Jensen Huang sees a ‘major second wave’ coming for AI

Silicon Valley might have started the generative AI boom, but Jensen Huang thinks the next act won’t belong exclusively to California.

Speaking virtually at Scaleway's ai-PULSE conference in Paris, the NVIDIA CEO declared that the AI industry is entering a major second wave. According to Huang, this phase is driven by two big forces: the global race to build sovereign AI, and the migration of AI from chatbots into heavy industries like manufacturing and biology.

Sovereign clouds meet European industry

Huang pointed out that countries and regions are realizing they need to control their own data and train models that understand their local cultural and industrial nuances.

Europe, in particular, sits in a sweet spot. While it lagged behind in early consumer chatbots, the continent holds deep expertise in digital biology, healthcare, and industrial manufacturing. Breakthroughs seen in large language models will soon translate directly into robotics and drug discovery.

NVIDIA is already entrenched in the region, working with 4,000 European startups, including over 400 in France like Mistral, Qubit Pharmaceuticals, and Poolside AI. But algorithms mean little without silicon, which is where local cloud providers come in.

Fueling the regional supercomputer

To power this shift, iliad Group subsidiary Scaleway launched Nabu, an NVIDIA DGX SuperPOD packing 127 DGX H100 systems for a total of 1,016 H100 GPUs.

The system is built explicitly to comply with stringent EU data protection regulations, giving European enterprises an alternative to routing sensitive data through US cloud giants. Scaleway is also offering cloud credits to startups to get them building immediately.

This is part of a broader European infrastructure push. Germany's upcoming JUPITER exascale supercomputer will pack 24,000 GH200 Grace Hopper chips, and NVIDIA's recently unveiled H200 GPU is scheduled to roll out in Q2 2024.

Why it matters

This marks a decisive pivot from centralized AI to regionalized infrastructure.

For European enterprises and startups, sovereign setups like Scaleway’s H100 cluster solve their biggest headache: compliance. Companies working in healthcare or manufacturing simply cannot ship proprietary data to overseas data centers under GDPR. Having compliant, top-tier compute close to home makes serious local model training actually viable.

For the broader industry, promoting sovereign AI is a masterclass in market expansion by NVIDIA. By convincing nation-states and regional telecom giants that compute independence is national security, NVIDIA creates massive new GPU demand outside the typical hyperscaler buyers.

We think the play is brilliant commercially, though it comes with real execution risks. Subsidized cloud credits only last so long, and European startups will eventually have to prove their specialized models can generate enough revenue to justify eye-watering GPU rental rates. Still, the era of relying solely on US-based AI infrastructure is clearly ending.