Google / Elyse Betters Picaro / ZDNET
Follow ZDNET: Add us as a preferred source on Google.
ZDNET's key takeaways
Gemini on Google Distributed Cloud is now available to customers.
The approach brings advanced models into enterprise data centers.
Gemini on GDC could support new capabilities for on-premise gen AI.
There are several obstacles to the successful deployment of artificial intelligence (AI) in the enterprise, including managing staff who are unsure how to use the technology, and cleaning and organizing the information that feeds AI services.
A boost to enabling corporate AI
Google has announced what it expects will be a boost to enabling corporate AI, with the company turning on the on-premise version of its Gemini family of large language model AI programs, as provided by its Google Distributed Cloud (GDC) on-premise offering.
Also: Forget plug-and-play AI: Here's what successful AI projects do differently
The announcement is a follow-up to the initial unveiling of on-premise Gemini that Alphabet made in April: "We are excited to announce that Gemini on GDC is now available to customers," said the company, "bringing Google's most advanced models directly into your data center."
Also: Google makes Gemini Pro available in AI Studio, Vertex AI tools
Google referred to prominent early customers for on-premise Gemini, including Singapore's Centre for Strategic Infocomm Technologies (CSIT), Government Technology Agency of Singapore (GovTech Singapore), Home Team Science and Technology Agency (HTX), KDDI, and Liquid C2.
New capabilities for on-prem use
Google's announcement suggested new capabilities for on-premise use of generative AI, including:
Language translation for large enterprises.
Fast decision-making with tools such as document analysis.
24/7 support for customers via chatbots.
Faster internal software development with Gemini code automation.
Safety measures via automatic filtering of "harmful content," and enforcing compliance measures.
The GDC offering includes several elements that work in concert with Gemini, including: Google's agentic AI framework, Agentspace; its managed programming tool for enterprises, Vertex AI; Google's open-source AI model family, Gemma; task-specific AI models; and all the Google Cloud hardware, such as Nvidia Blackwell 300 data center GPUs.
Also: First Gemini, now Gemma: Google's new, open AI models target developers
On the last point, Google emphasized its ability to manage on-premise infrastructure: "A fully managed Gemini endpoint is available within a customer or partner data center, featuring a seamless, zero-touch update experience. High performance and availability are maintained through automatic load balancing and auto-scaling of the Gemini endpoint, which is handled by our L7 load balancer and advanced fleet management capabilities."
Security measures include Intel's microprocessors that have "TDX" capability turned on, and Nvidia GPUs that have what Nvidia calls "confidential computing."
Also: This AI cloud: How Google Gemini will help everyone build things faster, cheaper, better
The announcement is replete with quotes from early customers, including Toru Maruta, the head of advancing business platform at Japanese telecom giant KDDI, who stated that the GDC offering "will bring cutting-edge AI capabilities, meet specific performance requirements, and address data locality and regulatory needs of Japanese businesses and consumers."
Sovereign AI trend
Google's offering will likely be an important element of what Nvidia CEO Jensen Huang has called the "sovereign AI" trend, where governments want specialized regional infrastructure that is not part of the public internet to run AI models. Huang has described sovereign AI as "a new growth engine for Nvidia."
Google has already shown a propensity to invest in creating independent regional cloud instances for whole countries.