Part I: Why We Should Pursue AGI in Saudi

Whenever Artificial General Intelligence (AGI) is discussed, a debate about its definition immediately follows. It is a fuzzy concept. You can define it through capability, a system that can do whatever humans do. You can define it economically, a system that can perform most economically valuable work. Or you can even define it philosophically, a conscious entity with superior intelligence. I’m not going to get into this debate, because no matter what you think AGI is, I am arguing we should pursue it here in Saudi. But before explaining why, I need to share how I see our local AI ecosystem today.

Let me start with LEAP 2023. It took place just a few weeks after I returned from studying AI in the US, and shortly after ChatGPT launched. It was my first look at the local scene, which then consisted of only a few AI players (fortunately, I joined one of them). Back then, AI was mostly confined to niche features inside enterprise software, alongside a few cool experiments. Fast-forward to LEAP 2026 just a few weeks ago, AI is the absolute mainstream. It was virtually impossible to find a single booth not pitching an AI solution. Compared to 2023, it felt as though an entire industry had emerged out of nowhere. In less than four years, we now have young, AI-native companies (like my friends at Sarj and Swat), mid-sized players (like my former employer, Mozn), and our local AI giant, HUMAIN.

Yet, as I walked around those booths in 2026, a specific pattern became hard to ignore. The vast majority of these AI offerings are simply variations of generic B2B enterprise workflow agents. In blunt terms, they look like Microsoft Copilot, only narrower and less capable. Their primary defendable moat relies heavily on local compute, data residency, or Saudi-specific deployments. These are temporary bottlenecks. With HUMAIN entering the market, global giants like Microsoft and Google bringing local data centers online soon, and other global players like Cohere, Brain Co, and Scale AI adopting forward deployment strategies, that geographical advantage will vanish quickly.

Just a few days after LEAP, OpenAI announced Astra. Greg Brockman described it as the beginning of the AGI era, complete with massive claims about leaps in intelligence and capability. Since then, other major labs have followed with their own frontier announcements, sparking intense debate on how fast the frontier is moving. Whether any of these systems should actually be called AGI is not the point here. What mattered to me was the contrast. Our local ecosystem is undoubtedly progressing, but the global frontier is moving at breakneck speed. It forced me to ask a critical question: where do we stand in Saudi Arabia compared to the frontier?

Starting at the bottom layer, infrastructure: local compute availability remains quite limited today, but this is actually the area of least concern. We are moving fast enough that compute will soon be a solved problem. Nearly every major infrastructure player has committed to Saudi expansion, while HUMAIN is building massive compute clusters aiming to make the Kingdom the world’s largest AI token exporter. HUMAIN has even stepped into the device tier with the Horizon laptop, a domain where Alat is also set to play a role. A full, sovereign AI infrastructure ecosystem goes well beyond compute nodes and laptops, and building those deeper hardware supply chains is insanely difficult. But the baseline infrastructure is already on a very fast track.

Moving one layer up—to data, intelligence, and foundational models—the gap becomes obvious. We still do not have a Saudi research lab producing frontier models. HUMAIN is the only major local player attempting to participate at that level. Releases like M3 and HUMAIN Voice are certainly meaningful steps, but if the benchmark is the global frontier, a very large gap remains. Then there is the application layer, where HUMAIN has announced a suite of products spanning developer gates, horizontal enterprise platforms, and vertical creative applications.

But HUMAIN is just a single player. We still need to ask: what does the Saudi AI ecosystem look like without them?

Remove HUMAIN from the equation, and the landscape looks very different. I am not aware of another meaningful local player tackling the algorithmic or foundational-model layer at a comparable scale. We also lack a deep local industry dedicated to proprietary data, model training, or core AI research. At the application layer, there are interesting bright spots like the infrastructure startups Deep.sa and CranL, and several companies building out verticals in customer service and healthcare. However, I have yet to see a Saudi horizontal AI product reach meaningful scale, and a consumer AI player simply does not exist here yet. In short, we still don't have an AI unicorn.

