Swing Vote of Global AI Competition: the Middle East

The global AI competition is often framed as a 1-on-1 battle between the US and China, with a “Cold War II” undertone as the backdrop. In this tussle, there is one region that is growing more powerful and critical to which side will eventually win (if there is any winning at all): the Middle East.

Akin to how a few thousand swing voters in a state like Pennsylvania, Michigan, or Virginia can decide the outcome of the US presidential election, a few countries in the Middle East, specifically Saudi Arabia, the UAE, and Qatar, are becoming the “swing vote” that could tilt the global AI competition in either the US or China's favor.

This dynamic is markedly different from the role that the Non-Aligned Movement played during the Cold War between the US and the Soviet Union. Instead of desperately seeking neutrality between the two power blocs, while trying to avoid the devastating fate of becoming a proxy war location, e.g. Vietnam and Korea, the Middle East holds natural advantages and real leverage over the two superpowers.

The region is being courted as a major customer and source of influence, not stomped upon as a convenient place to fight proxy battles. And the leaders who hold the “swing vote” – all kings and royal families of hereditary kingdoms, not heads of states elected by a democratic process  – are increasingly skilled in wielding their leverage.

Energy, Money, Geography

The Middle East’s natural advantages can be broken down into three simple elements: energy, money, and geography. Each has its own nuances. 

Energy is perhaps the most straightforward one to understand. Everyone who has followed the rise of generative AI knows that massive data center build outs require lots of energy. The logical chain is simple: GPUs are power hungry, we need more energy to power them, energy rich countries stand to benefit, and Middle Eastern countries are all energy rich (though not exactly clean energy rich).

Money is a bit more complicated than meets the eye. Being oil rich countries, the Middle Eastern kingdoms like Saudi Arabia and the UAE all sport massive sovereign wealth funds. Billions of dollars are being channeled from them to modernize the hardware of these countries, like fancy skyscrapers and new roads, futuristic new cities like Neom, and digital infrastructure projects, all of which naturally making them deep-pocketed customers for all things AI.

However, the Middle East’s willingness to splurge on AI has long predated the recent surge of AI infrastructure spending. As wealthy clients of leading venture capital funds across the US and China, the Middle East has been placing bets on AI for many years, just indirectly. A great case study is Sanabil, a subsidiary of Saudi Arabia’s sovereign wealth fund, Public Investment Fund, which is in charge of deploying $3 billion per year into venture capital firms or tech startups directly. An unexpectedly transparent organization, Sanabil shares all the VC firms it is an LP of directly on its website. Among them are the “who’s who” of the top-tier firms across the US – a16z, Altimeter, Coatue, Founders Fund – and China – 5Y, Yunqi, HongShan (fka Sequoia China).

These relationships guarantee exposure into the leading AI companies on both sides of the AI competition, regardless of whether the US or China is in the lead. More recently, the venture arm of Saudi Arabia’s state-owned oil company Aramco, called Prosperity7, participated in the latest funding round of Zhipu AI, one of China's leading generative AI foundation model startups. Many regard Zhipu as China’s closest comparable startup to OpenAI. Given how split the generative AI landscape is along geopolitical fault lines, Prosperity7’s move into Zhipu raises eyebrows as the first non-China investor to directly invest in a Chinese AI startup.

Given the level of sophistication Middle Eastern sovereigns are playing the “smart money game”, what Prosperity7 did may not be a one-off deal. It’s a strong display of leverage and independence in the face of strong geopolitical pressure to fall in line with either the US-led sphere or the Chinese camp. The Middle Eastern capital has, so far, been able to play both sides with adroit. 

Geography is one of the least appreciated advantages the Middle East possesses in the future of global AI development. The physical location of data centers play a crucial role in the speed and quality of delivery in digital services to end users; the closer the data center, the faster and more reliable the service. Furthermore, a properly distanced group of data centers is also necessary in providing disaster recovery backup in case a data center goes down, which is unavoidable.

