Low- and middle-income countries (LMICs) stand at a crossroads with AI: it could become a mechanism for asserting our sovereignty and delivering inclusive prosperity, or the latest tool of colonization and exploitation. For centuries, our labor, natural resources, and knowledge systems have been used to fuel progress in the Global North. The development and deployment of AI could follow this pattern, leaving LMICs without a stake in the technologies that will underpin our collective prosperity and ability to thrive.
But LMICs have an opportunity to avert this outcome. We already have the talent, resources, and vision to ensure that AI meets our needs. With increased coordination, investment in distributed computing, and grassroots innovation, we can establish a fairer technological order that creates value for Global South communities, strengthens their agency, and solves the most pressing challenges facing the planet.
Like colonial economies, the tech industry is fundamentally extractive. Modern AI systems, whether built by OpenAI or Meta, rely on data labeled by people in LMICs. But the Global North retains control of the industry and its profits, relegating Global South populations to the role of passive participants with high usage rates, rather than as innovators or equal actors.
Moreover, an estimated 2.6 billion people – one-third of the global population – are digitally unconnected and thus unaccounted for in the training of large language models (like the one powering ChatGPT). This reinforces Western-centric worldviews that dismiss our challenges, erase our histories, and stifle our potential.
Unless we take urgent action, the Global North will continue to dominate AI and develop new forms of economic and cultural dependency, widening the chasm between those who shape the future and those who are shaped by it. To move beyond colonial patterns of dependence, we must harness the potential of our youthful, digitally native populations, rather than chasing the largest models or building supercomputers. In this sense, the absence of legacy infrastructure, often framed as a barrier to innovation, is actually our biggest strength. Unburdened by outdated systems and rigid processes, we can build lean, purpose-driven data architectures that align with our needs and principles of data sovereignty.
Converting scarcity into innovation begins with education. That could mean introducing mandatory coding classes at schools or creating AI literacy programs to build a digitally fluent workforce. In Vietnam, for example, children are taught to code starting in third grade.
Developing local foundation models and open-source deep-tech tools, such as Amini’s Earth Foundation Model or Lelapa AI’s Vulavula, will be equally important. This requires maximizing existing resources and harnessing the benefits of Global North models while moving beyond them – similar to how DeepSeek challenged OpenAI’s dominance by focusing on efficiency instead of large-scale computing.
Embracing global technologies does not mean accepting them without question – we must also be prepared to create our own innovation ecosystem. Government programs, tax policies, and other measures are critical for supporting bottom-up initiatives in LMICs. Singapore has established one of the world’s most successful startup ecosystems partly through targeted investment and economic incentives.
Even as we chart our own path, our communities and technologists should engage in efforts to shape global AI ethics, localization, and governance, which requires establishing meaningful, equitable, and collaborative partnerships. And we should not hesitate to forswear initiatives that undermine our freedom to develop and use locally relevant technology. For example, the “green AI” and “frugal AI” movements imply that our developers can’t be trusted to make sustainable choices. But the high adoption rate of localized technologies like EDGE AI (formerly tinyML) in African countries shows that locally applied AI can create outsize, sustainable value at home and globally.
LMICs must act collectively to determine the best way to develop shared AI infrastructure and pool resources. To create a robust, secure, and inclusive ecosystem, some countries can host data centers, while others can build distributed computer nodes and processing centers. Resources must likewise be shared among the countries producing, transforming, and buying data, backstopped by targeted government initiatives that call for using local, rather than global, computational power and resources. Such a collaborative framework requires open dialogue, knowledge-sharing, and mutual support.
The question was never whether we in the Global South can “catch up” with the Global North’s AI dominance; it is whether we will use the technology as an equalizer to build a better world. LMICs have already started down this path, which suggests that the real AI revolution will unfold in Accra, São Paulo, Nairobi, and Jakarta, not in Silicon Valley. This is how it should be, because a data-rich, community-driven, and inclusive AI ecosystem in the Global South benefits everyone.
