Saudi Arabia wants to power 7% of global AI - but can it earn the world's trust?

The National

09-07-2025

When Saudi Arabia announced its new state-backed artificial intelligence company, Humain, it did so with a level of ambition consistent with its broader economic development plans. Backed by the $940 billion Public Investment Fund, and with $10 billion in venture capital to invest in AI startups, the message was clear: the kingdom wants to become a global AI leader.
While the scale of ambition is large, what's also essential is building trust, technical capacity and credibility over time. The real question is not whether Saudi Arabia has the money; it manifestly does. The question is whether capital alone is enough to establish global leadership in AI, or whether less-fungible factors – talent, data governance, geopolitical alignment – will determine who leads and who follows.
On paper, the scale is substantial. Saudi Arabia plans to build 6.6 gigawatts of AI computing capacity by 2034, a major commitment that would mark it as a serious regional player. It has begun construction on data centres as part of its Vision 2030 initiative to diversify the economy from oil, signed agreements with US technology companies including AMD and Amazon Web Services, and is investing $2 billion with Qualcomm to open a chipset design centre in Riyadh.
The kingdom also plans to open a 50-megawatt AI data centre by 2026, equipped with 18,000 Nvidia chips — the specialised hardware used to train large AI models and run them in practice, a process known as 'inference'.
But building infrastructure is not the same as building capability. Frontier AI models – such as GPT-4, Claude and Gemini – are typically developed in research labs by highly specialised teams. Saudi Arabia is not aiming to compete directly at this frontier, and current efforts appear focused elsewhere.
The kingdom is developing Arabic-language models, such as SaudiBERT and ALLaM, based on existing open-source frameworks such as Meta's Llama — publicly available model architecture that can be freely used, modified and adapted. This approach reflects an effort to address domestic and regional use cases rather than to compete globally at model level. And this makes a lot of sense.
That strategy is also pragmatic given the structural realities on the ground. Talent acquisition remains a known constraint for the kingdom, as well as for any other country. In a global market where top AI researchers command large incentives, including reported $100 million sign-on bonuses paid by Meta, building competitive teams will take time, co-ordination and sustained investment.
One less-discussed element of Saudi Arabia's AI strategy involves positioning itself as a data infrastructure partner. The kingdom has proposed a "Global AI Hub" that would allow for 'data embassies' – data centres on Saudi territory operating under foreign legal jurisdiction.
This model is intended to reassure foreign governments and companies that their data, while physically stored in Saudi Arabia, would not be subject to domestic law. The project reflects concerns around data sovereignty, now central to the geopolitics of AI and cloud services.
The underlying proposition is that countries across Africa and Asia, which are undergoing digital transition, will require data infrastructure partners. With relatively low-cost energy and its location between Europe, Asia and Africa, Saudi Arabia is positioning itself as a potential alternative to existing infrastructure hubs.
Saudi Arabia's data governance frameworks are still evolving. A national data policy was introduced several years ago, and the push to have more domestic managers under Vision 2030 may be contributing to a growing awareness of global digital norms within government and industry.
In parallel, many Saudis have studied abroad through initiatives such as the King Abdullah Scholarship Programme, launched in 2005. Some may be returning with relevant technical and enterprise experience, which could gradually support the growth of domestic AI capabilities.
US technology companies have shown commercial interest – agreements are in place, the construction of data centres is under way and hardware allocations have been announced. US policymakers, however, remain more cautious. Export controls on advanced AI chips continue to apply to Saudi Arabia, requiring licensing. These restrictions reflect US concerns about potential technology transfer, particularly to China.
Ultimately, reputation may prove a more complex barrier than infrastructure. Research excellence — the kind that draws international recognition and helps attract top talent — usually develops over time. It is shaped by strong academic networks, long-term funding and institutional support. Similarly, trust in data stewardship is not achieved through investment alone.
In recent years, Saudi Arabia has moved to open parts of its economy, attract a wider range of investment and put elements of its Vision 2030 strategy into motion. How these developments shape the country's position in the global AI space remains to be seen. The Gulf region as a whole has shown strong interest in AI development, including the UAE. But long-term leadership may depend as much on governing, transparency and academic freedom as it does on capital allocation.
If successful, Saudi Arabia could account for up to 7 per cent of global AI training and inference by 2030, in line with Humain's stated objective. It would have built a significant infrastructure base, expanded its digital footprint, and gained relevance in regional data flows.
If these efforts fall short, the reasons may include insufficient talent development, regulatory hurdles, or limited international trust in its digital governing model.
What is clear is that Saudi Arabia is now a participant in the global build-out of AI infrastructure. Data, like labour or capital, has become a factor of production. Infrastructure is emerging not in steel and cement, but in silicon and sensors.
The road ahead remains open. The path forward will hinge not merely on what Saudi Arabia builds, but on how trusted (and widely adopted) those systems become.