The Great AI Divide: As 1.2 Billion Adopt, 4 Billion Are Left Behind

The Great AI Divide: As 1.2 Billion Adopt, 4 Billion Are Left Behind  

Source Force | Tech News | Published: December 1, 2025

The generative artificial intelligence revolution is advancing at unprecedented speed, yet Microsoft’s AI Diffusion Report (October 2025) warns that this progress is cementing one of the most profound divides of the digital age.

In less than three years since ChatGPT’s debut, 1.2 billion people worldwide have used AI tools, making AI the fastest-spreading general-purpose technology in history outpacing the internet, smartphones, and personal computers. Yet, 4 billion people remain excluded , lacking the foundational infrastructure required to participate.

A World Split in Two: The Adoption Chasm

• Global North vs. South: AI usage reaches 23% of the working-age population in the Global North, compared to just 13% in the Global South.
• High-Adoption Leaders: The UAE (59.4%) and Singapore (58.6%) showcase decades of strategic digital investment.
• Low-Adoption Regions: Sub-Saharan Africa and parts of Asia struggle to reach 10%, hindered by systemic infrastructure gaps.

Microsoft CEO Satya Nadella acknowledges the challenge: “While the Global North leads in AI adoption, we are committed to bridging the digital divide and accelerating AI equity worldwide.”

 The Five Pillars of Access

Microsoft identifies five non-negotiable foundations for AI participation:

• Electricity
• Data Centre Proximity
• High-Speed Internet Connectivity
• Digital Skills
• Language Support

Failure in any pillar blocks meaningful access. Today, 750 million people lack electricity , while 4 billion fall short across these core areas . Infrastructure concentration worsens the divide: the U.S. and China command 86% of global data centre capacity , directly impacting latency, cost, and usability.

The Language Barrier: AI’s Unseen Frontier

Language emerges as a critical frontier:

• English dominates ~50% of web content, despite being native to only 5% of the global population.
• Countries with low-resource languages show adoption rates 20% lower than peers with high-resource languages.
• Swahili (200M speakers) has 500x less digital content than German .
• Model accuracy drops from 80% in English to below 55% in Yoruba , spoken by 50M people.

This linguistic imbalance ensures that even with connectivity, AI remains less effective for billions.

Concentrated Development and the National Race

AI development is even more concentrated: only seven countries host models ranked in the global top 200 ,  the U.S., China, France, South Korea, the U.K., Canada, and Israel.

• U.S. Leadership: OpenAI’s GPT models remain frontier leaders.
• China’s Challenge: DeepSeek V3.1 trails by less than six months.
• Israel’s Push: AI21 Labs’ Jamba Large 1.7 is just 11.6 months behind the frontier.

Microsoft VP Michal Braverman-Blumenstyk calls for national strategy:  “We are already a cyber powerhouse and we can and must also be an AI powerhouse.”

Analysts compare this race to South Korea’s semiconductor boom in the 1970s, which fueled 6.2% annual growth, versus the Philippines’ 1.8% without similar intervention.

The Path Forward: Equity or Entrenchment?

The report is both a celebration and a warning. AI adoption is historic, but structural inequality is deepening . Without deliberate intervention across the five pillars, the world risks entrenching a permanent AI underclass.

As Nadella emphasizes, AI has moved  “from hype to becoming a core part of how every organisation operates, innovates and delivers value.” The next phase will be defined not only by breakthroughs in algorithms but by global commitment to equitable infrastructure