A $1.5 billion deal with Microsoft won’t rupture the UAE’s ties to China.

Inferences from Minerva Technology Policy Advisors. Vol.11 - 22 April, 2024

Last week, it emerged that Microsoft's $1.5 billion deal with G42 had won approval from the US government, appearing to address concerns that the Emirati AI company could be a conduit for Chinese influence in the Middle East.

Inference: G42’s partnership with Microsoft suggests that the United Arab Emirates is gravitating towards the US’s technological orbit and away from China’s; the geopolitical reality is more complicated.

Why? The UAE’s pivotal role in the politics of the Middle East, its huge spending capacity, and its approach to AI-enabled government mean that US and Chinese interests will remain engaged in the region across a wide front of technologies and infrastructure. 

One deal will not be enough to break off China’s interests, or to assuage US concerns about China’s technological influence in the region.

So, what is G42? The group, founded in 2018 and based in Abu Dhabi, bills itself as “a global leader in creating visionary artificial intelligence for a better tomorrow.” It:

• has close ties to the Emirates government; Sheikh Tahnoun bin Zayed Al Nahyan, who also serves as the UAE’s national security advisor, is G42’s controlling shareholder;

• is involved in pretty much everything: developing data centers, exaflop supercomputers and AI systems for energy, healthcare, and beyond; and

• owned a stake in the company that developed a messaging app, ToTok, which Android and Apple removed from their app stores in 2019 following allegations that it was being used as a government surveillance tool.

What’s happened? At the end of last year, G42 announced a partnership with OpenAI, and moved closer to the US ecosystem and Microsoft (OpenAI’s sole cloud and compute services provider). 

More recently, the Abu Dhabi-based firm:

• drew attention from US officials; who first shared concerns in June alleging that G42 could be a backchannel through which advanced US technology might fall into Chinese hands. Then in January, the Department of Commerce started investigating G42. US lawmakers had also heaped on pressure, threatening trade restrictions;

• stripped back its physical presence in China, including its office sites, and focused on deploying more capital in the US;

• wound down its Chinese investments, including a stake in Bytedance, the parent company of TikTok, in order to maintain access to high-end Nvidia GPUs and;

• said, via its Chinese-born CEO Peng Xiao, that it “cannot work with both sides.”

Why was it significant? The deal between G42 and Microsoft is: 

• obviously political; the US government’s involvement in encouraging Microsoft into the move has been widely recognized; Microsoft and G42 said they had provided assurances on safety and compliance to both the US and UAE governments, which were closely consulted on the deal. Microsoft already had a direct agreement with the Emirati government to build out its cloud capabilities, as of September 2023.

• of wider infrastructural significance; G42 has promised to stop using Chinese-made equipment and software, including a number of Huawei systems; and use Microsoft’s Azure architecture exclusively. There is, however, likely to be a lag between this promise, and the reality; underlying components are cumbersome to swap and reliance on Huawei and other Chinese products will almost certainly run on for a time; and

• economically attractive; as the UAE (and the broader region) look for new economic opportunities beyond fossil fuels, finance and technology are coming to the fore. Also, nearly 400 million people around the world speak Arabic. G42’s Arabic LLM, Jais, was developed using Cerebras Systems’ supercomputer. After its partnership with OpenAI, there’s a chance that Jais gets an upgrade. That could give both OpenAI and Microsoft an opening in this, as yet, largely untapped AI market (in much the same way that Sam Altman reportedly plans to do with OpenAI Japan, announced last week). 

Still, the deal doesn’t signal a total realignment of the UAE’s commercial technology strategy, nor does it counter other significant economic links between China and the Emirates. 

Deep ties between China and the UAE are sustained by about 250,000 Chinese citizens who live in the UAE, among the million Emirati citizens in the country. The majority of China’s trade with Europe and Africa comes through Abu Dhabi and Dubai, and Chinese firms have built much of the UAE’s housing and infrastructure. Its leader, Sheikh Mohamed, wants to raise the bilateral trade volume with China to $200 billion by 2030. The UAE’s leading telecoms provider and Huawei are still engaged in a strategic partnership to build 5G and other advanced telecommunications networking infrastructure. Tencent likewise announced plans to make “huge new investments” in the UAE last month in data storage and cloud.

Moreover, the assurances provided by the new partners around AI safety and security don’t necessarily mean that Washington’s concerns about G42 are fully resolved; tough scrutiny of G42, its partners, and its management is likely to continue. 

The upshot. Saudi Arabia and other developing markets that are trying to build out their domestic tech capabilities will be watching this development, and its consequences closely. Without a clearer and more consistent plan from Washington for converting pressure on other countries to drop Chinese technology into more positive trade and commercial partnerships, the game is still very much afoot. G42’s partnership with Microsoft is a beginning, rather than an end to this process.

What we’re reading:

• The Washington Institute’s analysis on the China-UAE-US Triangle makes for an interesting read in light of the deal (it was published shortly beforehand).

• Anja Manuel, executive director of the Aspen Security Forum and a former US government official, opinion on US and Europe's need for a more cohesive strategy in the contest over AI.

• A testimony before Congress, by Jon B. Alterman, on China and the Middle East.

What we’re looking ahead to:

• 30 April: Deadline for UK regulators (including CMA and ICO) to submit strategies on AI governance.

• 8 May: RSA Conference in San Francisco — Please join Minerva for our second Global AI + National Security Symposium on May 8. Invite below.

• 17 May: Bank of England Prudential Regulation Authority (PRA) Supervisory Statement on Model risk management principles for banks comes into effect.

• 21-22 May: AI Seoul Summit in South Korea (virtual leaders’ meeting; in-person ministerial).

• 30 - 31 May: AI for Good Summit in Geneva.

• 13 - 15 June: Summit of G7 leaders in Borgo Egnazia.

• November: Global AI Safety Summit in France.