Iran's AI Roadmap 2026–2032: Implementation Strategy, Sanctions Reality, and Policy Priorities

Policy Context: Iran's AI Roadmap May 2025

In May 2025, the Islamic Republic of Iran formally adopted its National AI Roadmap, establishing the first comprehensive framework for AI development through 2032. This represents a strategic pivot: recognition that AI is a critical technology for national competitiveness and that government coordination is essential given sanctions constraints.

The roadmap commits to three primary targets:

1. Industry AI Integration: 45% of Iranian industries to integrate AI applications by 2032 (from ~5% today)
2. Economic Contribution: AI to contribute 12% of GDP by 2032 (from under 1% today)
3. Global Ranking: Iran to rank among the top 10 AI-harnessing nations globally

The National AI Organization, established in July 2024 under direct presidential oversight, serves as the coordinating body. The organization reports directly to the Office of the President and has mandate over research coordination, infrastructure deployment, workforce development, and regulatory framework development.

This represents a departure from previous technology initiatives in Iran. AI is being treated as a national security and economic priority, similar to how China approaches AI policy. Government commitment is real, funding is being allocated, and infrastructure is being built.

Financial Overview: $215M Commitment and Budget Mechanisms

The Iranian government has committed $215 million USD-equivalent for AI development and research through 2032. The funding breaks down across several channels:

Budget Allocation Mechanisms

• National Development Fund grants: Approximately $90 million for applied AI research, data center infrastructure, and university partnerships. These are competitive grants open to public research institutes, universities, and state-owned enterprises. Application periods: Q1 2026, Q3 2026, Q1 2027.
• Research Council Direct Funding: Approximately $65 million for fundamental AI research at Sharif University, University of Tehran, Amirkabir University, and Institute for Research in Fundamental Sciences (IPM). Funds support faculty recruitment, lab equipment, and doctoral scholarships.
• Infrastructure and Hardware: Approximately $45 million for domestic GPU procurement, data center expansion, and cloud computing infrastructure to circumvent sanctions-imposed restrictions on international cloud access (AWS, Google Cloud, Azure).
• Skills Development and Training: Approximately $15 million for teacher training programs, curriculum development, and online courses administered through the National AI Organization and Ministry of Education.

Fiscal Sustainability

Funding is sourced from several streams: (1) National Development Fund surpluses, (2) budget reallocations from other technology initiatives, and (3) expected revenue from AI-driven sectors (telecommunications, energy, e-commerce) generating 3–5% surcharges dedicated to AI reinvestment. The government's commitment is strong through 2026–2028; sustainability beyond 2028 depends on oil price dynamics and whether AI-driven sectors generate sufficient revenue to self-fund expansion.

Fiscal Risk: If global oil prices fall below $50/barrel (a geopolitical risk given current global supply), government budget pressures may force reductions in AI funding in 2029–2030. Current budget planning assumes oil averaging $60–70/barrel.

Implementation Timeline: 2026–2032 Milestones

Phase 1: Foundation (2026–2027)

• Establish domestic GPU data center operations in Tehran and Shiraz (2–3 petaflops capacity operational by Q3 2026)
• Launch first cohort of National AI Organization training programs (500 developers, data scientists per year)
• Publish open-source Farsi language models (Rakhsh AI v2.0) for public use
• Integrate AI curriculum into 50+ secondary schools and 15+ universities
• Award $50 million in competitive research grants to university consortia

Phase 2: Scaling (2027–2029)

• Expand domestic GPU capacity to 5+ petaflops across four regional data centers
• Deploy AI systems in government services: tax administration, healthcare, education (reaching 15+ government agencies)
• Increase industrial AI integration from 15% to 35% of Iranian companies
• Establish regional AI hubs in Mashhad and Isfahan (reducing Tehran concentration)
• Launch first domestically-trained AI PhD cohorts graduating with research publications
• Achieve Farsi NLP parity or superiority to English-language models in domain-specific applications

Phase 3: Consolidation (2029–2032)

• Achieve 45% industrial AI integration (target from roadmap)
• Scale domestic GPU capacity to 10+ petaflops, reducing dependence on external compute
• Export AI services and expertise to neighboring markets (Central Asia, Iraq, Afghanistan)
• AI contribution to GDP reaches 12% target
• Iran ranked globally as top 10 AI-harnessing nation

Execution Risk: These timelines are ambitious relative to Iran's current starting point. Achievement depends on sustained government commitment (vulnerable to political cycles), continuous funding (vulnerable to oil price shocks), and retention of trained talent (vulnerable to emigration). Realistic probability of achieving the full roadmap: 40–50% by 2032.

