On April 9, 2025, at Google Cloud Next 2025, Google Cloud announced a major collaboration combining its Google Distributed Cloud (GDC) with the Gemini AI model and NVIDIA's Blackwell architecture to deliver powerful, on-premises AI solutions for enterprises. This collaboration aims to address the stringent data sovereignty and security needs of various industries, driving the adoption of "Agentic AI" technology in on-premise environments.
Google Distributed Cloud will now support running the Gemini model within enterprise data centers. This is achieved through a partnership with NVIDIA, leveraging their newly launched Blackwell GPU system for high-performance computing. Dell, as a key partner, provides hardware support, ensuring enterprises can enjoy the flexibility of the public cloud while maintaining complete control over their data on-premises. Notably, this solution supports both networked and completely isolated "air-gapped" scenarios, ideal for government agencies, highly regulated industries, and enterprises with specific latency and data residency requirements.
A highlight of this collaboration is the integration of NVIDIA's Confidential Computing technology, ensuring end-to-end protection of data and prompts even from the cloud provider when using the Gemini model to process sensitive information. This combination of security and performance is considered a key step in unlocking the potential of on-premises AI. Sachin Gupta, VP and GM, Infrastructure and Solutions, Google Cloud, stated: "By combining the Gemini model with the groundbreaking performance and confidential computing capabilities of NVIDIA Blackwell, we're empowering enterprises to innovate securely without compromising on performance or ease of operation."
Furthermore, Google Distributed Cloud plans to launch GKE Inference Gateway, a tool integrated with NVIDIA Triton Inference Server and NeMo Guardrails, to optimize inference routing and load balancing, helping enterprises manage and scale AI workloads more efficiently. This feature is expected to enter public preview in Q3 2025, offering more businesses the opportunity to try it out.
Industry experts believe this collaboration marks a significant shift in AI deployment models. Previously, many enterprises couldn't fully leverage cutting-edge AI due to cloud-based deployments and security concerns. The Google and NVIDIA joint solution enables enterprises to run complex AI agents on-premises – agents that not only understand data but also reason, act, and self-optimize. This trend is seen as a crucial step towards "self-correcting" and "self-improving" enterprise AI systems.
The Google Cloud and NVIDIA collaboration extends beyond technology, reflecting a shared vision of democratizing AI. By bringing the Gemini model on-premises and combining it with the powerful performance of the Blackwell architecture, this solution promises to unlock new growth opportunities for industries like finance, healthcare, and manufacturing, while meeting stringent compliance requirements. As more details emerge and real-world applications are deployed, this collaboration may reshape the landscape of enterprise AI deployment.
