Sam Altman and the $7T geopolitics of AI chips | The AI Beat

Sure, the Super Bowl was exciting — Kansas City, Travis Kelce, Taylor Swift, yada yada yada. But not even the hometown challengers, the San Francisco 49ers, could keep Silicon Valley as engrossed as the news about OpenAI CEO Sam Altman and his quest for up to $7 trillion in AI chip funding, which the Wall Street Journal reported last Thursday.

That’s because Altman kept his X posts coming, one after the other, all weekend long — from short, cryptic bites like “chaotic good” and “also roon is my alt” to slightly longer missives insisting that he doesn’t “really know that much about this rumored compute thing” and pushing back on haters with “you can grind to help secure our collective future or you can write substacks about why we are going fail.”

But to me, Altman’s post-dump is far less interesting than the complex geopolitics of AI chips, which ground the funding reports in real news that we all should pay attention to. A quick online search on the issue, in fact, will take you on a fascinating trip around the globe.

Around the world in AI chip geopolitics

You can start your geopolitical AI chip tour in Taiwan — where TMSC manufacturers most of today’s chips suitable for AI workloads, including those designed by Nvidia, based in Santa Clara, California, which enjoys more than 80% of the market for high-end AI chips.

Next, head to China, which also relies on TMSC-manufactured chips, particularly from Nvidia, but also has an ambition to control Taiwan — which is why the US restricted Nvidia exports of its most advanced AI chips to China (which, in turn, led Nvidia to develop downgraded, China-focused chips).

Then, travel east to Washington DC, where you’ll find the Biden Administration working overtime to reduce reliance on TMSC, in an effort to fend off China in the global AI race. In August 2022, President Biden signed the $53 billion “CHIPS and Science Act of 2022” (“CHIPS Act”) into law. It aimed to “boost American semiconductor research, development, and production” by providing $280 billion in funding, primarily towards R&D.”

Just last Friday, Fortune reported that the Biden administration announced it would “direct $5 billion in CHIPS Act money toward a new training facility to boost workforce participation in a semiconductor industry dominated by foreign talent.” This is just a preview of what’s to come: “After a lengthy review period, the government will start giving out billions more over the coming months, primarily in the form of grants to domestic chip manufacturers such as Intel.”

After that, take a flight across the Atlantic Ocean until you hit the United Arab Emirates, where Sheikh Tahnoun bin Zayed al Nahyan, who has served as the country’s global security advisor since 2016, is the chair of AI firm G42.

AI chips come full circle in Abu Dhabi

That’s where it all starts to tie together: G42 recently announced agreements with European pharmaceutical giants like AstraZeneca and a $100 million deal with a Silicon Valley firm to build what the companies boast will be the “world’s largest supercomputer.” Last month, G42 announced a partnership with OpenAI, the creator of ChatGPT.

In addition, a big part of the Wall Street Journal report is that Sam Altman is reportedly in conversations with G42 to fund his $7T shopping spree, and today Bloomberg reported that G42 “is paring back its presence in China and has pledged to invest in key Western markets in an effort to assuage US concerns over the artificial intelligence firm’s ties to Beijing.”

But that’s not all: Nvidia, too, is courting UAE investors. According to The Circuit, CEO Jensen Huang appeared at the World Governments Summit in Dubai and told the UAE’s Minister of State for Artificial Intelligence Omar Al Olama that every country should have its own AI infrastructure and protect its “data sovereignty.” The report added that when Al Olama asked Huang about the $7 trillion OpenAI CEO Sam Altman is seeking, “Huang quipped that Altman was apparently looking to buy ‘all’ of the world’s AI chips.”

It’s not just about the money

Back in the US, Altman may be saying “show me the money,” but when it comes to AI chips, it’s not just about cold hard cash.

Ronen Dar, co-founder and CTO of compute orchestration platform Run.ai, told me by email that Altman’s discussions on chip manufacturing “highlight the strategic importance of computing in the AI world, as well as the critical need for chip supply among AI companies.”

But he added that regardless of geopolitical tensions, for an AI startup to rise up above Nvidia, it would take not just a significant financial investment from abroad, as well as a shift in reliance by data teams who have deeply integrated Nvidia’s software into their workflows and are accustomed to its ecosystem (something which I reported on deeply a year ago).

Chip industry analyst Karl Freund also added that he felt Altman’s efforts to solve the global supply chain problem while seeking lower costs for high-end AI chips are “at least one bridge too far.”

The amount of money needed, he told me by email, “would only be surpassed by the amount of talent he will struggle to attract and the time needed to get this right.”

In a follow-up article today, the Wall Street Journal agreed that Altman’s plan is “far more complicated than money.”

“Manufacturing chips is enormously capital intensive,” the article says. “It is also one of the most intricately complex industries in the world with a history of sharp cyclical swings that have made companies wary of radical expansion.”

And, of course, as I reported last week, Altman’s dream of AI chip domination is not just about the money, it’s about the massive potential environmental impact based on scarce water and mineral sources.

“If it does work out, the amount of natural resources that will be required is just mind-boggling,” Hugging Face climate lead and researcher Sasha Luccioni told me. “Even if the energy is renewable (which it isn’t guaranteed to be), the quantity of water and rare earth minerals required is astronomical.”