Workforce Policy: Retention, Training, and Brain Drain

The most significant constraint to implementing Iran's AI roadmap is not money or infrastructure—it is human capital. Between 150,000 and 180,000 skilled Iranians emigrate annually. For AI specialists, the emigration rate is higher: estimated 25–40% of trained ML engineers leave Iran within 5 years of career start. This creates a leaky bucket: government trains workers who emigrate to Canada, Australia, Germany, or US.

Retention Mechanisms Being Deployed

Premium Salaries: Government AI positions now offer salaries of $2,000–3,500/month, comparable to or exceeding private sector. This is 4–8x the national average and creates competition for talent.

Restricted Emigration for Grant Recipients: Researchers and engineers receiving government AI funding are contractually required to work in Iran for 3–5 years post-completion of training. Breach carries financial penalties. Enforcement is variable (estimate: 60–70% effective compliance).

Diaspora Engagement: The government is recruiting Iranian AI researchers living abroad to return temporarily (6–12 months) for specific projects, offering premium consulting rates ($5,000–10,000/month). This creates incentives for overseas Iranians to contribute without requiring permanent return.

Equity and Mission-Driven Framing: Government agencies and state-owned enterprises are offering equity stakes (small, ~0.5–2%) to early employees in AI divisions, framing work as "nation building" rather than just employment.

Projected Outcome

By 2027, estimate: 5,000–6,000 trained AI workers (from ~2,000 today), but emigration of 1,000–1,500 annually will offset gains. Net AI workforce growth: ~2,500–3,500 by 2027 (vs. roadmap target of 10,000+). This suggests the roadmap's workforce assumptions are optimistic by 40–50%.

Policy Implication

Workforce policy should shift from retention (which has limited effectiveness) to rapid upskilling of broader population. The goal should be: train 15,000 people with AI fundamentals (not just 5,000 specialists), accept that 30–40% will emigrate, and rely on the remaining 9,000–10,500 for domestic implementation. This is more pragmatic than expecting to retain 90% of trained workers.

Education Reform: University Partnerships and Curriculum Development

The roadmap emphasizes education as the foundation for workforce development. Policy initiatives include:

University Partnerships

Tier 1 Institutions (Sharif, Tehran, Amirkabir): Government is funding expansion of AI graduate programs, with target of 500 Masters graduates per year and 50–80 PhD graduates per year by 2028. Funding covers scholarships, faculty hiring, lab equipment. Sharif University homepage shows expanded graduate AI program accepting 120 students/year (up from 40 in 2024).

Secondary School Curriculum: Ministry of Education is integrating AI fundamentals into high school STEM curriculum. Target: 50+ secondary schools teaching AI by 2027. Curriculum focuses on practical applications (computer vision, NLP, recommendation systems) relevant to Iranian economy (agriculture, energy, e-commerce).

Online Education Platform

The National AI Organization is developing a national online learning platform (similar to Coursera but domestically hosted) offering free or subsidized AI courses in Farsi. Target: 10,000 registrations per quarter by 2027. Quality: variable, but serves workers unable to access university programs.

Professional Certification

Government is establishing a national AI certification program (similar to Microsoft or Google certs) recognizing competency in machine learning, data engineering, and AI ethics. This creates a credential pathway for self-taught engineers and bootcamp graduates. Estimated 2,000–3,000 certifications awarded per year by 2028.

Policy Assessment

Education reform is well-designed but faces implementation bottlenecks: (1) insufficient supply of qualified AI teachers (catch-22), (2) curriculum lag (textbooks, materials take 12–18 months to develop), (3) variable quality across secondary schools and universities. Impact by 2027: likely to produce 2,000–3,000 additional AI-trained workers per year (vs. roadmap assumption of 5,000+).

Sanctions Framework: Legal and Technical Constraints

US sanctions on Iran create hard constraints on AI development that policy cannot easily overcome. Understanding these constraints is essential for realistic planning.

Semiconductor and GPU Restrictions

US sanctions prohibit direct export of advanced semiconductors and GPUs to Iran. High-end NVIDIA GPUs (H100, A100) cannot be legally purchased. Iran has developed workarounds: (1) purchasing GPUs through intermediary countries (UAE, Turkey), (2) acquiring second-hand or refurbished hardware, (3) investing in domestically-designed compute chips. The $45 million infrastructure allocation includes ~$20 million for GPU procurement through informal channels.

Technical Reality: As of mid-2026, Iran has accumulated approximately 500–800 high-end GPUs through informal procurement. This is sufficient for training small-to-medium language models and running inference at scale, but insufficient for state-of-the-art research competing with US/EU institutions. Domestic data centers are 5–10 years behind Silicon Valley in terms of compute density and efficiency.

