On September 15, 2026, at Dreamforce 2026 in San Francisco, Salesforce and Google Cloud announced an expanded strategic partnership to build a "unified AI stack" enabling cross-platform agent reasoning and action on shared infrastructure. Simultaneously, Salesforce deepened integrations with AWS, signaling a decisive bet on open ecosystems over walled gardens.
For businesses evaluating AI agent platforms, this partnership reshapes the decision. The question is no longer which platform do I choose? but which ecosystem do I join?
What Was Announced
The partnership, detailed in a Google Cloud press release and reported by SiliconANGLE and Pulse2, covers four major integration layers:
| Integration | What It Does | Status |
|---|---|---|
| Hyperforce on Google Cloud | Runs Salesforce workloads natively on Google Cloud infrastructure | Live customer traffic now; broader migration Q4 2026 |
| Gemini Enterprise + Salesforce | Surfaces CRM data inside Gemini via headless architecture (MCP-based) | Announced; rolling out |
| Gemini inside Agentforce | Lets Salesforce customers build agents powered by Google's Gemini models | Announced |
| Commerce Cloud + Universal Commerce Protocol | Lets shoppers discover and buy products from within Google Search or Gemini | Announced |
Salesforce also announced parallel AWS integrations: Salesforce business context available natively in Amazon Quick via the Model Context Protocol (MCP), Agentforce access to Amazon Bedrock's model library, Data 360 Zero-Copy integration with AWS Glue-managed Apache Iceberg tables, and AWS DevOps Agent embedded in Slack.
Why This Matters: The MCP Revolution
The technical backbone of both integrations is the Model Context Protocol (MCP), an open standard that lets AI agents securely access external systems and data. Salesforce's headless architecture, built on MCP, means its capabilities can be surfaced inside any AI model — Gemini, Claude, or Amazon's models — without custom integration work for each.
This is the shift that matters for growing businesses. If your CRM, your analytics, your commerce platform, and your AI agent all speak the same protocol, you are not locked into one vendor's model. You can swap the underlying AI engine without rebuilding your agent infrastructure.
As Kevin Ichhpurani, President of Google Cloud's Global Partner Ecosystem, stated: "By bringing Salesforce into Gemini Enterprise, customers can now build trusted agents that reason across their business context and securely take action."
The Open vs. Walled Garden Decision
Before this announcement, the enterprise AI agent landscape required companies to make a platform bet: Salesforce Agentforce, Microsoft Copilot, or Google Gemini. Each had proprietary integration layers, and moving between them meant rebuilding agent workflows.
The Salesforce + Google Cloud partnership changes the calculus:
Before: Walled Gardens
- Agentforce worked best with Salesforce data, Salesforce workflows, and Salesforce-connected AI models.
- Gemini Enterprise worked best with Google Cloud data and Google-native agents.
- Combining the two required expensive custom integration.
After: Connected Ecosystem
- Salesforce data and workflows are accessible inside Gemini Enterprise via MCP.
- Gemini models can power Agentforce agents.
- Salesforce runs natively on Google Cloud infrastructure via Hyperforce.
- Agents from different platforms can collaborate through shared protocols.
This does not eliminate platform competition — Salesforce, Google, and AWS are still competing for the primary AI workload. But it gives customers meaningful portability and choice.
What It Means for SMBs: 3 Takeaways
1. Don't Lock In — Protocol-Native First
If you're building AI agents for your business, prioritize tools that support MCP and open agent protocols over proprietary platforms. The companies that built on MCP-compatible infrastructure today can adopt new models and platforms tomorrow without rebuilding. Those that locked into a single vendor's proprietary stack will face migration costs.
2. Your Data Foundation Matters More Than Your Model
The partners emphasized data gravity: Salesforce's value is its CRM context, Google's value is its infrastructure and Gemini's reasoning. Neither is useful without the other. For SMBs, this means investing in clean, structured, well-governed data is more valuable than choosing the "right" AI model — because models are becoming interchangeable, but data is not.
3. Commerce Is Going Agentic — Prepare Your Catalog
The Commerce Cloud integration with Google's Universal Commerce Protocol means products can be discovered and purchased directly from Google Search or Gemini, with payments processed by Salesforce. If you sell products or services online, you need machine-readable product data, clean APIs, and structured commerce endpoints — because AI agents are becoming a new sales channel.
The Competitive Landscape After Dreamforce 2026
Dreamforce also introduced AIforce, Salesforce's new AI interface layer, alongside a partnership with NVIDIA on the Koa reasoning model. The expanded partnership is a response to intensifying competition:
| Platform | Strategic Position | Agent Infrastructure |
|---|---|---|
| Salesforce + Google Cloud + AWS | Open ecosystem, MCP-native | Agentforce + Gemini + Bedrock + Amazon Quick |
| Microsoft Copilot | Deep Microsoft 365 integration | Copilot Studio + Power Platform |
| ServiceNow AI Agents | Enterprise IT and service management | Now Assist + specialized domain agents |
| Zendesk Specialized AI Agents | Purpose-built customer experience agents | Industry Agents + Custom Agents |
The pattern is clear: the winners are not the platforms with the best single model — they are the ones with the richest ecosystem of interconnected agents, data, and workflows across multiple clouds.
The NVIDIA Partnership: Koa Reasoning Model
Alongside the Google Cloud and AWS expansions, Salesforce announced a partnership with NVIDIA to develop the Koa reasoning model, reported by TechEDT on September 16, 2026. Koa is designed to target CRM-specific reasoning tasks — pipeline health analysis, account summarization, and case triage — rather than competing as a general-purpose foundation model. This vertical specialization mirrors the broader industry trend: the winning AI tools are not the most general, but the most specifically tuned to a high-value domain."
Read next from the blog : Where Do AI Assistants Get Their Supplier Recommendations? The Directory Discovery · Profound's $180M Raise: AI Search Visibility Becomes a Real Marketing Category · Zendesk Specialized AI Agents: The End of Generic Customer Service Automation · The Enterprise AI Agent Platform Decision: Copilot Studio, Agentforce, or Gemini?
