Nvidia CEO encourages Europe's lagging AI sector and offers tools to accelerate data center buildout - The Register-Herald

Nvidia's "toolkit" push into Europe is an ecosystem lock-in play, not philanthropy. The strategic calculus: embed CUDA and NVLink into European data center foundations before the EU AI Act crystallizes into rigid procurement rules. Once hardware selection locks in, switching costs become exponential—this is the 2016 CUDA-academic-lock-in playbook replayed at sovereign scale, with nation-states replacing universities as the captive audience. Upstream, this accelerates TSMC (Taiwan, China) CoWoS capacity and SK Hynix HBM3E demand. But the deeper tension: if the "tools" include network orchestration software, Europe's AI infrastructure control plane gets colonized by a US software stack—structurally at odds with Brussels' technological sovereignty narrative. On compliance, AI Act data-provenance mandates will inflate BOM costs by 8-12%. Paradoxically, GDPR cross-border restrictions favor localized deployment, eroding hyperscaler economies of scale. Competitively, AMD will weaponize MI350 price-performance and a "non-US-controlled" narrative for sovereign projects; Intel will contest mid-tier with Gaudi 3. But Nvidia's moat is developer inertia—90% of European AI startups train on PyTorch+CUDA. That migration cost is near-insurmountable within 24 months. 12-24 month outlook: Europe bifurcates into a dual-track market. Commercial cloud goes to Nvidia; sovereign projects spawn regional custom stacks. Nvidia's European revenue share should climb from under 15% to 25%+, contingent on convincing Brussels that American silicon doesn't equal American data control.