Software and Tool Restrictions

Iran cannot legally access NVIDIA's full CUDA toolkit, MATLAB (Mathworks), or certain cloud infrastructure (AWS, Google Cloud, Azure) through official channels. Government has responded by investing in open-source alternatives (PyTorch, JAX, TensorFlow) and building domestic tooling. This is workable but creates friction: researchers cannot use proprietary tools standard in global academia, creating a divergence from international best practices.

International Collaboration and Publication

Iranian researchers can publish in peer-reviewed journals and collaborate internationally (sanctions do not prohibit scientific exchange), but collaboration with sanctioned entities is restricted. This is a significant constraint for AI research: many top conferences and collaborative projects involve participants from sanctioned countries or companies. Iranian researchers navigate this by publishing in non-restricted venues and maintaining informal collaborations through non-sanctioned channels.

Policy Implication

The roadmap's AI targets are achievable within sanctions constraints, but only through focused investment in domains where Iran has data (e-commerce, energy, agriculture) and where foreign GPU access is not essential. Competing with global AI labs in frontier research (large language models, cutting-edge vision systems) will remain structurally difficult through 2032 without sanctions relief.

Infrastructure Investment: Domestic GPU and Data Centers

The government is investing heavily in domestic compute infrastructure to reduce dependency on international cloud services. Key initiatives:

Data Center Expansion

Tehran Data Center: Operational since Q3 2025 with ~2 petaflops capacity. Houses 200+ high-end GPUs, primarily used for government research and Farsi language model training.

Shiraz Data Center: Opening Q2 2026 with ~1.5 petaflops capacity. Location chosen for proximity to Persian Gulf (cooling advantages) and distance from Tehran (redundancy).

Mashhad Regional Hub: Opening Q4 2026 with ~1 petaflops capacity. Designed to serve regional AI research and industry (northeastern Iran, Central Asia exports).

Total Planned Capacity by 2028: 5–6 petaflops across four regional centers. This is comparable to a mid-tier global cloud provider (e.g., Scaleway, Digital Ocean) but far below NVIDIA's or hyperscale cloud capacity.

Pricing and Access Model

Government is offering GPU compute at cost (~$0.30–0.40/GPU-hour) to universities and state-owned enterprises, and at commercial rates (~$0.50–0.70/GPU-hour) to private companies. This subsidizes research while generating revenue for operations. Demand will likely exceed supply through 2028, making allocation a policy question (priority to government sectors, then academia, then private).

Fiscal Impact

Infrastructure capex is front-loaded: 60–70% of the $45M hardware budget is spent in 2026–2027. By 2028–2029, costs shift from capex to opex (electricity, maintenance, staffing). Electricity costs are significant: estimated $5–10M annually for full operation of all data centers. Iran's electricity prices (heavily subsidized by government) make this feasible, but energy sustainability is a political question in an era of power grid stress.

Policy Recommendations for 2026–2030

Recommendation 1: Revise Workforce Targets Downward

The roadmap's assumption of 10,000+ AI workers by 2032 is unrealistic given emigration. Government should adopt a two-track strategy: (A) deep expertise: train 3,000–4,000 specialists to elite competency, (B) broad literacy: train 10,000–15,000 workers with AI fundamentals. This maximizes impact with realistic retention rates.

Recommendation 2: Prioritize Domestic Data Advantages

Iran cannot compete with global AI labs in frontier research. Instead, government should focus AI investment on domains where Iran has unmatched data: consumer behavior (Digikala, Snapp), energy infrastructure (power grid, oil fields), agriculture (wheat, pistachios, saffron production). Developing world-class AI in these domains creates defensible competitive advantages and exportable expertise.

Recommendation 3: Establish Innovation Clusters Outside Tehran

AI research is concentrated in Tehran (Sharif, Tehran University, government labs). Policy should incentivize satellite hubs in Isfahan, Tabriz, and Mashhad through tax incentives and dedicated funding. This reduces brain drain (regional talent stays closer to home) and develops secondary cities.

Recommendation 4: Formalize Diaspora Engagement

Rather than restricting emigration, government should establish formal programs bringing diaspora AI researchers back for consulting, teaching, and collaborative projects. Estimated 500–1,000 Iranian AI researchers abroad would return for 6–12 month projects at premium rates. This generates intellectual capital transfer without requiring permanent repatriation.

Recommendation 5: Establish AI Ethics and Governance Framework

As AI deployment accelerates in government services, establish clear governance on bias, transparency, and accountability. This should include civilian oversight, independent auditing of government AI systems, and public transparency on AI use in sensitive domains (security, law enforcement, social services).