All of which brings me right back to LEAP: if there were so many AI companies on the floor, what are they all actually building?

A large portion of the market today is focused on custom enterprise AI solutions. We are seeing endless variations of internal workflow agents and wrappers built on top of existing models, heavily aimed at government and large corporate clients. Their true selling point is rarely a defendable technical moat; instead, they sell local deployment, strict data residency, and customization. There is nothing inherently wrong with that. These companies can solve real problems and generate real revenue, which is a very natural place for an ecosystem to start in a young market. The danger, however, is that our ecosystem has positioned itself to wait for the waves generated at the frontier to finally reach our shores so we can react to them. We are not making a serious effort to get out into the deep water and surf where the waves are actually being formed.

To be clear, I do not see our current position as a failure. Our rapid evolution is undeniably impressive. A few years ago, much of this ecosystem barely existed. But Saudi ambition should not be benchmarked against where we were five years ago; it should be benchmarked against the global frontier. As a football fan, I see a direct parallel. Today’s Saudi AI ecosystem is the Saudi Pro League before Cristiano Ronaldo. The league was not empty before Ronaldo arrived. There had already been years of investment, foreign players, and gradual improvement. But Ronaldo represented a fundamental shift in standards, raising the bar to demand top-tier status. Saudi AI still needs that Ronaldo moment.

To answer how pursuing AGI changes that ecosystem, we first need to look at why we landed in this position to begin with. I think the explanation is remarkably simple. Market players acted completely naturally and rationally, settling into a local equilibrium between risk and reward. Consider the custom enterprise AI companies I mentioned earlier. If you are building for large corporate clients or government entities, there is a massive amount of unmet demand. Creating value does not require inventing new technology. All it takes is fixing a broken workflow, integrating off-the-shelf models, satisfying local compliance rules, and closing the deal. Repeat that process, and you have a solid business. Because the market is young and competition remains limited, the bar for technical differentiation is still quite low.

Contrast that safe playbook with the reality of building something technically ambitious. You need elite, expensive talent. You burn through significantly more capital. You face painfully long development cycles before knowing if the product even functions, let alone scales. The probability of simply failing is far higher. If you survive that phase, you still have to figure out repeatability and unit economics. And the hardest truth? Even if you succeed, the local Saudi market is rarely large enough to justify the sheer cost of frontier innovation. To get the necessary reward for that level of risk, you have no choice but to compete on a global scale.

When the risk skyrockets but the local financial reward stays relatively flat, avoiding high-risk innovation becomes a rational calculation. This is a problem of incentives, not a lack of founder ambition. Once enough founders and investors act on those incentives, the market forms a self-reinforcing loop. Enterprise buyers continue purchasing custom workflows because they are accessible and low-risk. Founders build directly for those active budgets. Venture capitalists fund what is already generating reliable cash flow. Talent flows to where the jobs are, and the entire supporting infrastructure solidifies around this low-risk baseline.

This cycle can persist because our local market is still in its early stages. High unmet demand means companies can thrive without deep technical moats. Competition has not yet squeezed margins, buyers are still discovering what quality AI products look like, and generous local contracts sustain startups that would otherwise have to productize or go global to survive. Over time, this market dynamic will inevitably shift. As more players enter, generic wrappers become commoditized, and buyers get smarter, margins will compress. Companies will eventually have no choice but to engineer proprietary technology, build true products, and expand outside Saudi Arabia.

But relying on natural market maturity is a slow process, and the frontier is moving far too quickly for us to wait. Telling founders to magically become more ambitious or asking venture capitalists to accept worse economics will not fix it. As long as the current equilibrium offers comfortable rewards for minimal risk, people will reasonably choose that path. Instead, we must disrupt the equilibrium itself. We need an external pulling force that alters how the market naturally behaves, making the low-risk route insufficient and forcing companies to build genuine capability and differentiation just to maintain their standing.

