{"text":[[{"start":8.28,"text":"The writer is co-founder and chief investment strategist at Absolute Strategy Research"}],[{"start":14.68,"text":"“The United States is in a race to achieve global dominance in artificial intelligence.” These are the opening words of the White House’s AI action plan that is based on accelerating innovation, building AI infrastructure and leading in international diplomacy and security."}],[{"start":33.67,"text":"For the Trump administration, winning in AI is clearly at the heart of the battle for the 21st century and keeping the US as one of the leading (and richest) global economies. It is key to offsetting the productivity drag from worsening demographics. US trend productivity has slowed from a rise of more than 2.5 per cent a year in the 1960s, to about 1.5 per cent a year in the past 20 years. And US demographics will probably put further downward pressure on productivity, without labour-enhancing technological change."}],[{"start":70.95,"text":"As Yale associate professor Michael Peters argues in an article for the IMF, America must rediscover its dynamism if it is to maintain its global standing. This is why an explicit part of the US AI plan is to invent and embrace productivity-enhancing AI uses that the world wants to emulate. My company’s own analysis suggests that AI has already raised US labour productivity by 0.1 to 0.9 percentage points and might eventually raise global productivity growth by about a half point annually over the coming decade."}],[{"start":null,"text":"
"}],[{"start":111.13,"text":"However, increasingly, the key limit to the US leading the AI arms race is less technological and more the physical limits that AI places on US energy demand."}],[{"start":124.06,"text":"The International Energy Agency estimates that global data centre electricity demand will more than double, from 460 terawatt hours of electricity in 2024 to more than 1,000TWh by 2030 and will reach 1,300TWh by 2035. In the US, it expects data centres will account “for nearly half of electricity demand growth between now and 2030”."}],[{"start":157.6,"text":"But it is not just the scale of the energy demand needed for US AI that is the issue. It is also the role hydrocarbons play in fuelling this growth. The IEA expects more than half the electricity powering US data centres to still come from fossil fuels, notably natural gas, until after 2030. But even by 2035, the IEA forecasts more than 40 per cent of US AI energy being hydrocarbon-based due, in part, to the Trump administration’s cancellations of support for renewable energy initiatives."}],[{"start":null,"text":""}],[{"start":196.39,"text":"While China also relies on hydrocarbons (mainly coal) to power its data centres, it is looking to change this mix. Its strategy is focused on developing computing resources close to coastal renewable power sources. The IEA estimates that, unlike the US, China will see the level, as well as the share, of data-centre electricity generation coming from hydrocarbons falling after 2030."}],[{"start":225.07,"text":"One strategic risk from this US AI reliance on hydrocarbons is relative costs. US average electricity prices have already risen 38 per cent since 2020 with these cost increases often linked to rising AI data centre demands. These cost pressures are only likely to rise. In contrast, the cost curves for renewables continue to come down faster than for fossil fuels, potentially disadvantaging US AI vs Chinese AI in the longer run."}],[{"start":262.35,"text":"However, the bigger strategic challenge for US hydrocarbon-powered AI may not only be price, but also its physical impact on the economy. Energy, water and food production are intimately linked. The increased US energy needs for AI, when powered by hydrocarbons, will lead to more US water demand and stress than if AI growth were powered by renewables."}],[{"start":287.35,"text":"To make matters worse, two-thirds of the new US data centres built, or planned, since 2022 have been located in places of elevated water stress, according to Bloomberg analysis. As a result, the risks to longer-term US food security from hydrocarbon-powered AI could be a major constraint."}],[{"start":308.43,"text":"As the US “bets the farm” on using hydrocarbons to power a rapid data centre build-out, China is relying on a slower process, using a more sustainable energy mix."}],[{"start":320.32,"text":"Thus, the US strategy could be costly, not only in terms of higher electricity prices, potentially limiting the return on investments in AI, but also increased water stress and potential food insecurity. In exchange for these increased strategic risks, society will probably demand real, tangible benefits — a modestly faster search engine will seem a poor trade-off. The bigger strategic risk is that while the US may win the initial AI battle, it might end up losing the war due to its reliance on hydrocarbon energy."}],[{"start":367.10999999999996,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1766709139_4137.mp3"}