Interview: Tareq Amin, CEO, HUMAIN, on creating innovation sovereignty through architecture and intellectual property ownership
What role can hyperscale data centres play in anchoring artificial intelligence (AI) capabilities and advancing the digital economy?
TAREQ AMIN: Globally, the biggest constraint to AI infrastructure is energy – advanced models require enormous power, cooling, land and connectivity, and very few regions can offer all three at scale. The Kingdom uniquely fills this energy gap. With low-cost energy, available land, strong global connectivity and progressive regulation – including data residency and emerging data-embassy frameworks – the Kingdom can host AI compute not only to meet domestic demand, but also as an export opportunity. With the right package – competitive infrastructure pricing, robust data protection and secure cross-border connectivity – the Kingdom can become the world’s third major centre for AI compute, behind the US and China.
How can the Kingdom scale AI-ready data centre capacity to lead in model training?
AMIN: Reaching multi-gigawatt capacity means rejecting traditional construction. Instead of bespoke facilities, the build must be modular and prefabricated – more like industrial manufacturing than real estate. Core components should be produced off-site in standard 100-MW blocks and assembled on-site, enabling fast deployment and scaling in line with demand. Equally important is pairing this with a commercial strategy, that is, secure compute offtake agreements first, then build and price infrastructure to remove cost barriers. Modular build, combined with committed demand, is the most efficient path to gigawatt-scale capacity.
Which strategies can attract AI talent in chip design, graphics processing unit (GPU) computing and model development for innovation sovereignty?
AMIN: Deep technical talent exists in the region – especially among engineers trained abroad – but what is often missing is the opportunity to work on frontier challenges. People do not come only for compensation; they come to build things that matter: foundational models, new chip architectures and globally scaled systems. When the Kingdom positions itself as a place to design the next generation of GPUs, not just deploy them, talent follows. Strategic partnerships are essential – including agreements that bring global chip design teams into the Kingdom and pair them with programmes that mentor Saudi engineers. Innovation sovereignty is created through architecture and intellectual property ownership, not just manufacturing.
In what ways can developing Arabic-first, multi-modal AI models strengthen AI capabilities?
AMIN: Models reflect their training data. If culturally aligned data is absent, the model will not understand local language, nuance or values, thus creating dependency. Training on high-quality regional datasets that are sourced from trusted governmental and private sources enables the Kingdom to develop models that are fluent in Arabic dialects and aligned with regional norms and guardrails. Once that foundation exists, sector-specific applications – such as those in finance, health care, education and law – become possible.
Where should investment focus in next-generation GPU or accelerator AI factories to position the Kingdom as a regional hub for model development?
AMIN: A divide is emerging, in that there are developing economies that can afford compute and those that cannot. If compute remains expensive, AI will unfortunately benefit only a few. Investment should focus on inference, not just training. Training hardware will continue advancing, but inference is where economics breaks down. Leading in efficient, lower-cost inference accelerators would redefine global compute affordability. This means backing emerging accelerator companies early, co-developing architectures locally and providing large-scale global serving from regional data centres.
Read More from OBG
