AI isn’t just upending businesses, but it is also poised to structurally change how entire nations operate. Compared to companies, countries are larger, more complex, and have lower manoeuvrability for those governing them, and navigating the AI revolution will be trickier – and fraught with greater consequences – for their leaders.
The 5-layered cake
Given how successive technological revolutions are faster than the other – the printing press required 300 years to be mainstream, while smartphones took a decade – countries don’t have long to adapt. In order to both protect themselves from potential downsides, and partake in the gains, leaders can divide their efforts along the 5-layer “AI cake” that NVIDIA CEO Jensen Huang has described. Broadly, countries need to insert themselves within as many of these five layers as possible, based on their demographics, relative strengths and access to natural resources.
1. Energy
Energy is the lowest-hanging fruit, and will become perhaps the most underappreciated strategic asset of the AI age. AI data centers are extraordinarily power-hungry, and countries with surplus cheap electricity — geothermal in Kenya and Ethiopia, hydro in the Congo, and solar across the Middle East — could have a structural cost advantage that Silicon Valley simply cannot replicate. Even electricity sources that are no longer popular in the west – such as coal or nuclear – will do, and could provide a differentiated edge. Electricity is the biggest input to AI, and having excess supply will stand countries in good stead.
Energy generation is also something that doesn’t require much technical expertise – just quick execution and political will to set up suitable plants. The tiny Himalayan nation of Bhutan is an example. It sells excess hydropower to India, and over the last decade ended up minting 13,000 bitcoin – worth nearly 20% of its GDP – using its excess electricity. Other countries can follow this example, and look to become energy-surplus to prepare for the AI age. Another aspect that could be worth investing in is transmission capacity – it too could end up being crucial.
2. Chips
Semiconductor chips are considerably more complex. Only a handful of countries produce AI chips, and they’d want to double down on their efforts. Countries that have the requisite abilities but aren’t producing chips would want to move down the value chain and see where they can fit – India, for instance, is setting up semiconductor assembly and testing plants that could be the precursor to full-blown advanced chip production. Malaysia has had success in this area, and other countries with similar developmental profiles can look to emulate its example.
Outside of chips, there could be niche areas that countries could hunt down and develop expertise in. A Japanese company named Nittobo makes nearly all the thin glass that is used in chip manufacturing. While it won’t be easy, countries could look to replicate such companies – or acquire them – to have a seat at the AI decision-making table.
3. Infrastructure and datacenters
Datacenters will be the factories of the AI boom. They also present a unique opportunity – there is considerable opposition building up to build datacenters in the US which could lead to many projects being cancelled, and other countries can look to capitalize. Datacenters only require land and access to electricity – other component parts can be imported – and could help all kinds of countries participate in the AI race. And after datacenters in the Middle East were attacked in the Iran war, countries can pitch their geopolitical stability to woo investors.
Opportunities also abound to take strategic bets on creating components for datacenters. Datacenters require coolants (Vertiv Holdings (VRT), fibre optic cables (Corning) and HVAC systems, and countries could set up companies manufacturing these components.
4. Models
Models aren’t trivial to create anymore. One can’t simply throw money and build frontier models – as Meta and xAI have now realized, building world-class models requires top researchers, compute for them to experiment with, and plenty of luck. This also appears to be a layer that is getting commoditized – the frontier labs are converging in capabilities, and open-source isn’t too far behind. It could be prudent for most countries to let this layer be, and focus their energies elsewhere.
But there could still be opportunities in the space. Some countries like Kenya and Nigeria have been contributing to providing human feedback to models, and they can explore similar business models for the short term.
5. Applications (SaaS, Robotics, Healthcare)
This is perhaps the most level playing field at the moment. AI will likely cause massive disruption in the software and applications space, and the spoils could go to countries and companies that can make the most of this disruption. Countries need to incentivize their workforces to build products and services atop the AI stack. Fresh products or services started now with the advantages of network effects or other moats could end up becoming enduring businesses for decades.
The Military Caveat
The neat analysis described above doesn’t hold true for countries which harbour ambitions to be global superpowers, or are under imminent military threats. AI is already being used in warfare, and will likely be the determining factor in military success in the coming years. AI abilities are already correlated with military might – there’s significant overlap between the top arms exporters and the top countries with AI models.
As such, countries will be left with two choices. They can either join the American or Chinese poles, and hope that their side provides them with the required AI technology when they’re at war, or develop their own. If a country wants to stay relatively independent over the next few decades, it would need to have control over the entirety of the AI stack. India and the EU could be potential candidates to make this move.
These are big changes for countries to implement, but in AI, nations are faced with possibly the biggest change in centuries. And successful countries in the past have rallied together in times of change – US’s GNP doubled from $99.7 billion in 1940 to nearly $212 billion in 1945 during World War 2, its number of aircraft carriers rose 15x, and its production of planes rose 30x. AI needs countries to come up with a similar response – and those who react quickly can put their nations in good stead for decades to come.