Nvidia Buys into Europe’s AI, Accounts for 10-15% of Startup Investments in 2026

Set Trending Topics as a preferred source on Google.

Anyone founding a serious A.I. start-up in Europe now has a good chance of finding Jensen Huang on the cap table. Over the past three years, Nvidia has bought into 26 European tech companies, in some cases across several rounds, and it has now announced 13 more stakes. In Paris, Huang spoke of roughly $8 billion that Nvidia deployed in the region over the past year, plus $2 billion in new investments. Counting the announced deals, that is $10 billion within a year, and it buys the chipmaker something money alone rarely secures: a box seat in almost every important layer of European A.I. infrastructure, models, applications and customer relationships.

The money is there. With a market value of about $5.65 trillion, Nvidia is the world’s most valuable company, and in the first two quarters of its current fiscal year it has already earned about $118 billion (roughly €102 billion) in net income: $58.3 billion in the first and $59.7 billion in the second quarter. The $10 billion for Europe amounts to less than a tenth of that profit.

What exactly is inside the $10 billion, Huang has not explained. The figure comes from his speech in Paris, and Nvidia has not broken it down into equity and infrastructure. The company also keeps the size of individual checks under wraps almost every time. The tables below therefore list the total size of each round, and Nvidia’s share of it is usually not public.

Almost Every Major European A.I. Start-up Has Nvidia on Board

The list spans foundation models (Mistral AI, Black Forest Labs, AMI Labs), voice A.I. (ElevenLabs, PolyAI, Gradium), applications (Lovable, n8n, Legora, Synthesia), robotics (Neura Robotics, Flexion), autonomous driving (Wayve, Oxa), quantum computing and A.I. cloud (Nscale, Nebius). Nvidia is present across the entire stack: among the buyers of its chips, among the providers of computing power and among the companies that turn both into products. Gaps are rare.

Thirteen More Stakes Announced Around Huang’s Europe Tour

Around Jensen Huang’s trip to Europe, Nvidia announced 13 additional minority stakes. The company does not name amounts. Nvidia speaks of “planned” investments, so formal closing is not confirmed for every deal. Doctolib announced its planned stake itself and said it will support the expansion of its A.I. offerings for health care professionals. (The column “Latest known round” shows the companies’ funding status and says nothing about Nvidia’s share.)

Unicorns Without Nvidia: Helsing, Parloa and DeepL

The list is long, but it has gaps. The defense start-up Helsing recently raised $1.8 billion at an $18 billion valuation, with Dragoneer, Lightspeed, Iconiq, JPMorganChase, CPP Investments and General Catalyst, but without Nvidia among the named investors. The Berlin customer service start-up Parloa was valued at $3 billion after raising $350 million, led by General Catalyst, with EQT Ventures, Altimeter, Durable and Mosaic. At the translation specialist DeepL, Index Ventures led a $300 million round at a $2 billion valuation, with Iconiq and Teachers’ Venture Growth. Nvidia appears in none of the three announcements. That does not rule out smaller checks or earlier rounds, because Nvidia does not always disclose its stakes.

Twice as Much as the Scaleup Europe Fund, in a Fraction of the Time

For comparison: The E.U.’s Scaleup Europe Fund, managed by Sweden’s EQT, is meant to close Europe’s late-stage funding gap. The fund starts with a volume of at least €5 billion (initially €3 billion according to Science|Business, including €1 billion from the E.U. budget) and is meant to finance 30 to 40 companies over three to four years. A second phase targets €20 billion. The first investments are planned for the fall.

In nominal terms, $10 billion is twice as much as €5 billion, and depending on the exchange rate the factor is somewhat lower. The time frame widens the gap: The fund comes to roughly €1.25 to €1.7 billion a year, while Nvidia cites $10 billion for twelve months. Per year, that is many times the E.U.’s most important instrument for scale-ups. One caveat remains: Part of Nvidia’s money flows into infrastructure and listed stakes such as Nebius, and the sum is not reported as pure start-up capital.

How Large Is Nvidia’s Share of All European Venture Capital?

