Salesforce is putting its core infrastructure on a rival’s servers. At Dreamforce 2026 in San Francisco on September 15, Salesforce and Google Cloud announced an expanded partnership that moves Hyperforce, the architecture underpinning Salesforce’s entire CRM platform, onto Google Cloud infrastructure, with agents from both companies able to read and act on the same enterprise data without custom integration work. General availability for North America is set for November 2026, with select US customers migrating as early as Q4 2026.
The move is bigger than a routine cloud partnership renewal. Hyperforce launched on Amazon Web Services and has stayed there as its primary home for six years. Bringing Google Cloud in as a first-class host, not just an API integration partner, signals that even the biggest SaaS vendors no longer want to bet their infrastructure on a single provider. It also ties Salesforce’s Agentforce platform directly into Google’s Gemini Enterprise, a pairing that puts pressure on both AWS and Microsoft Azure at the exact moment enterprise AI agent spending is accelerating.
What Salesforce and Google Cloud Actually Announced
The centerpiece of the announcement is Hyperforce running natively on Google Cloud. Hyperforce is the infrastructure layer Salesforce built to decouple its software from any single data center footprint, letting the company deploy on public cloud regions instead of Salesforce-owned facilities. Until now, AWS carried the bulk of that load. Salesforce and Google Cloud confirmed that Hyperforce on Google Cloud is already serving live production traffic for early customers, with a broader wave of select US customers migrating in the fourth quarter of 2026 and full North America general availability targeted for November 2026. Germany is named as the next region in 2027, with additional countries to follow as part of what the companies describe as a progressive rollout.
Alongside the infrastructure shift, Salesforce connected its headless architecture, the decoupled front-end and back-end model it has been building toward for several years, to Gemini Enterprise. The practical result: an AI agent built on Gemini Enterprise can pull a Salesforce account history, read Tableau analytics, and update a CRM opportunity, all without a developer writing custom middleware to bridge the two platforms. Salesforce also confirmed that Gemini models are now available as a model option inside Agentforce, giving customers a choice beyond Salesforce’s own models and other partner LLMs.
Data plumbing got attention too. The companies expanded “Zero Copy” data sharing between Salesforce Data Cloud (branded Data 360) and Google BigQuery, letting each platform query the other’s data without duplicating it first. Tableau analytics are exposed to Gemini Enterprise workflows through a Tableau MCP (Model Context Protocol) connector, so a Gemini-based agent can pull a chart or a metric straight out of Tableau mid-conversation. On the commerce side, Salesforce Commerce Cloud will start powering shopping experiences inside Google Search and the Gemini app this fall, a smaller piece of the announcement but one that puts Salesforce merchant data directly in front of Google’s consumer search traffic.
Why Salesforce Is Diversifying Away From a Single Cloud
Salesforce has not abandoned AWS. Reporting around the Dreamforce announcement makes clear that Salesforce is simultaneously deepening its AWS integrations even as it brings Google Cloud on board, and AWS remains described as Hyperforce’s first and still-primary cloud. What is changing is the assumption that Salesforce runs on one dominant cloud with everything else treated as a bolt-on integration.
Three forces are pushing that shift. First, enterprise customers increasingly standardize on a specific cloud provider for regulatory, procurement, or existing-infrastructure reasons, and a Salesforce that only runs cleanly on AWS leaves money on the table with Google Cloud-native accounts. Second, the AI agent era rewards proximity between compute and data: an enterprise running its data warehouse on BigQuery wants its CRM agents reasoning over that data with minimal latency and no brittle sync jobs. Third, the Zero Copy data-sharing layer between Data Cloud and BigQuery only pays off if Salesforce workloads actually sit near Google’s infrastructure rather than reaching across the public internet for every query.
Meir Amiel, President and Chief Trust and Infrastructure Officer at Salesforce, framed the move as a change beneath the product rather than a change to what customers get. “What changes here is the infrastructure underneath our customers, not the trust they’ve built with us. Hyperforce running natively on Google Cloud will extend the same security, compliance, and resilience standards we hold ourselves to. Enterprises will get Google Cloud’s flexibility without giving up an inch of the trust bar they expect from Salesforce,” Amiel said in the joint announcement.
