Nvidia Corp billionaire boss Jensen Huang (黃仁勳), clad in his signature leather jacket, has been crisscrossing European capitals and sharing the stage with the likes of British Prime Minister Keir Starmer and French President Emmanuel Macron as he pitches “sovereign” artificial intelligence (AI), a vision of new data centers offering essential compute power within national borders rather than via dominant tech firms from abroad. But if there were prizes for irony, it is a concept that might win a few.
Huang’s pitch has understandably struck a chord with leaders desperate for new sources of productivity gains and for ways to avoid falling terminally behind in a tech race dominated by the US and China. Recent announcements include a partnership with French AI start-up Mistral to build a cloud platform powered by 18,000 Nvidia Blackwell chips and a Germany-based industrial cloud for European manufacturing built with 10,000 Blackwell chips. It is not just Europe — Nvidia has cut big sovereign AI deals in the Middle East — but the Old Continent is where Huang sees computing capacity increase by a factor of 10 over the next two years.
“It’s coming,” he said.
Illustration: Constance Chou
However, this does not much look like sovereignty. From Nvidia’s point of view, the company is certainly positioning itself as a geopolitical actor, engaging directly with heads of state like Macron as the ultimate tech enabler to boost AI adoption. That is good for Nvidia amid a wider Sino-American trade war that has seen it lose US$15 billion in Chinese sales due to export controls and as Europeans become warier of US tech providers such as Alphabet Inc and Microsoft Corp. Bloomberg Intelligence last month estimated sovereign AI investments could add US$10 to US$15 billion in annual revenue for Nvidia in a de-globalizing world.
However, from Europe’s point of view, we are still a long way from the tech autonomy leaders like Macron want to offer anxious voters. The hardware and infrastructure powering these big AI projects are ultimately Nvidia’s, a US company with 80 percent market share whose dominance would be entrenched through chips that are updated or replaced every few years.
To the extent that there is a European supply chain, it exists elsewhere, in the brainy engineers and open-source models offered by the likes of Mistral. However, it remains to be seen if that would be enough to secure Europe’s AI future when US rivals are so dominant. Mistral’s 1 billion euros (US$1.2 billion) in capital raised so far is a fraction compared with OpenAI’s; and as ambitious as its plans are, Europe today has just 4.8 percent of estimated global AI compute power, according to Epoch AI.
We have seen this movie before. France and Germany once pinned their hopes on a sovereign cloud to protect user data from the extraterritorial reach of the US and China. Today, US tech companies still account for about 70 percent of cloud services in Europe. And for every attempt to reduce dependence, such as the Danish municipalities quitting Microsoft, there is an opposite move like the German military’s cloud deal with Google.
In search, European tools promoted as alternatives to Google have relied instead on Microsoft’s Bing — so when Bing went down during last year’s global outage, so did they. Is it any wonder that Microsoft is now offering “sovereign cloud” services to Europe without a hint of irony?
Misnomers and fuzzy language aside, some would argue this will not matter much if Huang’s vision of AI as an essential technology like electricity or the steam engine pays off. At last week’s Research and Applied AI Summit, a panel I moderated compared AI sovereignty to a national airline: The flags and operations are what count, not the origin of the aluminum tube and its engines.
However, there have been tangible costs to not worrying enough about dependencies — like Russian natural gas or Chinese exports — and AI might be the same, University of Amsterdam researcher Leevi Saari said. With tech still very much driven by globalized supply chains, outsourced labor and dominant vendors, what is on offer today looks more like “sovereignty-as-a-service” — the wrapper of autonomy at high cost. After all, maybe it is the existence of alternatives like Airbus SE that makes airlines so relaxed.
If AI sovereignty is a worthy goal, Europe will need to do more than come up with new wrappings for the same chips. It has advantages: talent, skills, companies like ASML Holding NV and an automotive-industrial base ripe for innovation. However, what it lacks is an ecosystem with plentiful research spending, financing and end-user demand — the kind that helped US start-ups raise more than double the funding of their European counterparts last year, AVP said.
That would not be changed in a day, but it should be part of any sovereign vision — as should investing in chip independence to secure “good enough” alternatives and diversify risk, Nathan Benaich of Air Street Capital said. The alternative, he reckons, is digital colonialism. As China heads down its post-DeepSeek path and the US hugs its hemisphere closer, expect to see Huang’s leather jacket more frequently.
Lionel Laurent is a Bloomberg Opinion columnist writing about the future of money and the future of Europe. Previously, he was a reporter for Reuters and Forbes. This column reflects the personal views of the author and does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
Chinese agents often target Taiwanese officials who are motivated by financial gain rather than ideology, while people who are found guilty of spying face lenient punishments in Taiwan, a researcher said on Tuesday. While the law says that foreign agents can be sentenced to death, people who are convicted of spying for Beijing often serve less than nine months in prison because Taiwan does not formally recognize China as a foreign nation, Institute for National Defense and Security Research fellow Su Tzu-yun (蘇紫雲) said. Many officials and military personnel sell information to China believing it to be of little value, unaware that
Before 1945, the most widely spoken language in Taiwan was Tai-gi (also known as Taiwanese, Taiwanese Hokkien or Hoklo). However, due to almost a century of language repression policies, many Taiwanese believe that Tai-gi is at risk of disappearing. To understand this crisis, I interviewed academics and activists about Taiwan’s history of language repression, the major challenges of revitalizing Tai-gi and their policy recommendations. Although Taiwanese were pressured to speak Japanese when Taiwan became a Japanese colony in 1895, most managed to keep their heritage languages alive in their homes. However, starting in 1949, when the Chinese Nationalist Party (KMT) enacted martial law
“Si ambulat loquitur tetrissitatque sicut anas, anas est” is, in customary international law, the three-part test of anatine ambulation, articulation and tetrissitation. And it is essential to Taiwan’s existence. Apocryphally, it can be traced as far back as Suetonius (蘇埃托尼烏斯) in late first-century Rome. Alas, Suetonius was only talking about ducks (anas). But this self-evident principle was codified as a four-part test at the Montevideo Convention in 1934, to which the United States is a party. Article One: “The state as a person of international law should possess the following qualifications: a) a permanent population; b) a defined territory; c) government;
The central bank and the US Department of the Treasury on Friday issued a joint statement that both sides agreed to avoid currency manipulation and the use of exchange rates to gain a competitive advantage, and would only intervene in foreign-exchange markets to combat excess volatility and disorderly movements. The central bank also agreed to disclose its foreign-exchange intervention amounts quarterly rather than every six months, starting from next month. It emphasized that the joint statement is unrelated to tariff negotiations between Taipei and Washington, and that the US never requested the appreciation of the New Taiwan dollar during the
RSS