Jensen Huang's Tough Love Tour: Why Europe Must Stop Consuming AI and Start Producing It

In October 2026, Jensen Huang did something unusual for a CEO courting a continent's business. He told Europe it was thinking too much.

Over the course of a week, the NVIDIA CEO visited London, Paris, Stockholm, Berlin, and Dublin. He met with heads of state, launched a supercomputer, signed a joint declaration with Germany's chancellor, and invested in startups across five countries. But the headline from every stop was the same: Europe has everything it needs to lead in AI, and its biggest obstacle is itself.

This was not a sales pitch wrapped in flattery. It was a lecture, delivered with the directness of someone who has spent forty years building a company now worth more than $3 trillion. Huang's message to European leaders was simple, uncomfortable, and repeated in every city: stop overthinking, stop over-regulating, and start building.

"You Overthink Things"

The sharpest version of the message came in London. Speaking to a packed audience, Huang addressed Europe's tendency to find reasons not to act:

"You have smart people, you have great culture, you have young people, and the energy is incredible. But for some reason, if there were a million reasons to win, you will find the one reason not to do it. I think you overthink things" [1][2].

In Berlin, standing beside Chancellor Friedrich Merz, he went further. Addressing Germany directly, the country that invented the automobile and the printing press, Huang said:

"Nothing pains me more, frankly, than to see one of the most technologically advanced countries in the world falling so far behind... Maybe it is the European way to overthink everything. Just get to work. Instead of looking for reasons not to do something, you need to look for ways to do something" [3][4].

When Merz raised the EU AI Act and its approach to regulating AI technology, Huang's response was blunt:

"You can't regulate if you don't have the technology. There is nothing for regulation to look for if the industry itself is not thriving... We should regulate the applications right away, not the underlying technology" [5][6].

The pattern was unmistakable. In four countries across ten days, Huang delivered the same diagnosis: Europe's intellectual reflex to analyze, debate, and regulate before building is the very thing preventing it from competing.

Sovereign AI: Don't Outsource Your Intelligence

Alongside the tough love, Huang advanced a specific strategic argument. He calls it "sovereign AI," the idea that nations must build, train, and control their own AI models rather than importing intelligence from American or Chinese companies.

In Paris, during a panel with President Emmanuel Macron, he put it in terms no policymaker could ignore:

"You can't afford to outsource your intelligence. Your country's intelligence is encoded with and embodies the knowledge of your people, your history, culture, common sense and your values. That is the definition of intelligence. You must not outsource that" [7].

In Stockholm, at the launch of the Sferical One supercomputer, a joint venture between Ericsson, AstraZeneca, Saab, SEB, and the Wallenberg Foundations, Huang reinforced the point:

"There's no reason for you to export your data, only to import back the intelligence" [8].

And at the GTC Berlin keynote, the message became a directive: "Europe needs to stop just consuming AI. It needs to focus on producing AI" [9][10].

The sovereign AI argument is more than rhetoric. It addresses a real structural risk. Today, the vast majority of AI models that European enterprises use are trained, hosted, and controlled by American companies. The data flows out, the intelligence flows back in, and the value capture happens somewhere else. Huang's point is that this is not just a competitive problem. It is a sovereignty problem.

But sovereignty is not a single thing. It is layered. Owning the compute infrastructure is one layer. Choosing which models to run, and whether to train your own, is another. Controlling where the data lives and how it moves is a third. And maintaining operational security, the ability to patch, segment, and isolate without depending on a foreign provider's timeline, is a fourth. The enterprises and governments that treat sovereignty as a checkbox rather than an architectural discipline will find they have traded one dependency for another.

Physical AI: Electrons Will Never Replace Atoms

For European manufacturers, Huang offered a different message: AI will not replace your industrial base. It will make it more valuable.

At the Invest in Germany summit in Berlin, he addressed a room of industrial executives directly:

"AI can only augment your capability. It will not replace your capability. So take it" [11][12].