Inside the Agentforce and Gemini Enterprise Integration
The technical shape of the integration matters more than the press-release language suggests. Salesforce and Google Cloud are not simply letting two chatbots talk to each other. They are letting agents on either platform execute actions, not just retrieve information, against the other platform’s business logic. A Gemini Enterprise agent working a sales pipeline can now trigger a Salesforce workflow, create a case, or update an opportunity stage in real time, without the request bouncing through a middleware layer that a developer has to maintain and patch.
This builds on groundwork Salesforce and Google Cloud laid earlier in 2026. An April 22 integration introduced Agentforce Sales inside Gemini Enterprise, letting sales agents engage leads, draft meeting briefs, flag at-risk deals, and update pipeline records without leaving the Gemini Enterprise interface. The September expansion generalizes that pattern across more of the Agentforce and Gemini Enterprise stack, and adds the option to run Agentforce agents on Gemini models directly rather than only calling into Gemini Enterprise as an external system.
Google Cloud’s own messaging positions this as part of a broader agent marketplace strategy. The company’s Gemini Enterprise blog describes “an open partner ecosystem to discover and deploy a wide range of third-party agents from leaders like Oracle, Salesforce, and ServiceNow, all within a secure and governed environment,” according to Google Cloud’s official blog. Elsewhere, Google Cloud characterized the Salesforce tie-up specifically: “Our expanded partnership with Salesforce brings more of the enterprise AI stack together, with Gemini Enterprise and Salesforce’s agentic capabilities working across trusted cloud infrastructure,” the company wrote in its blog post announcing partner-built agents in Gemini Enterprise.
One thing is absent from the public disclosures: a dollar figure. Neither company published a contract value, a minimum spend commitment, or a term length. That is typical for hyperscaler-to-SaaS infrastructure deals, where the money moves through committed cloud spend rather than a headline number, but it also means outside analysts are estimating the deal’s scale from customer migration counts rather than disclosed figures.
Cloud Market Context: Where AWS, Azure, and Google Cloud Stand
The timing lines up with a broader shift in cloud market share that has been building through 2026. AWS has been ceding percentage points off its historical dominance while Google Cloud has posted the fastest growth rate of the three hyperscalers for several consecutive quarters. Landing Salesforce as a native Hyperforce host, even as a second option alongside AWS, gives Google Cloud a marquee enterprise SaaS win that AWS cannot easily counter without matching Salesforce’s own AI agent roadmap feature for feature.
| Cloud Provider | Role in Salesforce Hyperforce | Primary AI Agent Platform | Enterprise SaaS Hosting Angle |
|---|---|---|---|
| AWS | Original and still-primary Hyperforce host since launch | Amazon Bedrock (hosts OpenAI, Anthropic, Meta models) | Deepening AI agent integrations alongside the new Google Cloud option |
| Google Cloud | New native Hyperforce host, GA November 2026 in North America | Gemini Enterprise, 200+ foundation models via Agent Platform | First hyperscaler to host core Salesforce infrastructure beyond AWS |
| Microsoft Azure | Not part of Hyperforce hosting | Microsoft Foundry, Copilot, Dynamics 365 | Relationship centers on Slack, Teams, and Office integrations, not core CRM hosting |
Microsoft’s position in this story is the most interesting by omission. Azure is not mentioned as a Hyperforce host in any of the Dreamforce disclosures, and Salesforce’s relationship with Microsoft has historically run through Slack, Teams, and Office 365 connectors rather than infrastructure hosting. That leaves Microsoft’s enterprise AI answer, the combination of Copilot, Dynamics 365, and Azure AI Foundry, competing against a Salesforce-Google Cloud pairing that now spans infrastructure, data, and agents in one stack. For customers who have already standardized on Dynamics or Copilot, that gap matters less. For the large pool of Salesforce customers evaluating which cloud to consolidate their AI agent spend on, Microsoft is effectively out of the conversation on this specific deal.
Historical Context: Hyperforce’s Six-Year Path to Multicloud
Hyperforce was never designed as an AWS-exclusive architecture. Salesforce built it specifically to decouple its software from proprietary data centers and give the company region-by-region flexibility for data residency and compliance. In practice, AWS became the default and, for most of Hyperforce’s life, the only major public cloud actually running Salesforce’s core workloads at scale. That made Salesforce’s own multicloud story mostly theoretical for customers who cared which hyperscaler their CRM actually touched.
