The myth of AI independence

A version of this article originally appeared in Quartz’s AI & Tech newsletter. Sign up here to get the latest AI & tech news, analysis and insights straight to your inbox.

A trillion-dollar bet is enticing governments across the globe, built on a promise of digital independence. Construct your own data centers, process your citizens' data domestically, and capture economic value from the AI boom instead of watching it flow exclusively to Silicon Valley. From Southeast Asia to the Middle East to Europe, governments are committing billions to what they're calling "sovereign AI," a new form of digital independence for the era of intelligent machines.

It's being framed as digital decolonization. Just as colonial powers once extracted oil and minerals while offering minimal benefits to locals, the argument goes, a handful of American tech giants now extract data, transform it into intelligence, and sell it back to the countries it came from. Sovereign AI promises to break that cycle.

But there's a fundamental contradiction at the heart of this giant buildout. Countries are pursuing sovereignty by making themselves more dependent on the very foreign companies they claim to be protecting themselves from.

The sovereignty paradox

The pattern is playing out everywhere. In Indonesia, the country's second-largest telecom plans to increase its AI server capacity 25-fold by 2028 — a roughly $5 billion investment it can't make without Nvidia $NVDA chips and American cloud partnerships.

Saudi Arabia secured preliminary approval to buy 18,000 Nvidia chips during President Donald Trump's visit in May and expects to purchase several hundred thousand of the company's top processors over the next five years. 

India is receiving a $15 billion investment from Google $GOOGL. Thailand is getting $5 billion from Amazon $AMZN Web Services. France is building what it calls Europe's largest AI data campus through Mistral, its domestic AI leader —  in partnership with Nvidia. 

The justifications for AI sovereignty vary. Some countries want to ensure AI models incorporate local languages and values. Others worry about sensitive healthcare data flowing into foreign systems. Smaller nations argue that domestic infrastructure guarantees cheaper, more reliable access for local companies and researchers who are otherwise "always at the back of the line" for computing resources.

But none of these projects deliver anything close to self-sufficiency. Nvidia, based in the heart of Silicon Valley, dominates the AI chip market with a roughly 90% share. Its only serious rival, AMD, is also American. The servers that house these chips come mostly from Dell $DELL and Supermicro — both American. Even China, which has built something close to a self-sufficient AI stack, hasn't yet developed an alternative to American processors.

U.S. export controls on advanced AI chips give Washington leverage over every country seeking to build AI infrastructure. Saudi Arabia's chip purchases are still pending final approval months after Trump's visit. Future sales may be tied to individual trade deals, and countries find themselves caught between U.S. security demands and their existing relationships with Chinese tech companies.

The cheaper alternative nobody wants to admit exists

Amid giant expenses and already tight budgets, there's another complication: The tech giants could deliver the same computing capacity for less. Amazon and Microsoft $MSFT already pitch "sovereign cloud" offerings with enhanced data controls and dedicated local infrastructure. These products promise data localization and security controls, letting countries build national AI models on top of existing infrastructure.

The hyperscalers' scale gives them bargaining power with suppliers like Nvidia that individual countries lack. They can deliver computing capacity faster and cheaper than governments building from scratch.

But these commercial solutions can't deliver true sovereignty either. U.S. companies remain subject to American jurisdiction — specifically the CLOUD Act, which allows American authorities to access data these companies hold anywhere in the world. When French legislators pressed Microsoft about whether it could guarantee that French citizens' data wouldn't be transmitted to the U.S. without French approval, the company couldn't make that promise.

This creates what critics call "sovereignty as a service," with countries getting the appearance of control while foreign corporations maintain the actual power. The arrangement resembles colonial-era build-operate-transfer schemes, where European companies constructed critical infrastructure in colonies under the guise of development.

The modern version may be worse. These public-private partnerships operate with minimal transparency. Contracts with tech giants often hide crucial details about data access and algorithmic control from legislative scrutiny. Subscription-based models give corporations the ability to remotely update or disable systems, meaning countries risk finding their "sovereign" infrastructure turned off if they don't comply with corporate interests.

None of this means countries shouldn't invest in AI capabilities. But the gap between sovereign branding and actual control is widening. Countries that have achieved meaningful digital independence have invested in domestic technical expertise, adopted open standards, and built in the ability to switch providers. 

Without those elements, national AI projects risk becoming expensive showcases for foreign technology — rather than genuine infrastructure under local control.