Google Cloud Next 25: partner go-to-market refined, agentic AI marketplace launched

Google Cloud has honed its partner go-to-market approach and has launched a new marketplace for AI agents amid a bevy of other announcements at its ‘Next 25 conference.

Outlined in a blog post penned by the vendor’s president of its global partner organisation, Kevin Ichhpurani, partners are set to see new processes to capture and share their contributions with Google Cloud’s sales team.

This includes increased visibility into co-selling activities such as workshops, assessments and proofs of concept in addition to migration, application modernisation and other managed services by partners.

“Ultimately, this information will better enable our sales team to connect customers with the correct ISV and services partners,” Ichhpurani said.

On the earnings side, the blog post said Google Cloud is introducing AI-powered capabilities in the Earnings Hub to benchmark performance against similar partners, receive personalised tips on how to boost earnings and warn when customers change their consumption patterns, among other features.

Additionally, the vendor said it would expand its partner training efforts after having invested more than $100 million in it over the last four years, with Google Agentspace and Workspace listed as two “critical areas” for knowledge development.

Agentic AI

Aside from Google Cloud’s partner go-to-market updates, most of the vendor’s new updates revolved around AI – particularly agentic AI.

Among these announcements was the launch of an AI Agent Marketplace section for Google Cloud Marketplace to enable customers to buy and manage AI agents built by partners.

Agents from Accenture, BigCommerce, Deloitte, Elastic, UiPath, Typeface and VMware are available on launch, with solutions from Cognizant, Slalom and Wipro coming soon.

Google Cloud also launched the preview of Agent2Agent, a new open protocol for AI agents to communicate with each other, exchange information securely and coordinate actions across a number of enterprise platforms and services, like Atlassian, Box, Cohere, Intuit, Langchain, MongoDB, Salesforce, SAP, ServiceNow, UKG and Workday.

“We believe the A2A framework will add significant value for customers, whose AI agents will now be able to work across their entire enterprise application estates. More partners can begin building with the A2A framework today,” Ichhpurani said.

A number of service partners contributed to A2A, including Accenture, BCG, Capgemini, Cognizant, Deloitte, HCLTech, Infosys, KPMG, McKinsey, PwC, TCS, and Wipro, and will support its implementation.

These partners have expanded their Google Cloud practices “significantly” over the last year with new experts and technical resources, Ichhpurani continued.

“This means our customers now have access to a global community of highly-trained AI experts who can help them develop AI agent use cases and strategies with interoperability in mind, prototype new applications, train and manage AI models and ultimately deploy AI agents across their businesses,” he added.

Also on the cards were new AI features for Workspace, which includes Google Workspace Flows – an AI-powered multi-step processing agent that utilises Gems – custom AI agents built using Gemini to take on specific tasks.

Other Workspace announcements were focused on the generation of audio overviews and the writing refinement tool Help me refine for Google Docs, proactive analysis and visualisation capabilities for Sheets, generating video content in Google Vids and enabling Gemini to provide summaries and answer queries in Google Meet.

Google’s Vertex AI platform is receiving multiple updates, one of which includes the open-source Agent Development Kit (ADK) for designing agents built on the same framework that the vendor uses for its Agentspace and Customer Engagement Suite (CES) agents.

Within the ADK is the preview launch of Agent Garden – a collection of pre-made samples and tools to speed up development processes.

Additionally, Agent Engine – a managed runtime for agents – was announced to have entered general availability, while Grounding with Google Maps had a preview launch, enabling agents to provide responses related to geospatial information, but only for US locations.

Vertex AI is also set to see Gemini 2.5 Flash, which features dymamic and controllable reasoning like Gemini 2.5 Pro and automatically adjusts processing time, or thinking budget, based on query complexity to enable faster answers for simpler requests. Users are given granular control over the budget for explicit tuning of the speed, accuracy and cost balance for specific needs.

That flexibility is considered key to optimising Flash performance in high-volume, cost-sensitive applications, according to a blog post penned by Vertex AI director of product management Jason Gelman and Google DeepMind director of product management Tulsee Doshi.

Various generative media models in Vertex AI are also being launched, with the text-to-speech Lyria in preview with allowlist, editing and camera controls for video generation model Veo 2 in preview with allowlist, instant custom voice for Chirp 3 and improved image reconstruction capabilities for image-to-text model Imagen 3.

On the AI Hypercomputer side, Google Cloud announced its seventh-generation Tensor Processing Unit (TPU) Ironwood, claiming it to be the vendor’s most performant and scalable custom AI accelerator to date and the first designed for inference.

It has also added new inference capabilities in Google Kubernetes Engine (GKE), including generative AI aware scaling and load balancing features, and Pathways, the vendor’s own internal machine learning runtime developed by Google DeepMind.

Other announcements include adding the preview of Gemini availability to Google Distributed cloud for on-premises environments, Cloud WAN – a managed enterprise backbone to transform enterprise wide area network (WAN) architectures with Google’s planet-scale network.

Announcement also includes the next generation of its Customer Engagement Suite, which includes human-like voices, comprehension and understanding of emotions for agents to adapt better during conversations and purpose-built vertical agents for specific needs like food ordering, automotive and retail.

Google Cloud also released several security advancements at ‘Next 25, emphasising the integration of AI and agents to enhance threat detection and response.