NVIDIA is collaborating with Google Cloud to deliver agentic AI to enterprises searching for to domestically harness the Google Gemini household of AI fashions utilizing the NVIDIA Blackwell HGX and DGX platforms and NVIDIA Confidential Computing for information security.
With the NVIDIA Blackwell platform on Google Distributed Cloud, on-premises information facilities can keep aligned with regulatory necessities and information sovereignty legal guidelines by locking down entry to delicate info, corresponding to affected person data, monetary transactions and labeled authorities info. NVIDIA Confidential Computing additionally secures delicate code within the Gemini fashions from unauthorized entry and information leaks.
“By bringing our Gemini fashions on premises with NVIDIA Blackwell’s breakthrough efficiency and confidential computing capabilities, we’re enabling enterprises to unlock the total potential of agentic AI,” stated Sachin Gupta, vice chairman and common supervisor of infrastructure and options at Google Cloud. “This collaboration helps guarantee clients can innovate securely with out compromising on efficiency or operational ease.”
Confidential computing with NVIDIA Blackwell gives enterprises with the technical assurance that their consumer prompts to the Gemini fashions’ software programming interface — in addition to the information they used for fine-tuning — stay safe and can’t be seen or modified.
On the similar time, mannequin homeowners can shield towards unauthorized entry or tampering, offering dual-layer safety that permits enterprises to innovate with Gemini fashions whereas sustaining information privateness.
AI Brokers Driving New Enterprise Purposes
This new providing arrives as agentic AI is remodeling enterprise expertise, providing extra superior problem-solving capabilities.
In contrast to AI fashions that understand or generate primarily based on realized information, agentic AI techniques can motive, adapt and make selections in dynamic environments. For instance, in enterprise IT help, whereas a knowledge-based AI mannequin can retrieve and current troubleshooting guides, an agentic AI system can diagnose points, execute fixes and escalate complicated issues autonomously.
Equally, in finance, a conventional AI mannequin may flag doubtlessly fraudulent transactions primarily based on patterns, however an agentic AI system may go even additional by investigating anomalies and taking proactive measures corresponding to blocking transactions earlier than they happen or adjusting fraud detection guidelines in actual time.
The On-Premises Dilemma
Whereas many can already use the fashions with multimodal reasoning — integrating textual content, photographs, code and different information sorts to resolve complicated issues and construct cloud-based agentic AI purposes — these with stringent safety or information sovereignty necessities have but been unable to take action.
With this announcement, Google Cloud shall be one of many first cloud service suppliers to supply confidential computing capabilities to safe agentic AI workloads throughout each setting — whether or not cloud or hybrid.
Powered by the NVIDIA HGX B200 platform with Blackwell GPUs and NVIDIA Confidential Computing, this answer will allow clients to safeguard AI fashions and information. This lets customers obtain breakthrough efficiency and vitality effectivity with out compromising information safety or mannequin integrity.
AI Observability and Safety for Agentic AI
Scaling agentic AI in manufacturing requires sturdy observability and safety to make sure dependable efficiency and compliance.
Google Cloud immediately introduced a brand new GKE Inference Gateway constructed to optimize the deployment of AI inference workloads with superior routing and scalability. Integrating with NVIDIA Triton Inference Server and NVIDIA NeMo Guardrails, it affords clever load balancing that improves efficiency and reduces serving prices whereas enabling centralized mannequin safety and governance.
Trying forward, Google Cloud is working to reinforce observability for agentic AI workloads by integrating NVIDIA Dynamo, an open-source library constructed to serve and scale reasoning AI fashions throughout AI factories.
At Google Cloud Subsequent, attend NVIDIA’s particular tackle, discover classes, view demos and speak to NVIDIA specialists.
