Dive Brief:
- Google Cloud’s enterprise customers can now access Gemini via Google Distributed Cloud, its on-premises and edge cloud solution, the company said Thursday.
- The tech giant positioned the move as a security play. “For years, enterprises and governments with the strictest data security and sovereignty requirements have faced a difficult choice: adopt modern AI or protect their data,” Google said in the announcement. “Today, that compromise ends.”
- The company partnered with Nvidia in April to bring its Gemini models to the chipmaker’s Blackwell systems as it looked to enhance the privacy and security of GDC. Google’s Agentspace, a service for pre-built AI agents, remains in preview.
Dive Insight:
Enterprises have concerns about keeping data and intellectual property secure when using AI. On-premises deployments offer a type of oasis.
“Bottom line, it’s about empowering enterprises everywhere to tap leading frontier models that deliver best-in-class capabilities in a controlled, secure, compartmentalized, sovereign manner,” Chirag Dekate, VP analyst at Gartner, told CIO Dive.
The on-premises AI market is still evolving.
“OpenAI currently does not have this approach, and Anthropic also does not have this sort of delivery approach,” Dekate said. “It’s not very clear whether OpenAI or Anthropic has the incentive to address the on-prem opportunity.”
Vendors see on-premises as a potential differentiator in the enterprise market as more consumer-focused AI companies capture other experiences, such as through apps or online access.
Demand for on-premises access to frontier models is growing alongside data privacy concerns.
“Enterprises have some workloads and data assets that will never move to the cloud,” Dekate said, citing compliance, security and data sovereignty considerations for highly-regulated industries and government ecosystems.
On-premises data centers are starting to see a boost from AI and GPU deployments, and Nvidia has had a hand in supporting a large chunk of those workloads. In just one quarter, large cloud providers added an average of roughly 72,000 Nvidia Blackwell GPUs per week, executives said during the company’s Q1 2026 earnings call.
“It’s really hard to move every company’s data into the cloud, so we’re going to move AI into the enterprise,” CEO Jensen Huang said to investors in May. “We’re going to see AI go into enterprise, which is on-prem, because so much of the data is still on-prem.”
PC manufacturers are also helping to bolster the availability of on-premises AI.
In May, Dell executives touted expanded partnerships with Google and Cohere as it looks to accelerate customers’ on-premises AI deployments. Dell had more than $12 billion in AI orders during Q1 2026.
