Race for AI dominance is creating competition and new global alliances

Nvidia boss Jensen Huang in Washington DC last month. The AI surge has created the need for massive investment into power generation and data centres. AFP

OpenAI’s recent $1 trillion spending spree on chips and computing power shows that artificial intelligence now depends as much on energy and infrastructure as on algorithms.

The ChatGPT maker has signed pricey long-term deals this year for greater computational ability – enough to draw the energy of roughly 20 nuclear reactors – as it scrambles to secure the power to keep its large language models running.

The scale of the spending signals a deeper shift: the race for AI dominance now depends less on code or talent and much more on who controls the physical systems that power it.

That realisation is spreading beyond boardrooms to governments as nations pour money into chips, data centres and energy. But which nations will come out ahead – and might this race redraw the map of global power as deeply as oil once did?

Generative AI has triggered an infrastructure boom of staggering proportions. Data centres – the beating heart of AI – consumed about 4.4 per cent of total US electricity in 2023, a figure projected to reach as much as 12 per cent by 2028. Between 2017 and 2023, electricity use by data centres more than doubled.

Ensuring responsible use of AI top of the agenda as Gitex kicks off

02:03

The boom shows no sign of slowing. Spending by the four largest US technology companies could top $320 billion this year. The splurge by Microsoft, Meta, Alphabet and Amazon has already soared 63 per cent to record highs last year. Much of the investment is flowing into data centres, chips and energy systems.

In parts of the US, electricity demand from data centres now exceeds available supply, forcing utilities to delay connections. The same pressures are visible elsewhere.

Even Saudi Arabia, one of the world’s largest oil producers, is accelerating investment in renewable and nuclear power to diversify its energy mix and free up hydrocarbons for export. The kingdom expects rising demand from digital infrastructure, heavy industry and a growing population to test its existing capacity.

The International Energy Agency projects that, between 2024 and 2030, electricity use by data centres will grow by about 15 per cent a year, roughly doubling total consumption.

Cooling is also a huge cost: many data centres consume hundreds of millions of gallons of water annually to control heat. In cooler climates, such as Norway, that burden can be reduced by lowering the need for artificial cooling.

Together, these costs are redrawing where it makes sense to build, and who stands to benefit. At present, a handful of countries dominate this new hierarchy.

The US remains strongly positioned, owing to its deep capital markets, ample energy and lead in advanced chip design – driven by Nvidia, the Silicon Valley company whose processors power many of the world’s leading AI models.

China, meanwhile, is pouring state money into building its own chip industry, adding more large data centres. Even with US export controls on advanced chips, China’s ability to fund and build at scale means it will remain a key player in the global AI race.

Elsewhere, in the Middle East, many countries have substantial oil and gas reserves. However, its position in the AI race will depend on people as much as power, which is why governments in places like the UAE are investing in developing a high-skilled technology workforce. But doing that at scale remains a formidable challenge.

Tech-power alliances

Europe’s position is fragmented: countries like France and Norway benefit from steady nuclear or hydro power, giving them a base to fuel the expansion of data centres. Others face high electricity costs, slow grid expansion and local opposition to big projects, making it harder to build at scale. In Ireland, for example, planning limits and public pushback have slowed or stopped several new data-centre projects around Dublin.

Many governments want more control over this infrastructure and are now talking about “sovereign AI” – the goal of building and running their own systems without relying on foreign technology. In practice, however, few can achieve full independence.

Most still depend on partners for chips, software and expertise. Just as oil – and the pipelines that move it – still shape global power, control over chips, data networks and electricity grids is starting to form the backbone of alliances in the AI age.

Two alliances stand out. First, the US and its allies are building supply chains and standards around Western chipmakers and cloud providers. Second, China is developing its own ecosystem to reduce dependence on foreign technology. But even as the world splits into rival technology blocs, both sides still depend on the same few critical chokepoints.

Taiwan’s dominance in advanced chipmaking is one of the biggest in the global technology system. Most of the world’s cutting-edge processors are made by Taiwan Semiconductor Manufacturing Company. All of its factories are based on the island. That concentration means any disruption could ripple through the entire global economy.

These pressures have, in part, set off the global race to expand capacity, with countries investing in chips, data centres and energy to secure their place in the AI supply chain.

But doing so will depend on access, not isolation. The nations best able to combine computing power, energy and cooling – and to build the partnerships needed to sustain them – will lead in the early years of the AI age.

Just as oil defined – and still shapes – the geopolitics of the last century, infrastructure may define the next. The race to power AI is not only a technological contest, but a new map of global power in the making.

Amit Joshi is professor of AI, analytics and marketing strategy at IMD