Does Kenya's AI policy account for the physical infrastructure AI actually requires?
A civil, structural and environmental engineer's reading of the draft Kenya AI & Emerging Technologies Policy — and what it still leaves ungoverned.
Not yet. Kenya's draft AI policy treats AI mainly as a digital system. At scale, it is also a physical one — large compute needs electricity, water, land, cooling and environmental approval, the same disciplines already applied to roads and dams. Without an explicit readiness framework, investment risks landing on unprepared grids, water-stressed sites and unassessed land.
Why this matters now
Kenya's ambition to become a leading African AI hub is achievable, and the draft policy is a real governance foundation. The Kenya AI Strategy 2025–2030 already names data centres, cloud infrastructure and connectivity as enabling foundations, and national energy plans prioritise expanding renewable generation. What's missing is the bridge between those two facts: a framework that coordinates energy, water, land, environment and climate sectors before investment commitments and site decisions are locked in.
This isn't an argument against data centres. It's an argument for attracting the right investment to the right locations, under conditions that protect Kenya's energy security, water resources, host communities and long-term competitiveness — the same logic already applied to every major road, dam and transmission project in the country.
AI's physical footprint, dimension by dimension
| Dimension | Why it matters | What's missing today |
|---|---|---|
| Electricity & grid | AI compute creates large, continuous, sometimes variable loads | Site-specific transmission and distribution capacity isn't checked before commitments are made |
| Water & cooling | Cooling technology and local climate set direct water demand | No water-balance or drought-resilience requirement for water-stressed sites |
| Land & services | Facilities need secure sites, roads, drainage, security, logistics | Poor siting shifts infrastructure cost onto counties and communities |
| Environmental & social | Noise, emissions, hazardous materials, e-waste, occupational health | Assessment isn't yet proportionate to AI-scale operations |
| Cumulative impact | Clustered facilities strain shared grids and catchments | Projects assessed individually; cluster effects go unmeasured |
| Climate risk | Heat, drought and flooding affect reliability and demand | Climate screening isn't built into siting and design |
| Social licence | Public trust depends on transparency and fair benefit-sharing | No disclosure or benefit-sharing requirement yet exists |
Eight provisions the final policy should add
| # | Provision | What it requires |
|---|---|---|
| 1 | Recognise AI infrastructure as national infrastructure | Define large-scale compute, data centres and AI-support facilities as strategic infrastructure needing cross-sector planning |
| 2 | Create an AI Infrastructure Readiness Framework | Assess power, water, cooling, connectivity, land, climate risk and community impact before major approvals |
| 3 | Require early environmental & social screening | Move assessment upstream, before land acquisition, technology lock-in and power or water commitments |
| 4 | Introduce resource-use disclosure | Significant facilities report projected electricity, water, cooling, backup generation, emissions and e-waste |
| 5 | Address cumulative & regional impacts | Assess clusters of facilities, not only individual projects, on grids, catchments and host communities |
| 6 | Establish green, climate-resilient standards | Renewable baseload, efficiency, water-sensitive cooling, heat recovery, circularity, resilient design |
| 7 | Build Kenyan professional capacity | Multidisciplinary capability across power systems, civil and environmental engineering, EIA, and safeguards |
| 8 | Create transparent monitoring | Published indicators, thresholds, responsible institutions and periodic national readiness reports |
How it could be phased in
| Action | Lead | Timing | Outcome indicator |
|---|---|---|---|
| Interim screening guidance | ICT Ministry, NEMA, Energy, Water, Lands, counties, KEBS | 0–12 months | Guidance issued and applied to major proposals |
| National Readiness Framework | ICT Ministry with cross-sector taskforce | 0–18 months | Framework approved; indicators published |
| Disclosure & assessment thresholds | NEMA, EPRA, WRA, KEBS, counties | 0–18 months | Thresholds gazetted or regulated |
| National project register | ICT Ministry, NEMA, regulators | 12–24 months | Public register operational |
| Capacity programme | Universities, professional bodies, industry, government | 12–36 months | Trained public and private practitioners |
| Biennial readiness report | ICT Ministry and taskforce | Every 2 years | Report published with power, water, environment and investment indicators |
Suggested policy language
Two clauses that could be inserted directly into the digital infrastructure or responsible-deployment pillar:
The Government shall establish and periodically update a National AI Infrastructure Readiness Framework to guide the planning, siting, approval, development and monitoring of large-scale compute, data-centre, cloud and other AI-enabling infrastructure — integrating electricity and grid capacity, renewable energy, water availability, cooling technology, connectivity, land-use planning, climate risk, environmental and social assessment, cumulative impacts, circularity, public participation, local professional capacity and community benefit.
Within twelve months of adoption, the responsible Ministry shall convene relevant national and county institutions to issue interim screening guidance and define thresholds for strategic assessment, project-level EIA/ESIA, resource-use disclosure and public reporting for AI-enabling infrastructure.
Related questions
Does a data centre in Kenya need an Environmental Impact Assessment?
Yes. Under the Environmental Management and Coordination Act and NEMA's EIA regulations, qualifying projects require an EIA before commencement. Read the full approvals brief →
Who regulates AI infrastructure siting in Kenya?
No single body — the ICT Ministry, NEMA, county governments, EPRA and the Water Resources Authority each hold part of the mandate, with no framework coordinating them yet. Read the full regulatory-map brief →
How much water and power does an AI data centre actually use?
It depends on scale, cooling technology and local climate, so figures vary widely by design. What matters for policy is a disclosure requirement, not a single number. Read the full disclosure brief →
Does Kenya require climate-resilient data centre design?
Not by name yet, but the Climate Change Act and National Adaptation Plan already apply to infrastructure generally — the link to data centres just hasn't been made explicit. Read the full climate-design brief →