The shift toward Google Cloud did not happen overnight. Salesforce and Google Cloud had already been layering integrations for years, from BigQuery connectors to earlier versions of Data Cloud interoperability. The April 2026 Agentforce Sales integration inside Gemini Enterprise was a clear signal that the two companies were building toward something more structural than a feature-level partnership. What the September Dreamforce announcement did was formalize that trajectory: instead of Google Cloud being a data destination Salesforce could optionally connect to, it becomes an infrastructure host in the same tier as AWS.
That history matters for reading how durable this move is. Salesforce has spent years insisting it is cloud-agnostic even when its infrastructure told a different story. The Hyperforce-on-Google-Cloud rollout is the first time that agnosticism has real infrastructure behind it, which raises the obvious follow-up question of whether Azure eventually gets the same treatment, or whether Microsoft’s competing Copilot and Dynamics stack makes that partnership less likely than the AWS and Google Cloud pairing.
Enterprise AI Stack Fragmentation: The Problem This Solves
The stated goal behind the partnership, according to both companies, is eliminating friction between fragmented enterprise systems where data, agents, and applications operate in silos. That framing reflects a real and growing complaint from enterprise IT teams: a typical large organization now runs a CRM on one vendor’s infrastructure, a data warehouse on another’s, and multiple AI agent platforms layered on top, with each connection requiring bespoke engineering to keep data in sync and permissions consistent.
Zero Copy data sharing is the specific mechanism meant to address that. Rather than exporting Salesforce Data Cloud records into BigQuery and maintaining a pipeline to keep both copies current, Zero Copy lets each platform query the other’s data directly. That cuts out a category of integration bugs that come from stale or partially synced data, and it reduces the storage cost of holding duplicate copies of the same enterprise dataset. It also, less discussed in the announcement, deepens the technical relationship between Salesforce and Google Cloud, since an enterprise that builds workflows against Zero Copy is building against a Salesforce-Google Cloud-specific integration rather than a portable standard.
The Tableau MCP connector follows a similar logic. Model Context Protocol has emerged over the past year as a de facto standard for connecting AI agents to external tools and data sources, and Salesforce exposing Tableau through MCP means any Gemini Enterprise agent, or in principle any MCP-compatible agent, can pull Tableau visualizations and metrics into a conversation without a custom Tableau API integration.
Comparing This Deal to GPT-6 Astra’s Multicloud Rollout
The Salesforce-Google Cloud announcement lands in the same week as another multicloud story: OpenAI’s GPT-6 Astra, the company’s newest and most capable frontier model, reached general availability on Amazon Bedrock on September 8, roughly a week before Dreamforce. Astra is now available through OpenAI’s own API, Microsoft Foundry, and Amazon Bedrock, making it a genuinely multicloud model in the same way Salesforce is becoming a genuinely multicloud SaaS platform.
The pattern across both stories is the same: the biggest names in enterprise software and AI are refusing to be exclusive to one hyperscaler, even when one hyperscaler was their original or most natural home. For OpenAI, spreading Astra across AWS, Azure, and its own infrastructure maximizes reach into whichever cloud an enterprise customer has already standardized on. For Salesforce, spreading Hyperforce across AWS and Google Cloud does the same thing for CRM and AI agent workloads. Enterprises are the beneficiaries of that competitive pressure, since it means less lock-in risk and more leverage when negotiating hosting terms.