The Middle East’s location – nestled in between three large clusters of Internet users from Europe, Asia, and Africa – to speak nothing of its own growing digital citizenry, presents perhaps the perfect location to train and deploy generative AI models and applications. As we explored in our recent post on the Houthis accidental slashing of the cables that connect data centers under the Red Sea, 80% of the web traffic between Asia, Europe, and Africa goes through the Middle East. That is a lot of traffic!

Dedicated GPU data centers fueled by cheap power makes it an ideal location for training large models. Proximity to millions of web users make the same data centers ideal for running inference workloads to deliver generative AI services. Being in the middle (literally) of so many dynamic regions also make it a good location to build backup data centers for redundancy and disaster recovery.

Geography may have played an indelible role in the ongoing political instability and developmental tragedy of the Middle Eastern region. But if it knows how to take advantage of the same geography to enable a future of artificial intelligence, then it can turn a curse into a blessing. 

Courting the Swing Vote

The relative efforts between the US and China to court the Middle East have, so far, been uneven.

China’s engagement is more aggressive and tangible. This makes sense, because China is also being aggressively “boxed in” by US export control and coordinated measures by other US allies. Thus, at least when it comes to generative AI development, China is still meaningfully behind the state of the art development in the US. To make up for this disadvantage, more proactively partnering, investing, and winning friends among the Middle Eastern countries is a sensible, if not only, strategy.

Huawei, unsurprisingly, is leading the way by expanding its cloud data center presence in Saudi Arabia. In a more “culturally sensitive” tactic, Huawei Cloud used its in-house AI model, Pangu, to develop the first Arabic language LLM deployed in its first cloud region in Egypt. Alibaba Cloud has also stepped up its expansion effort with a joint-venture partnership with the Saudi Cloud Computing Company. On a more official level, bilateral cooperation in AI was very much on the agenda when Sheikh Mohammed bin Zayed al-Nahyan, who rules the UAE, made a state visit to China in May.

Meanwhile, the US is sending more mixed signals. On the one hand, Microsoft’s $1.5 billion investment into G42 of the UAE with the explicit backing and encouragement of the Biden administration, is a clear move in pushing US technology more aggressively into the region while lessening China’s. Part of the deal is conditioned upon G42 removing Huawei equipment from its stack and cutting ties with China.

On the other hand, the US export control umbrella has reached the Middle East last year, restricting the most advanced GPUs from Nvidia and AMD from being sold to some Middle Eastern countries for fear of Chinese firms gaining access to them there. More recent developments, like Nvidia’s new partnership with Ooredoo, Qatar’s national telecom company which has 26 data centers throughout the Middle East, Africa, and even Indonesia, is drawing national security scrutiny, even though the Nvidia products Ooredoo is procuring should fall outside current export control restrictions. Even the Microsoft-G42 deal is being re-litigated; Republicans in Congress have recently requested an official intelligence community assessment of the agreement.

The US’s mixed posture is understandable as well. While pushing out Huawei around the world has been a priority for many years, the US has not had a good answer to offering a decent alternative to Huawei technology when it comes to 5G. In AI, the US holds most of the cards in advanced technology, so it offers a clearly superior alternative to Huawei. Yet it is also concerned about restricting access, because the way AI models are accessed is mostly via cloud APIs, which is a leaky bucket – a topic I explored a bit previously in a discussion of OpenAI’s API ban on China.

At the end of the day, the US’s courtship of the Middle East is lukewarm because it can afford to be. America is home to all the most consequential generative AI companies and most of the technologies. It does not need the Middle East “swing vote” nearly as much as China does. However, it would serve America’s interests well to start treating the Middle East like the crucial “swing vote” that it is, given all its aforementioned advantages, and not adopt a “my way or the highway” Cold War mentality, like what some of the more immature voices on the geopolitics of AI is saying.

As a region and culture with centuries of history (so more like China less like the US), the Middle East has as strong a sense of its place in world history and human civilization as any. It is the quintessential battleground region, whose “swing vote” may decide the eventual winner of the global AI competition.