Microsoft is riding on geopolitics to extend its AI footprint

Welcome to AI Decoded, Fast Company’s weekly newsletter that breaks down the most important news in the world of AI. You can sign up to receive this newsletter every week here.

Microsoft extends its grip on global AI with $1.5B share of the UAE AI company G42

Microsoft’s recently announced deal to acquire a $1.5 billion share of United Arab Emirates-based AI company G42 is the culmination of months of coordination and negotiation among the companies and their respective governments. G42 will now begin getting computing power from Microsoft’s Azure data centers, and will resell certain Microsoft AI services.

Behind the U.S. government’s involvement in the deal is a desire to push back on China’s growing influence in the Gulf region and its bid for AI supremacy. The terms of the deal require that G42 and Microsoft comply with special security protections of technology assets and intellectual property, presumably to guard against leakage to China or other U.S. adversaries. G42 also agreed to strip out and replace all Chinese-made (Huawei) servers from its infrastructure.

For the U.S. government, the deal is a move in a burgeoning cold war with China—a rivalry where the chess pieces aren’t nukes and proxy armies but technological and economic influence. The Chinese government has announced more than $110 billion in technology merger and acquisition deals since 2015, says the Brookings Institute. The Chinese government is able to easily access AI developed in its private sector for use in intelligence gathering and defense—a concept the Chinese call “military-civil fusion.” The Chinese threat, including in AI, is the main justification for a yearly U.S. defense spend that could approach $1 trillion in FY2024. 

For Microsoft, the G42 deal represents a major step in a different kind of cold war. Microsoft has emerged as a central player in the AI arms race, and has been extremely successful in buying influence over some of the space’s most important companies. 

Microsoft was the first big tech company to realize the potential of OpenAI, and invest. In early 2023, it paid $10 billion for an estimated 49% ownership in the company. (It also backed CEO Sam Altman during the company’s brief and tumultuous leadership changes last fall.) In February, Microsoft bought a small $16 million share in the French AI upstart, Mistral, to have a key player in the global AI race. Microsoft was also an investor in Inflection AI, the company founded by DeepMind founder Mustafa Suleyman and LinkedIn billionaire Reid Hoffman. And it was ready when Inflection gave up its plans to be a consumer AI company, hiring away Suleyman and most of Inflection’s staff and nabbing the rights to distribute Inflection’s emotionally intelligent AI model through its Azure cloud.

“Satya [Nadella] gets the game,” says Emerj Artificial Intelligence Research founder Dan Faggella, speaking about the Microsoft CEO. “If the U.S. is unable to provide the power for his data centers, he is fully aware that he may be able to find the power he needs in the Middle East. There are companies in the Middle East that are in stealth mode that are raising huge amounts of money and building giant data centers that are purpose-built for pure AI work, for training and inference.” 

Microsoft has a long and friendly relationship with the U.S. government, including numerous contracts to build computer networks and even AR glasses for the military. It has legions of lobbyists in D.C., as well as people like Brad Smith, who is seen in the Capital as a trusted and friendly face from the tech world (much like Eric Schmidt). By helping the Biden administration in the sphere of geopolitics, Microsoft furthered its own influence in the AI world.

On X, a low-key scuffle between tech pundits and “builders”

Humane, a company founded by ex-Apple engineers, sent some A-list reviewers its new product, the AI Pin. Reviewers reviewed. Reviewers panned. Hard. The Verge’s David Pierce titled his review, “Humane AI Pin review: not even close.” YouTube star Marques Brownlee also said the wearable fell far, far short of expectations.

On X, some people have rushed to Humane’s defense. “I find it distasteful, almost unethical, to say this when you have 18 million subscribers,” Danial Vassallo said of Brownlee’s review. “Hard to explain why, but with great reach comes great responsibility. Potentially killing someone else’s nascent project reeks of carelessness.” Intercom founder Eoghan McCabe seemed to concur. “The anon haters so badly want to see this fail,” he said (though it’s not clear whom he was referring to). “I’d love if the rest of us builders could get on their side and give them a chance to win.” Marc Andreessen also chimed in with support for the AI Pin. Numerous others pointed out that Brownlee was just doing his job, and that it was Humane’s job to release a product that lives up to its hype. 

There’s probably something to both sides of this argument. I did find some of the reviews—the titles of the reviews, in particular—to be pretty cutting. Indeed, the title of Brownlee’s review (“The Worst Product I’ve Ever Reviewed . . . For Now”) was a bit more salacious than the review itself, which was pretty even-handed and almost gracious. On the other hand, Humane’s prelaunch presentation seemed slick and Apple-esque. Lowering the temperature on the marketing and PR might have helped lower the expectations of reviewers. 

Some findings from Stanford’s 2024 AI Index

Stanford’s Institute for Human-Centered AI (HAI) is out with its annual AI Index, which tracks recent trends in AI R&D, model performance, responsible AI, economics, policy, and public opinion. The report is meant to keep business leaders and policy makers informed on the latest goings-on in the burgeoning AI industry. Here are some of the research’s stand-out findings:

• AI is beating humans on some tasks, but not on all. AI beats humans on image classification, visual reasoning, and English understanding. Humans are still better at tasks like multitask language comprehension and visual common-sense reasoning. 

• Frontier models got way more expensive. The median costs of training state-of-the-art AI models nearly doubled in the past year. OpenAI’s GPT-4, for example,  used an estimated $78 million worth of compute to train, while Google’s Gemini Ultra cost $191 million for compute.

• Investment in generative AI has skyrocketed. Investment in AI overall declined last year, but funding for generative AI grew 8 times since 2022 to reach $25.2 billion in 2023. 

• AI is making workers more productive and leading to higher quality work. In 2023, numerous studies were published to assess AI’s impact on labor, suggesting that AI enables workers to complete tasks more quickly and to improve the quality of their output. 

• In the U.S., AI regulations have sharply increased. In 2023, 25 new AI-related regulations passed into law, a stark increase from just one in 2016. Last year alone, the total number of AI-related regulations grew by 56.3%.

More AI coverage from Fast Company: 

• Could the AI revolution crash our data centers?
• Big Tech is on a generative AI hiring spree
• What happens when the AIs turn on each other?
• It’s those deepfakes of local officials that are most worrying

Want exclusive reporting and trend analysis on technology, business innovation, future of work, and design? Sign up for Fast Company Premium.