I argue that raising the market bar to its highest visible form is precisely that kind of pulling force. The higher the bar, the less room there is for a low-risk equilibrium to survive underneath it. The market gets pulled upward because the threshold for what counts as good enough has shifted. Today, the highest visible bar in the AI industry is AGI. This is why I maintain that the exact definition of AGI is secondary to the argument. What matters is that AGI represents the pinnacle of the industry's ambition.

The key question then becomes: can setting the highest possible target at the top truly pull the entire ecosystem along with it?

History gives us a compelling template for this in the Apollo program. When the US committed to landing a man on the moon, it set a bar far beyond what 1960s engineering was naturally capable of achieving. To meet the extreme weight and power constraints of spaceflight, Apollo designers could not wait for commercial computing to organically mature. Driven by an extraordinarily ambitious goal, NASA bought up nearly 60% of all US integrated circuits in the early 1960s, enforcing unprecedented manufacturing standards. This massive demand forced early chip manufacturers like Fairchild to master quality control and mass production almost overnight. Whether or not astronauts ever reached the moon, that single ambitious target forever transformed the semiconductor industry and sparked the birth of Silicon Valley.

The parallel is clear, but let's look at how this economic ripple effect plays out locally. Assume HUMAIN seriously pursues AGI. The first obvious bottleneck is talent, so let's use that as a concrete example. Frontier AI talent operates in a global market with global options. To recruit them, HUMAIN must pay globally competitive compensation, a single move that resets the local talent market overnight. Top Saudi and regional engineers can suddenly work on frontier problems and earn top-tier salaries right at home. Mid-sized companies immediately face the threat of losing their best people. To retain them, they are forced to raise salaries, driving up baseline operating costs. To sustain higher payrolls, they can no longer rely on low-friction client projects; they must generate significantly more revenue. That forces a pivot toward building scalable products, developing genuine technical differentiation, or expanding internationally to justify their new cost structure. In trying to survive the talent shock, they are forced to become better companies.

This impact extends far beyond company payrolls. As frontier recruitment accelerates, it forces a painful question into the open: why is our local talent pool unable to supply these roles? The talent shortage stops being treated as an inevitable market condition and transforms into a critical problem that demands an institutional response. Universities face a clear and demanding test: are their computer science programs actually equipping students to work on frontier AI systems? That pressure forces universities to update outdated curricula, recruit top research talent, and invest in real lab infrastructure. The benchmark for a great computer science program is no longer whether its graduates can land a standard tech job at a local enterprise. The new benchmark is whether its top graduates can compete at the frontier.

Talent is just one domino. The exact same pressure propagates across every layer of the ecosystem. It pushes venture capital to adjust to longer investment horizons and technical risk. It reshapes procurement, compelling enterprise buyers to pay for real intellectual property rather than basic localization. It accelerates infrastructure, demanding the deployment of complex compute and developer stacks.

The point is not that pursuing AGI magically fixes each of these layers directly by itself. Rather, it sets a standard so demanding that structural weaknesses across the ecosystem suddenly become binding constraints. And once those constraints become impossible to ignore, the low-risk equilibrium breaks, forcing the entire system to respond and evolve to survive.

The clearest pitfall in this strategy is declaring an AGI ambition without a credible, practical execution path. That simply leaves us with empty press announcements. The tragic irony is that superficial claims propagate through all layers of the system just as quickly as real ones, creating hollow incentives. Pursuing AGI only acts as a real catalyst if we seriously target it and take concrete steps toward attaining it.

If the pursuit is genuine, whether we actually achieve AGI or not becomes secondary. As long as it leads to the birth of a vibrant, local frontier ecosystem, the mission succeeds. There will always be a next target to set and a higher bar to chase.

Which brings us to the much harder question: can Saudi Arabia realistically attain AGI if we pursue it, and how? Ambition and capital alone are not enough, and simply copying the playbook of established frontier labs in Silicon Valley or China is a dead end. Yet, I believe Saudi Arabia possesses unique structural advantages that make attaining AGI a realistic target.

That, insha Allah, will be the scope of Part II.