Nvidia’s $10 billion corresponds to roughly €8.6 billion at an exchange rate of about 0.86. Extrapolated to the full year 2026 using data from Dealroom ($44.2 billion in the first half) and Tracxn (about €30 billion in the first half, as reported by Trending Topics), this results in a share of 11 to 16 percent:

“Weaker” means the second half falls short of the first by the same ratio as in 2025 (Tracxn: €24 billion versus €30 billion). “Stronger” means it exceeds the first half by 14 percent, as it did in 2025. The calculation is an approximation, because Huang’s figure also includes infrastructure, and Nvidia’s money is simultaneously part of the round totals.

Taking a more conservative approach and subtracting the two large stakes in Nebius ($2 billion) and Nokia ($1 billion) leaves $7 billion, or roughly €6 billion. Nvidia would then account for 7 to 11 percent of tech investments in Europe (Dealroom: 7 to 9 percent, Tracxn: 9 to 11 percent). It remains unclear, however, whether Huang’s $10 billion figure includes these $3 billion at all or whether they come on top. If they come on top, the total would be $13 billion and the share 14 to 21 percent. If they are included in the $10 billion, the 11 to 16 percent shown above apply.

What Nvidia Gets Out of It: Insight Into Business Models, Growth and Customers

The thesis: With its minority stakes, Nvidia buys knowledge in addition to returns. Anyone invested in Mistral, ElevenLabs, n8n, Lovable or Doctolib sees firsthand which models customers demand, how quickly computing needs grow and which applications people actually pay for. Nvidia also plays a dual role as supplier: Nscale and Nebius, in which Nvidia holds stakes, are among the European cloud partners that rent out Nvidia chips. How deep the insights go in each case depends on contract details such as information rights or observer seats, and those are not public.

Nvidia describes its own role as that of an ecosystem builder. In the announcement launching the Nemotron Coalition, Jensen Huang says: “Open models are the lifeblood of innovation and the engine of global participation in the AI revolution.” Among the founding members are two companies from the table above, Mistral AI and Black Forest Labs. Nvidia is developing the coalition’s first base model with Mistral, trained on Nvidia’s DGX Cloud and then released openly. It is meant to become the foundation for the Nemotron 4 model family. On the planned stakes, Huang reportedly said that Europe needs to “get in the race” and pick up speed on open-source models, technological sovereignty and robotics.

The counterposition comes from Nvidia’s own account: The coalition partners receive computing power and know-how from Nvidia, the jointly developed model is meant to be openly available, and Nvidia describes the goal as an ecosystem in which intelligence is developed locally. From this perspective, the start-ups benefit from the relationship at least as much as the chipmaker does.

Hugging Face as a Lever for Sovereign A.I. With Nemotron

With its acquisition of Hugging Face for $12.9 billion (about €11 billion), Nvidia could directly serve the demand for sovereign A.I. in Europe. The platform was founded by French entrepreneurs and has about half of its workforce in Paris. It is the central hub for open models, where public agencies and companies download, adapt and run their own versions of open-weight models locally. Nvidia had already described exactly this in earlier announcements on European language models: Intelligence should be “developed and served locally,” the models should cover the E.U.’s 24 official languages, be made more efficient with Nemotron techniques and be fine-tuned on European infrastructure through a Hugging Face integration.

That suggests a possible chain: The stakes provide insight and proximity to the model builders, Hugging Face provides the distribution platform, and Nvidia’s own Nemotron family provides the product for customers who want a model that can run in Europe and is not hosted by a U.S. company. Nvidia has not said that this is the intention. Hugging Face co-founder Clément Delangue stresses that the platform will keep operating as an independently neutral unit inside Nvidia. Observers expect competition authorities in Paris and Brussels to review the deal. Advocates of open models, on the other hand, see a chance to reduce dependence on Big Tech, because open models can be run and audited locally.

Competitors: Other Corporations Are Investing Too

Nvidia is not the only strategic backer in the European A.I. market. ASML led Mistral AI’s Series C with €1.3 billion, Samsung leads the latest round, Salesforce Ventures appears at Mistral, Black Forest Labs and Lovable, Google at Ineffable Intelligence, Amazon at Neura Robotics and AMD Ventures at Recursive Superintelligence. What sets Nvidia apart is its breadth: No other strategic investor appears in the table at as many companies.