Secure Agent AI: Google & NVIDIA Unite

In this week’s episode of “AI is Changing Everything (Again),” NVIDIA and Google Cloud have joined forces to deliver agentic AI capabilities to enterprise customers. This partnership will let businesses run Google Gemini AI models on their own systems using NVIDIA’s powerful Blackwell HGX and DGX platforms.

The collaboration also incorporates NVIDIA Confidential Computing to keep company data safe and secure while organizations deploy these advanced AI solutions in-house.

“By bringing our Gemini models on premises with NVIDIA Blackwell’s breakthrough performance and confidential computing capabilities, we’re enabling enterprises to unlock the full potential of agentic AI,” said Sachin Gupta, vice president and general manager of infrastructure and solutions at Google Cloud. “This collaboration helps ensure customers can innovate securely without compromising on performance or operational ease.”

NVIDIA Blackwell’s confidential computing offers businesses peace of mind by ensuring that their interactions with Gemini models remain private. This guarantees that prompts and any data used to customize these models cannot be accessed or tampered with by unauthorized parties.

At the same time, companies that own these AI models can prevent unauthorized access, creating a double layer of protection. This security setup lets businesses innovate with Gemini models while keeping their data private.

AI agents: The next wave of business technology

AI agents are already shaking up how businesses use technology by solving more complex problems. Unlike generative AI tools that just respond based on what they’ve learned, AI agents can think through problems, adapt to new situations, and make decisions as things change.

For example, in IT support:

• A standard AI might just show you troubleshooting guides
• An AI agent can actually diagnose what’s wrong, fix issues itself, and call for help with complicated problems

In finance:

• A regular AI might flag suspicious transactions based on patterns
• An AI agent can investigate unusual activity and take action—like blocking transactions before they happen or updating fraud detection rules on the fly

Breaking barriers for regulated industries

While cloud-based multimodal AI has transformed the operations of many businesses, organizations with strict security requirements or data sovereignty concerns have largely been unable to participate. Healthcare providers, financial institutions, and government agencies- often those with the most to gain from AI- have faced an almost impossible choice between innovation and compliance.

As a result, the NVIDIA and Google Cloud partnership could represent a critical turning point for regulated industries. The ability to run sophisticated AI models on-premises while maintaining full data control effectively eliminates the main barrier to adoption.

NVIDIA’s new Blackwell chips feature built-in privacy mechanisms that allow you to use Google’s Gemini AI safely on their HGX B200 systems, which is good news for anyone concerned about securing AI. Google Cloud also launched a more innovative method for deploying AI, called GKE Inference Gateway. It streamlines AI operations and integrates well with NVIDIA’s tools. 

As AI evolves into systems that can reason and make decisions independently (i.e., agentic AI), Google Cloud is collaborating with NVIDIA Dynamo to enhance our ability to monitor and comprehend them. This development makes these AI technologies more trustworthy and usable for companies everywhere.

For MSPs, this means a chance to offer clients secure access to powerful AI like Google’s Gemini on NVIDIA’s platform. To do this well, though, MSPs will need to develop expertise in these advanced AI and security technologies. 

IT giant Cisco also recently announced a new agentic AI-powered collaboration solution that will create “future-proofed” experiences for customers, employees, and IT teams.