Pakistan’s relatively weak mobile broadband penetration and gaps in electricity access are limiting the infrastructure needed for widespread artificial intelligence adoption, according to the World Bank.
The October 2026 Middle East, North Africa, Afghanistan and Pakistan Economic Update said Pakistan is simultaneously building a strong AI talent base, with around 75,000 IT graduates produced annually and record ICT services exports of $4.6 billion in FY2025-26.
Broadband and Electricity Remain Key Constraints
The World Bank said internet usage across the MENAAP region is often higher than income levels would suggest, but the bigger issue is the quality and depth of connectivity and reliable access to electricity.
Pakistan is among eight MENAAP economies where mobile broadband subscriptions are lower than expected relative to income levels. The others are Djibouti, Egypt, Iraq, Jordan, Oman, Syria and Yemen.
The report does not provide a standalone internet usage percentage for Pakistan and does not simply classify the country as having low internet use. Instead, it identifies mobile broadband penetration and electricity access as more important barriers to advanced digital and AI adoption.
Electricity access in Pakistan stands at 95.7% of the population.
| Economy | Electricity Access |
|---|---|
| Pakistan | 95.7% |
| Afghanistan | 87.8% |
| Syria | 89.3% |
| Libya | 77.4% |
| Djibouti | 69.2% |
Pakistan is one of only five economies in the comparison where electricity access remains below 99.5%. The other 15 economies covered by the comparison have universal access.
The World Bank said reliable electricity and connectivity increase the reach, continuity, and productive value of AI tools.
Where broadband penetration remains limited or electricity access is incomplete, firms and workers may struggle not only to adopt advanced AI applications but also to sustain the digital activity on which those systems depend.
Pakistan Is Still Building a Strong AI Talent Base
Despite these infrastructure gaps, the World Bank places Pakistan among MENAAP’s “talent and ecosystem builders.”
It said Pakistan is positioning itself as a talent-led AI services exporter, supported by its 75,000 annual IT graduates and $4.6 billion in ICT services exports in FY2025-26.
Pakistan is also pursuing what the report calls “AI sovereignty” through a $1 billion programme running through 2030.
The programme includes:
- Shared GPU infrastructure
- Development of a sovereign multilingual AI model
- 1,000 AI PhD scholarships
- AI training for one million non-IT professionals
The World Bank groups Pakistan with Egypt, Jordan, Morocco and Tunisia as middle-income economies that have growing technical talent and digital ecosystems but still face major AI-readiness gaps.
Pakistan Leads MENAAP in Claude Conversations
Pakistan is already a significant user of AI tools.
Data from Anthropic’s Claude platform showed that Pakistan recorded the highest absolute number of conversations among MENAAP economies.
However, usage was concentrated in arts and media activities, including graphic design, content creation and translation, rather than software development.
The World Bank said this may reflect Pakistan’s large presence on international freelance platforms, where these services are in demand.
It also noted that AI usage in most MENAAP economies remains concentrated in a relatively small number of occupations compared with the United States, where adoption is spread across a broader professional base.
Urdu Remains Underrepresented in AI Training Data
Pakistan also faces a major localisation problem because Urdu has limited representation in the data used to train AI systems.
Urdu accounted for just 0.03% of global URLs collected by Common Crawl in July 2026.
| Language | Share of Common Crawl URLs |
|---|---|
| English | 40.6% |
| Arabic | 0.7% |
| Persian | 0.7% |
| Urdu | 0.03% |
Urdu has somewhat better representation in open-source AI training datasets hosted on Hugging Face, where it accounts for 0.6% of datasets.
| Language | Share of Hugging Face Datasets |
|---|---|
| English | 64.4% |
| French | 11.3% |
| Chinese | 4.8% |
| Arabic | 2.0% |
| Persian | 0.9% |
| Urdu | 0.6% |
The World Bank said language representation matters because it affects the quality of AI outputs for people who rely on Urdu, Arabic or Persian in professional, public-service, and everyday use.
Three Gaps Could Limit AI Productivity Gains
The World Bank identified three broader structural gaps across MENAAP:
- Localisation
- Usage
- Foundational human and physical capital
The physical component includes reliable broadband and electricity, while the human-capital challenge includes digital skills, AI expertise and creative-thinking capabilities.
The Bank said these problems reinforce each other. Poor localisation can discourage adoption, limited use prevents firms and governments from reorganising work around AI, and weak skills or infrastructure restrict workers’ ability to use the technology productively.
‘Small AI’ Could Help Work Around Infrastructure Limits
The World Bank said countries do not need to wait for universal high-speed connectivity before adopting AI.
It identified “small AI” as a possible route for economies with connectivity and electricity constraints.
These are lower-cost, purpose-built systems that can operate using low-bandwidth connections, basic mobile phones, intermittent electricity, SMS, voice calls, lightweight apps, or offline functions.
Pakistan is already developing such solutions designed for local contexts and languages, according to the report.
Overall, the World Bank said Pakistan represents both sides of the region’s emerging AI divide: a large technology talent base and growing export sector, but also significant broadband, electricity, localisation and wider readiness gaps that could restrict how broadly AI spreads across the economy.
The post Pakistan’s AI Push Faces a Broadband and Power Problem appeared first on ProPakistani.
