transparency, talent poaching).  This is an area of future vulnerability – if we do not invest now what happens to the future of research – say 45-60 years? USA, China (and lagging behind is EU, South Korea and Japan) are heavy investors in AI R&D – as observed by patents, but most of this is really on commercialization not always basic research.

Second, you cannot have AI without hardware. According to McKinsey – this requires a $7 trillion investment.  This hardware is what drives your internet (contrary to what people think, satellite still needs data cables for the last mile). It is expected that by 2028 5G will be 2/3 of all mobile subscribers who want videos, agentic AI and edge computing. This is largely the domain of the telecommunications industry. Most telecom providers are government or government affiliated and hence there are many trade barriers where foreign capital may not be easily accepted. Data cables are mostly private sector initiatives or consortiums with familiar players - Meta, Microsoft, China Mobile International, Amazon, Alphabet/Google, Mobily, Orange, Vodaphone, Airtel – so are cube sats like Elon Musk’s’ Starlink.

Hardware is in your phones or computers or data centers and often associated with the semiconductor industry (though not necessarily so). The biggest challenge in the semiconductor industry which is super specialized is the logistics of materials and components. To manage uncertainties, new factories or foundries are being set up in sovereign jurisdictions or using the strategy of friendshoring (to reduce dependency on China). China dominates the semiconductor supply chain and has a huge cost advantage having also managed to ensure their supply chain. However, during the pandemic, the supply chain was disrupted leading to a delay of 15-22 weeks, affecting 169 industries. Today, we have tariffs but that is another story. Geopolitical uncertainties are another reason for uncertainty – Ukraine supplies 50% of the neon used for fabrication.

Semiconductors are also used in hyperscalers (80% of data centers) and in 2024 $57 billion was invested in this sector. This AI semiconductor industry needs deep pockets so any country investing in the semiconductor industry will not find it profitable unless it reaches scale OR has big subsidies. Since more governments are jumping on the bandwagon – the latter scenario is needed. So, who has deep pockets – say perhaps those with oil resources or those countries printing money? Despite the need to diversify energy, oil seems to be doing well with all the global insecurity. Hence recently we saw some interesting deals in the UAE, KSA and Bahrain– all private-public partnerships.

AI in its current form needs massive energy infrastructure so this is another area of investment. Big bets are being made on nuclear, natural gas, hydrogen, biofuels, solar, but we need to manage this with the stress of climate change. The demand is massive – according to IEA Executive Director, “Global electricity demand from data centres is set to more than double over the next five years, consuming as much electricity by 2030 as the whole of Japan does today.” Again, this is a cost intensive sector and while the private sector is jumping in – look closely – government funding is a must.

Most of the large players in the AI sector (from USA, China, South Korea, few from Europe, India) have a tremendous amount of power – and this makes deals with them interesting. Take for example, India’s UPI payment launched in 2016 and available in Bhutan, France, Mauritius, Nepal, Singapore, Sri Lanka, and the UAE, which now handles more transactions than VISA per day. Or the fact the NVIDIA has a market share of 90% of data server GPUs. Or Microsoft 365 Copilot is now being used by  70% of Fortune 500 companies. Or Tencent’s WeChat that has 1.34 billion subscribers. But think about smart devices like autonomous cars and the way they connect to your mobile, or devices like Alexa, Siri or Google who listen as you speak, or the internet of things. Who has your data? Where in the world is it shared? And how is it being used? Lots of geopolitical uncertainty here for data flows.

AI is global if you look at data flows. Take a look at the 559 cables and 1636 landings that transfer 97% of the words data across the Earth. This is what keeps the global economy running. Just to give you an idea of its impact, s $10 trillion in daily financial transactions flows through these data cables. Look at the ~8000 Starlink satellites that circle the world. Then consider the fact that a single chip can travel over 25,000 miles and cross 70 borders before finding a home in your device. What is not yet truly global is research funding. As of 2020 the 10 largest R&D-funding countries (~$2 trillion in R&D expenditures), accounted for 85.0% of the global total. Hopefully this last bit changes.

AI is global – it cannot be local as the service and third party providers are based across the world.