Advisory
Sovereign AI strategy: evaluating twenty-plus models before committing capital
20+
AI models and data assets evaluated
Challenge
A sovereign investor was standing up a national AI company and had to decide, quickly and defensibly, which products, models and data assets to build it around. The candidate portfolio ran to more than twenty AI products, large language models and datasets, developed by different entities at different levels of maturity, with no common basis for comparing them. Capital in the hundreds of millions depended on the selection, and a board would need to see how it was made.
Approach
I led the strategic assessment. We built a structured evaluation framework covering technical performance, data assets, commercial potential and strategic fit, and applied it consistently across the whole portfolio, including benchmark testing of the language models on Arabic-language tasks. The assessment fed an end-to-end AI strategy for the new company. I then co-authored the carve-out business plan with the fund’s legal and valuation teams: operating model, capital requirements, phasing and governance.
Outcome
The assessment identified the flagship Arabic language model, a 70-billion-parameter model that scored 74.8 per cent on average across the benchmark suite, and ranked the wider portfolio for investment. The business plan secured board approval for multi-hundred-million-dollar capital deployment. HUMAIN was launched publicly by the Public Investment Fund in 2025.
Source
Engagement record. Public context: the Public Investment Fund’s announcement of HUMAIN, 2025.
What decision are you weighing?
For advisory, training and speaking enquiries, or a conversation about a decision you are weighing.