He expanded on this at the GTC keynote, framing Europe's manufacturing heritage as an irreplaceable asset:

"The next wave of AI is physical. Electrons will never replace atoms... Every single industry in Europe, whether it is the automobile industry because we're building self-driving cars, the way we manufacture with robotics, the way we discover new drugs in healthcare, and all the way to financial services, all of these different industries have to become AI companies. Every single one of them" [12][13].

Huang called this "Physical AI," the application of AI to the tangible world of factories, vehicles, robots, and supply chains. He argued that no region on earth has dedicated itself to mechatronics, electronics, and mechanical systems the way Europe has [10]. That expertise, he said, cannot be replaced. But it can be amplified.

The Money Behind the Words

Huang backed his message with commitments. Over the course of the European tour, NVIDIA announced nearly $10 billion in cumulative investments across Europe's technology ecosystem [10][14].

The investments spanned infrastructure and startups. In the UK, NVIDIA committed $250 million in compute for agentic safety and security research [15]. In France, Huang met with Macron and backed four French startups, with Mistral AI's compute platform running on 18,000 GB200 systems [6][14]. In Sweden, the Sferical One supercomputer went live with 1,152 Blackwell Ultra GPUs [8][16]. In Germany, NVIDIA signed a Joint Declaration of Intent with Chancellor Merz and invested in four German startups, including Agile Robots, Avelios Medical, and Reflex Aerospace [11]. Germany's federal government is backing the initiative with hundreds of millions of euros in funding for computing capacity in Deutsche Telekom's Industrial AI Cloud, built on NVIDIA Blackwell hardware [17].

NVIDIA also announced investments in 14 European startups during the tour week alone [14]. The investments were not charity. They were infrastructure for a thesis: if Europe builds its own AI compute, trains its own models, and applies them to its own industries, NVIDIA sells the hardware that makes it possible.

From Sovereignty Talk to Sovereign Architecture

Huang's message lands differently depending on who is listening. For heads of state, sovereign AI is about national competitiveness. For enterprise architects, it translates into a concrete set of infrastructure decisions: where AI workloads run, who controls the data, and what happens when the provider's priorities diverge from yours.

This is where the gap between consuming AI and producing AI becomes tangible. Most European enterprises today run AI inference through API calls to models hosted on American hyperscale clouds. The data leaves the enterprise boundary, crosses jurisdictions, and lands on infrastructure the enterprise does not own, cannot inspect, and cannot control. For regulated industries, healthcare, financial services, defense, this is not a theoretical risk. It is a compliance problem that already limits what they can build.

The public cloud model was designed for general-purpose compute elasticity. It was not designed for the specific demands of sovereign AI: data residency, model isolation, accelerator-level telemetry, and the ability to operate without any dependency on external control planes. When Huang says "don't outsource your intelligence," he is describing the architectural problem that enterprises face every time they send training data or inference requests to infrastructure they do not control.

VMware Cloud Foundation addresses this at the platform layer. VCF provides the compute, networking, storage, and security stack that lets enterprises run AI workloads on infrastructure they own, whether on-premises or in a sovereign cloud operated within national boundaries [18].

What makes this more than a marketing claim is the engineering. VCF 9.1 introduced Private AI Model and GPU Metrics that give operators model-level visibility and utilization telemetry in the same console they use for the rest of the estate [19]. NVMe Memory Tiering extends effective memory pools by combining DRAM and NVMe for memory-bound AI workloads. Topology Aware Scheduling places workloads with GPU and NUMA locality in mind. And VCF supports multiple accelerator vendors: Enhanced DirectPath I/O provides near-bare-metal GPU passthrough for AMD Instinct MI350 GPUs [20], while VMware Private AI Foundation with NVIDIA integrates NVIDIA AI Enterprise for Blackwell-based deployments with vGPU, GPUDirect RDMA, and high-speed networking through ConnectX-7 and BlueField-3 NICs [21].

