For the Global South, it is not only about utilizing more technology; rather, it involves designing technologies that use less energy and are climate-aware and relevant to local contexts. In this article, a case is made for developing a green, inexpensive, and distributed model of AI that serves public interest without making energy insecurities worse.

The new development tool that everyone talks about is artificial intelligence.AI can help farmers predict rain, help clinicians in overcrowded clinics, increase disaster relief and speed up governmental processes. Artificial intelligence seems to be a surefire way for the Global South to leap forward quickly, since development is urgently needed here and there is a growing ambition in terms of digital technology. 

AI and Energy Dilemma

However, there is one more factor in all of this. Artificial intelligence requires electricity, which is not widely available, equally accessible, and cheap in most developing nations. This means that the question today is not whether the Global South should embrace artificial intelligence. It is how to do so without worsening the energy situation.

The problem is compounded since AI technologies are not low-tech systems. Large models require strong computers, and the computers run from data centers, which require high amounts of power. Data centers require continuous cooling, adding yet another strain on power grids. 

With the growth of AI technology around the world, the demand for energy from data centers will continue to increase rapidly, and is predicted to do so even further by 2030. This problem is particularly relevant in countries where the power grid infrastructure is underdeveloped, as this is no longer a future problem but rather a present one.

This is the energy dilemma that faces the Global South. Countries could increase their emissions and weaken their climate promises by pushing AI development using fossil fuels. They would miss out on technological progress by waiting until perfect green energy sources are in place before developing AI capacities. Both of these are poor solutions. The former leads to dirty development. The latter leads to slow development.

However, this is not where the story ends. All AI technologies are not created equal when it comes to their energy needs.Big frontier models consume more electricity than small and targeted systems. Research has shown that training large models consumes an enormous amount of electricity and also generates a large amount of carbon, while smaller models have the potential to greatly reduce energy consumption. 

Designing Region-Specific AI

The reason for highlighting this point is that not all of the most relevant uses of AI in the developing world need large-scale systems. An agricultural advice system, language translation tool, or triage tool could be effective in smaller models.

This is why the future needs to be built based on green AI rather than just more AI. Green AI is not merely rhetoric. It is a real practice in which one can question the usefulness, efficiency, and feasibility of a system within a country where such a system is intended to operate. 

In the case of the Global South, this would mean designing AI according to the demands of that particular place and within the bounds of its available power. It would mean opting for small language models, frugality, and less energy-intensive infrastructure over costly mimicry of Silicon Valley.

This fact needs to be considered when developing policy as well. Governments need to begin by investing in AI applications that will consume less energy but solve existing problems in the society. The universities and new companies need to focus on creating local language applications, agriculture predictions, and service-based systems that will not involve much energy consumption. On the other hand, all large AI projects need to pass the energy and carbon assessment to determine how their use will affect the grid.

What follows is linking digital and clean energy planning. It makes no sense to create data centers as independent units that use a lot of power. Data centers should have links to solar or wind or any other form of renewable power where this is feasible. Cooperation between regions will decrease the price and risks involved. Countries can cooperate to purchase good-quality hardware, exchange their know-how, and set up common standards for digital facilities.

Required: Low-Carbon Digital Infra

In the long run, discussions on artificial intelligence policies, energy policies, and climate policies need to go hand in hand. What is required is a financial mechanism that facilitates the creation of low-carbon digital infrastructure and rewards energy-efficient innovations. Equally important is the development of local capability for maintaining and adapting such innovations so that they don’t remain permanently dependent on costly imports. This is what constitutes digital sovereignty.

The lesson is plain. AI success in the Global South should be measured not by model size or server count. The real test is whether AI makes lives better without creating new problems. A digital tool to save water in agriculture is only a success if it does not overload the power grid. A health app is only as good as its reliability in places where electricity is still shaky. In other words, development should not come at the expense of energy security or climate responsibility.

The path to AI self-creation for the Global South can only lie in building green, decentralized, low-energy, high-resource systems that are efficient and resilient. It must resist the allure of mimicking the world's giants. 

We must seek to meet our own national needs and constraints using these systems; we will have then produced not just another promise of artificial intelligence, but artificial intelligence development.

(The author is an undergraduate student, Department of International Relations, Jahangirnagar University, Bangladesh. The views expressed are personal. She can be contacted at sajutimahata2@gmail.com )