The $100 billion alliance that's changing the rules of the AI market

BarcelonaNvidia and OpenAI have announced the largest investment in the history of technology: $100 billion (€95 billion) to build the largest artificial intelligence (AI) infrastructure ever conceived. The deal, which surpasses the $70 billion Microsoft paid for Activision Blizzard, could determine who controls the future of AI.

The scale of the project is unprecedented. The planned 10 gigawatts (GW) of capacity is double all the chips Nvidia sold last year and is equivalent to the output of ten nuclear reactors, even surpassing the five gigawatts of Donald Trump's Stargate project, which involves Oracle, OpenAI—yes, also OpenAI—and Soft. (€475 billion).

However, the alliance has a peculiarity that raises doubts among some analysts: it is a circular system in which Nvidia funds OpenAI so that it can buy its products. According to New Street Research, for every $100 that Nvidia invests, OpenAI will spend $350 on chips from the Californian company. Financial Times warns: "This agreement of mutual favors is reminiscent of the bubble in financing telecommunications operators by their suppliers starting in 2000."

Two companies in different situations

For Nvidia, the world's most valuable company, with a market capitalization of €3.6 trillion, the transaction comes at the height of its technological dominance. With a quarterly cash flow of €60 billion, second only to Apple, the company led by Jensen Huang can afford this gigantic bet.

And not only that: Nvidia's strategy goes beyond OpenAI. The company not only has a stake in xAI, Elon Musk's company, but has also just committed another €5 billion (€4.75 billion) to Intel, its traditional rival, to jointly develop chips that combine Intel's x86 processors with Nvidia's GPU architectures. This diversification reduces dependence on Taiwan's TSMC in manufacturing and reestablishes Intel as a strategic partner in the American supply chain.

For OpenAI, the investment comes at a critical time. Despite generating annual revenues of 12 billion euros (11.4 billion euros) and having 700 million users, Sam Altman's company expects to lose 44 billion euros (41.8 billion euros) by 2029. The alliance with Nvidia and the implicit guarantee that this represents incipient means that the money pays an interest rate three times that of a chip manufacturer.

Rivals respond

The Nvidia-OpenAI alliance has shaken up the industry. Microsoft, OpenAI's flagship partner, sees this privileged relationship threatened, despite having more than quintupled its annual spending on Nvidia chips, from $2 billion to $11.5 billion. However, the company led by Satya Nadella is already developing its own Maia 100 AI chips to reduce this dependence.

Meanwhile, Google is accelerating the development of the v5p generation of its TPUs (tensor processing united), a direct competitor to Nvidia's H100s, and participated in 120 venture capital rounds in 2024, more than double Nvidia's 49. Finally, Amazon is developing its own Trainium and Inferentia AI chips for its AWS cloud and has signed a €4 billion (€3.8 billion) agreement with Anthropic that stipulates the exclusive use of its semiconductors.

Looking at Nvidia's most recent report, the concentration of power around it is evident: just six direct customers provide 85% of its quarterly revenue, and two of these unspecified customers account for 39% of the total.

China: Total Ban and Own Alternatives

China's response has been forceful and coordinated. The Cyberspace Administration of China has ordered a complete ban on the purchase of AI chips from Nvidia by the country's largest companies, from Alibaba to ByteDance to Tencent. This measure affects both the H20 and RTX Pro 6000D chips and is equivalent to expelling Nvidia from the world's second-largest market.

Furthermore, the ban has spurred domestic innovation. Cambricon, a Chinese AI chip manufacturer, saw its revenue grow by 4,400% in the first half of 2025 and has become the best-performing stock on the Chinese stock market. Huawei has unveiled an aggressive roadmap for its Ascend AI chips and will announce new models between 2026 and 2028.

Chinese tech giants have responded with massive investments. Tencent has returned to the markets after four years to fund its AI expansion, Alibaba has secured $3.2 billion (€3.04 billion) through convertible notes, and has launched more than a hundred open-source AI models (some of which even work in Catalan). https://www.ara.cat/media/ia-whatsapp-nega-respondre-catala-ne_1_5457006.html), while Baidu has issued bonds worth 4.4 billion yuan (570 million euros).

"The consolidation of the US AI infrastructure is further evidence of the attempt at technological containment," argues Chinese official media, justifying the push toward complete technological sovereignty.

Europe: ASML and regulation

Europe has found its champion in ASML. The Dutch company, now the most valuable on the continent with a market capitalization of €266 billion after surpassing SAP, has invested €1.3 billion in the French startup Mistral AI, making it its largest shareholder.

The transaction creates a strategic triangle in Europe: Nvidia provides computing power, Mistral develops AI models, and ASML integrates these capabilities into its lithography equipment to reduce defects by up to 15% and thus improve the performance of semiconductor manufacturing processes.

Meanwhile, while the €43 billion European chip law aims to double the European market share to 20% by 2030, the European Commission is intensifying its regulatory response. The French authorities have opened an antitrust investigation into Nvidia.

A new competitive landscape

The alliance marks the emergence of a paradigm in which computing becomes a strategic raw material. The global AI market, which will reach $1.5 trillion (€1.4 trillion) by 2025, is no longer governed by pure competition, but by integrated ecosystems.

Technological fragmentation is accelerating: the United States, China, and Europe are building incompatible ecosystems. Nvidia's dual strategy (OpenAI and Intel) ensures simultaneous demand and supply. It remains to be seen whether this circular financing model is sustainable.