Google Cloud Platform adds WAN and on-premises AI services
A handful of infrastructure announcements for the Google Cloud Platform debut at Next 2025 including a Cloud WAN service, on-premises Gemini AI and new AI chips.
Advancements for Google's Gemini AI model family took center stage at the hyperscaler's Google Cloud Next 2025 show, but a few new infrastructure capabilities are arriving to Google Cloud Platform.
New services and capabilities previewed at Next include an on-premises version of Gemini; Cloud WAN, a managed networking service; and a new generation of Google's Tensor Processing Unit (TPU) AI accelerator chips.
The breadth and depth of its AI services gives Google an advantage against its chief cloud competitors, AWS and Microsoft Azure, according to Chirag Dekate, an analyst at Gartner. But AI might not be as pressing a purchase concern for IT executives who may have other market challenges, like tariffs, on their minds.
"All three cloud providers are now leading with an AI-forward message," Dekate said. "But this year, cost efficiency and other vectors dominate [executive] mind share."
AI on-premises
Google Distributed Cloud, Google's fully managed on-premise cloud appliance, will include Gemini AI models for GDC units with the Nvidia Blackwell GPU, according to the company.
GDC provides the entire Google Cloud Platform infrastructure stack air-gapped and sold through various channel partners. The appliances previously supported Vertex AI, a collection of pretrained models for various tasks, but Gemini on GDC brings generative AI capabilities to highly regulated industries.
The new release will also support an on-premises version of Google Agentspace, a multi-modal AI agent service that acts as a chatbot search for enterprise data.
Until now, enterprises were locked out of first-party frontier models.
Chirag Dekate Gartner, Analyst
Privately owned and developed GenAI models by companies like OpenAI and Anthropic are typically limited to just cloud access, Dekate said. Google's release of Gemini on GDC is among the few non-open source LLMs available for enterprise purchase.
"Until now, enterprises were locked out of first-party frontier models," he said.
A public preview for Gemini and Agentspace on GDC will start in the third quarter of this year, the company said.
Later this year, Google's AI services in the cloud will be available via access to Ironwood, the latest version of Google's AI accelerator chips.
Ironwood will be available in two sizes: a 256-chip configuration and a 9,216-chip configuration. The chip should achieve double the performance of Trillium, Google's TPU released last year, according to the vendor. Customers will be able to link tens of thousands of Ironwood TPUs together for more demanding AI application development.
Ironwood is optimized for the Google Pathways software stack, a large language model developed by Google.
Ironwood will be available later this calendar year.
Google networking
Cloud WAN, available today according to Google, provides a fully managed WAN services for enterprises using the same network developed, used and maintained by Google.
Specific Cloud WAN use cases highlighted by Google include connectivity for data centers or migration services for edge and branch networks. Cloud WAN will have an open ecosystem with numerous networking partners, including Cisco, Fortinet, Juniper Networks, Netskope, VeloCloud and others.
The service offers dedicated, private connections to Google Cloud and will enable cross-cloud connectivity with other public cloud services, including AWS, Oracle Cloud Infrastructure and Azure. All three also offer their own WAN services.
Tim McCarthy is a news writer for Informa TechTarget, covering cloud and data storage.