That multi-vendor support matters. Sovereign AI means choice at every layer, including the accelerator. An enterprise locked into a single GPU vendor's ecosystem has traded one dependency for another. VCF's ability to run AMD and NVIDIA accelerators on the same platform, managed through the same operational tooling, is a structural advantage that no hyperscaler offers, because hyperscalers have no incentive to give you that choice.

The sovereign cloud model, where VMware Cloud Service Providers operate VCF-based infrastructure within national boundaries, directly addresses the data residency requirements that Huang was pushing European leaders to take seriously. For regulated industries, this is not optional. It is the only architecture that satisfies both the AI ambition and the compliance reality.

The Uncomfortable Question

There is a tension at the center of Huang's European tour that he did not resolve, and probably could not. NVIDIA's business model depends on selling GPUs. The more AI factories Europe builds, the more GPUs NVIDIA sells. Sovereign AI is good strategy for Europe, and it is also very good business for NVIDIA.

That does not make the argument wrong. The structural risk of dependency on foreign AI platforms is real, regardless of who points it out. The investments in compute, startups, and research infrastructure are tangible. And the competitive gap between Europe and the US in AI is not a perception. According to multiple analyses, Europe ranks third globally in AI investment, behind the United States and China [6].

But it is worth noting that when Jensen Huang tells Europe to build more AI infrastructure, he is also telling Europe to buy more NVIDIA hardware. The best arguments are the ones where the speaker's interests and the audience's interests are perfectly aligned. This is one of those arguments. Whether Europe acts on it is another question.

Huang's message to European leaders was about nations. But the same logic applies at the organizational level. Your data, your models, your inference, your infrastructure. The alternative is dependency, and dependency is not a strategy.

Disclosure: The author is an employee of Broadcom. This article represents the author's technical analysis and does not represent Broadcom's official position.

References

[1] Jensen Huang Warns Europe: Move Faster on AI or Risk Falling Behind↗ - Business Insider [2] Jensen Huang London Speech↗ - LinkedIn [3] Jensen Huang Warns Germany to "Get to Work"↗ - YouTube [4] Mario Nawfal on Jensen Huang's Europe Tour↗ - X (Twitter) [5] NVIDIA's Jensen Huang Scolds Europe on AI Race↗ - Bloomberg [6] Jensen Huang Warns Europe Not to Overthink AI Rules↗ - Startup Fortune [7] Jensen Huang and Macron on Sovereign AI↗ - YouTube [8] Ericsson and Partners Launch Sferical One AI Supercomputer↗ - Ericsson [9] Jensen Huang Urges Europe's Industries to Become AI Companies at GTC Berlin↗ - Quartz [10] NVIDIA Huang Highlights AI Benefits for Europe↗ - Yahoo Finance / Reuters [11] Germany Wants to Lead the World in Industrial AI: How NVIDIA Will Help↗ - NVIDIA Blog [12] Jensen Huang Warns Europe: Move Faster on AI↗ - Business Insider [13] NVIDIA CEO Urges German Manufacturers to Adopt AI↗ - Broadband Breakfast [14] NVIDIA CEO Says It Invested Almost $10B Across Europe↗ - Newsquawk [15] GTC Berlin 2026 News↗ - NVIDIA Blog [16] Sweden's Sferical One Supercomputer Goes Live with 1,152 NVIDIA GPUs↗ - CryptoBriefing [17] NVIDIA CEO Jensen Huang Warns Germany Is Falling Behind in AI Race↗ - Euronews [18] VMware Private AI Foundation with NVIDIA in a Sovereign Cloud↗ - VMware Blog [19] VCF 9.1: Secure, Cost-Effective Private Cloud Platform for Production AI↗ - VMware Blog [20] AI with VCF 9.1 on AMD GPUs: Build with Open Frameworks and Simplify Management at a Lower TCO↗ - VMware Blog [21] VMware Private AI Foundation with NVIDIA at Explore25↗ - VMware Blog