Oil wars are bad news for equal AI access

Of all the many fallouts and risks arising from the ongoing strife in West Asia, there is one sneaky problem that no one is thinking about: By sending the oil prices into a tizzy, the Iran war could reshape our relationship with AI and the contours of that industry.

For months, the key talking point surrounding AI has been its potential to replace human workers. A recent study by Anthropic found that there has been “no systematic increase in unemployment” caused by AI since late 2022. But the study also noted that the hiring of younger workers has slowed in occupations exposed to competition from AI.

But the ubiquity of AI transcends the workplace. People of all ages and skills currently use AI chatbots as their own personal interns and assistants. They ask them how to respond to a tricky email from their bosses, how to write code in a new language, and even how to fix their messy sleep cycles.

At the personal and social level, AI’s most promising positive use might be in helping people reach a level of skill and expertise that they may otherwise have struggled with, particularly in new domains of work. It can help you learn how to build a robust Excel model, even if you haven’t done it before. It can help you create a new smartphone app based on a whimsical idea, even if you know little to nothing about programming.

These powers prompted the Australian economist, Justin Wolfers, to recently argue that the way gains from AI are shared will be decided by who owns AI. If you own access to AI as an individual, it can turn you into a thinker, doer, and entrepreneur in a completely new field. If your boss owns access to AI, she can probably use it to do large parts of your job.

Shared and equitable access to AI is a rapidly evolving game – and much of it will perhaps be determined by global energy markets, which are now in disarray. More than most recent tech advances, AI is an energy guzzler on an industrial scale. According to the International Energy Agency (IEA), electricity consumption from AI data centres has grown at around 12% a year over the last five years. From 2024 to 2030, the IEA estimates that consumption could grow more than four times faster than the consumption of all other sectors combined.

These challenges would especially hurt smaller countries. In Ireland, the energy demand was so debilitating that the government had to impose a moratorium on data centres in Dublin between 2021 and 2025.

Energy costs may also dictate the kind of AI you get access to. Not all AI models are created equal, and more sophisticated models – which perform better and make fewer mistakes – demand more energy to function. That also makes them costlier to access, and in result, some of the best, most sophisticated AI models may automatically fly out of the reach of individuals, smaller businesses or poorer economies.

In a similar sense, prohibitive energy costs may also price out newer players in the AI space. The apparent abundance of AI today is driven, at least in part, by early subsidies offered to frontier labs. But as costs rise, most AI model development may become concentrated in the hands of a few operators. The costs of training AI models are already high, given that frontier labs are still in the process of increasing their sophistication: Some believe that by 2027, a single training run could cost billions of dollars.

These are challenges that few have taken up just yet in global governance. Most global powers remain focused on improving national competitiveness by chasing the development of the best AI systems – rather than attempting to foster market competition in the sector, or improving access to those systems among individuals or small businesses.

But even the effort to build those systems – or the infrastructure to sustain them – remains concentrated in a handful of countries, particularly China, the United States, and Western Europe. Most developing economies and lower-income regions are yet to enter the conversation in a meaningful way.

The best way to fend off the risk of job displacement caused by AI is to ensure equitable access to sophisticated AI systems. That would not just improve the odds of re-skilling; it would also increase market competition, and with it, innovation.

The writer is a student of all things global and, self-confessedly, master of none, notwithstanding his Columbia Master’s, a stint with the UN and with monarchs in the Middle East.

Disclaimer: The views expressed above are the author's own. They do not necessarily reflect the views of DH.