The future of artificial intelligence in Saudi Arabia will not be decided inside data centers. It will be decided inside math classrooms.
Over the past while, I have been going through the numbers on the Kingdom’s AI ecosystem, and the more complete the picture became, the more I kept arriving at a single conclusion: the challenge ahead of us is not a shortage of capital, or infrastructure, or energy. In all of those areas, the Kingdom holds very strong cards.
The real challenge is human — specifically: how do we build, starting from school, the person capable of leading this field?
This is not a call for pessimism. Quite the opposite. Because the problem, in my view, is clear and manageable — if the investment is directed to the right place.
First: what has been achieved in recent years genuinely matters
We have more than 38,000 graduates in AI-related fields between 2019 and 2023. 86% of Saudi universities offer AI-related bachelor’s programs, 56% offer master’s programs, and 9% offer doctoral programs connected to the field. The number of graduates in these disciplines rose sharply over the same period.
These are meaningful numbers, and they reflect a real expansion of the Saudi talent base in computer science, computer engineering, AI, and related fields.
It is a major leap compared to where things stood ten years ago.
But the more important question is:
Has the base expanded as fast as the summit?
That is where the real problem begins.
Second: what do the PISA results tell us?
In the 2022 PISA assessment, Saudi fifteen-year-olds scored as follows:
- Mathematics: 389, against an OECD average of 472.
- Reading: 383, against 476.
- Science: 390, against 485.
But I believe the averages alone do not tell the most important part of the story.
More telling is the distribution of students across proficiency levels.
Only about 30% of Saudi students reached Level 2 proficiency in mathematics or higher, compared with 69% on average across OECD countries.
As for Levels 5 and 6 — the highest performance levels in mathematics — only a very small number of Saudi students reached them, against an OECD average of 9%.
Why does this number matter to me?
Because building a researcher capable of working at the advanced frontier of AI does not begin in the PhD.
It begins much earlier.
Mathematics is not everything in AI, but it forms an essential part of the foundations on which much of advanced machine learning research rests: linear algebra, probability, statistics, optimization, differential equations, and more.
More importantly, differences in mathematical ability do not appear suddenly at university entry. Many of them start taking shape years before that.
So when we later ask:
Why aren’t we producing more Saudi AI researchers?
the answer may not lie in the number of PhD programs alone.
We may need to step back.
Back to school.
That said, there is an important positive sign. The Kingdom improved its mathematics score compared to 2018 — a notable improvement — and OECD data points to a narrowing gap between the highest- and lowest-performing students in mathematics over 2018–2022.
So the direction is not wrong.
But the level still needs serious work, and what we fix today will not show its full results for many years.
There is a time lag that no administrative decision can compress.
Third: the second problem is research depth
There is a big difference between having a large number of AI graduates and having a large number of people capable of pushing the field’s frontier forward.
The skills of advanced research are not acquired from coursework alone; a large part of them is built through daily work alongside an experienced researcher: reading papers, reproducing results, forming hypotheses, failing, trying again, and learning from all of it.
And here another bottleneck appears:
the number of research-active researchers and supervisors who are genuinely working at the edge of knowledge.
You can open dozens of academic programs, but you cannot produce outstanding PhD supervisors at the same speed.
The result is that we may see growing numbers of graduates who can use AI techniques, while the number of people who can develop new techniques remains limited.
Both are needed.
But one does not substitute for the other.
Three factors make the matter even more complicated.
1. The speed at which the field changes
AI changes in months, while accrediting and updating academic programs can take years.
Some content can become outdated before it even finds its place in the curriculum.
2. The center of knowledge is no longer only inside the university
A large share of real learning today happens in published papers, on arXiv, on GitHub, and in open research communities.
A student who does not know how to enter this ecosystem will find themselves outside an important part of the knowledge — even with a good degree in hand.
3. Academic incentives
If a researcher’s evaluation depends heavily on paper counts and teaching hours, it becomes natural for researchers to drift toward safer, easier-to-publish projects.
But building a strong research ecosystem also requires room for high-risk projects — the kind that may fail many times before producing something important.
Fourth: what can we do?
In my view, there is a set of practical steps worth discussing.
1. Go back to the root of the problem: math and science
The first — and perhaps the most important in the long run — is returning to the root of the problem: mathematics and science at school, especially from fourth grade through high school.
What is needed is not just more math periods, but raising the level of the teacher and their ability to teach the subject in depth.
This is not a policy that delivers quick results.
We could start today and not see its real effect on researchers for more than ten years.
But that is precisely the kind of investment that is hard to market in the media — even though it may be among the highest-return investments of all.
2. Identify gifted students early
We need a national pathway for identifying students who excel in math and science at an early age, then giving them a genuinely different environment: specialized training, mentors, competitions, research projects, and early access to universities and labs.
Instead of waiting for the outstanding student to reach university, why not start working with them at 13 or 14?
3. Open up compute to students and researchers
We can leverage an advantage the Kingdom holds today: compute.
Providing GPU hours to researchers and students in exchange for a simple research proposal could cost far less than building new facilities — and at the same time enable thousands of experiments and projects that could turn into real research.
4. Concentration over dispersion
I believe we need a greater degree of focus.
We do not need dozens of strong AI programs.
It may be better to have fewer centers — but each with outstanding researchers, PhD students, compute capacity, stable funding, and real research freedom.
Depth sometimes matters more than spread.
5. Fix the “return contract”
It is not enough to send a student abroad for a PhD and then expect them to come back only to drown in teaching and administrative procedures.
If we want the researcher to return, they must find, upon returning, a lab, funding, compute, students, a team, and protected time for research away from administrative burdens.
In other words:
we don’t just need a “scholarship contract” — we need a “return contract.”
6. Bring the university closer to industry
The relationship between universities and industry could be far stronger.
Joint master’s and PhD programs with Aramco, SABIC, the telecom sector, and the banking sector, for example, could connect the student to a real problem and real data — so that the thesis does not end at the boundary of an academic document, but becomes a system or a model that can actually be deployed and tested.
7. Change what we measure
The number of programs and graduates matters, but it does not necessarily tell us about research strength.
We also need to look at indicators such as:
- the number of high-impact publications;
- the participation of Kingdom-based researchers as lead authors at major conferences;
- the number of PhD students graduated under active researchers;
- the share of returning researchers who stay in the Kingdom for many years.
Because what we do not measure well, we struggle to manage well.
And what can an individual do?
This part is perhaps the closest to me, because many of the solutions above require institutional decisions.
At the level of individuals, though, there is something we can start building today:
mentorship density.
We need communities that are small but serious: paper-reading groups, projects that reproduce published results, open-source contributions, and ongoing conversations between researchers and students.
This may look very modest next to the billions of riyals being invested in AI.
But it is, in truth, one of the things money cannot easily buy.
In the end, the existence of just one team inside the Kingdom that publishes consistently at major conferences, builds open tools, and works on real problems could have a larger effect than we expect on an entire generation of students.
Because when a nineteen-year-old sees a researcher close to them doing exactly that, their answer to a simple but very important question may change:
“Could I do that too?”
Conclusion
In my view, the gap is not in the university alone.
A large part of it begins between the ages of ten and sixteen.
University education numbers can look reassuring because they measure those who made it to university — but they do not always tell us about those who never reached the pathway we need in the first place.
The good news is that this is not an intractable problem.
The Kingdom has the money, the energy, the infrastructure, the universities, and the ability to attract top talent.
What we need is to direct a larger share of that effort to the place where the problem actually begins.
AI is not merely a matter of compute. In the end, it is a matter of minds.
Sources
- OECD, PISA 2022 Results – Saudi Arabia: Country Note
- OECD, Education GPS – Saudi Arabia: Student Performance, PISA 2022
- SDAIA / State of AI in Saudi Arabia — education, graduate, and AI-program data.
- Saudi Press Agency — report on the Kingdom in data and AI technologies.