| Metric | Salesforce Hyperforce on Google Cloud | GPT-6 Astra on Amazon Bedrock |
|---|---|---|
| Announced | September 15, 2026 (Dreamforce) | September 8, 2026 |
| Hosting clouds | AWS (existing) plus Google Cloud (new) | OpenAI direct, Microsoft Foundry, Amazon Bedrock |
| GA timeline | North America, November 2026 | Already generally available |
| Primary driver | AI agent interoperability, data proximity to BigQuery | Reach into enterprise customers standardized on a specific cloud |
| Next expansion | Germany and additional regions in 2027 | Rolling out to all paid ChatGPT tiers and developers |
What This Means for Enterprise Cloud Buyers
For IT teams evaluating Salesforce, the practical near-term impact is limited. Existing customers on AWS-hosted Hyperforce are not being forced to move, and the Q4 2026 migration wave is explicitly described as covering select US customers rather than a mandatory cutover. The bigger impact is on procurement leverage and future architecture decisions. A Google Cloud-native enterprise that previously treated Salesforce as an AWS-adjacent dependency can now evaluate running its CRM data, analytics, and AI agents inside a single cloud boundary, which simplifies compliance reviews, reduces cross-cloud egress costs, and shortens the list of vendors involved in a security audit.
FinOps teams should also pay attention to the Zero Copy expansion. Cross-cloud data duplication has been a quiet but persistent cost center for any enterprise running Salesforce alongside a BigQuery-based analytics stack, since every sync job that copies Data Cloud records into BigQuery, or the reverse, carries both storage and compute costs. Cutting that duplication out should show up as a measurable line-item reduction for organizations that adopt it, though neither company has published a specific percentage or dollar estimate of expected savings.
Competitive Response: What AWS and Microsoft Do Next
AWS is not standing still. The same week’s coverage of the Dreamforce announcement notes that Salesforce is deepening its AWS integrations in parallel with the Google Cloud expansion, which suggests AWS will push its own version of tighter Agentforce-to-Bedrock interoperability rather than cede AI agent workloads to Google Cloud by default. AWS’s broader AI infrastructure momentum, including GPT-6 Astra’s Bedrock availability and continued chip investment, gives it leverage to keep Hyperforce’s largest share of workloads even as Google Cloud enters as a second host.
Microsoft faces a tougher strategic question. Azure was not brought into the Hyperforce hosting conversation at all, and Microsoft’s own enterprise AI stack, built around Copilot and Dynamics 365, is a more direct Salesforce competitor than a natural hosting partner. That positions Microsoft to respond less by matching this specific deal and more by pushing customers toward Dynamics 365 as a Salesforce alternative that keeps everything inside the Azure and Copilot ecosystem, rather than trying to become Hyperforce’s third cloud host.
Predictions: Where the Salesforce-Google Cloud Partnership Goes Next
Based on the disclosed timeline, the technical scope of the integration, and the broader multicloud trend already visible in the OpenAI and Bedrock story, five outcomes look likely over the next 12 to 18 months.
- North America general availability in November 2026 will trigger a wave of enterprise case studies from Google Cloud highlighting Zero Copy cost savings, used as a sales tool against AWS-hosted competitors.
- Germany’s 2027 rollout will be the first real test of how Hyperforce on Google Cloud handles EU data residency and sovereignty requirements, an area where Google Cloud has invested heavily to compete with AWS and Azure in Europe.
- AWS will announce a matching or deeper Agentforce-Bedrock integration within two to three quarters to prevent Google Cloud from becoming the default choice for Salesforce customers building new AI agent workflows.
- Microsoft will not pursue Hyperforce hosting and will instead intensify Dynamics 365 and Copilot competitive positioning against Salesforce directly, treating this partnership as validation that Salesforce and Microsoft are rivals, not infrastructure partners.
- Other large SaaS platforms with single-cloud dependencies will face increased customer pressure to announce their own second-cloud hosting options, following the precedent Salesforce just set with a name-brand competitor’s infrastructure.
The Bigger Picture for Multicloud Enterprise Software
What makes this deal notable is not the technology. Cross-cloud data sharing and agent interoperability have been in development across the industry for a couple of years. What stands out is the willingness of two competitors in adjacent markets to build genuine infrastructure dependency on each other. Google Cloud competes with Salesforce’s own AI ambitions in some enterprise segments, yet it is now hosting Salesforce’s core platform. That tension gets resolved by a simple calculation: the revenue and market positioning both companies gain from a deeper partnership outweighs the strategic risk of empowering a partial competitor.
For the wider cloud computing market, the message is that hyperscaler exclusivity is becoming a liability rather than a feature for major SaaS platforms. Enterprises want their CRM, their data warehouse, and their AI agents to work together regardless of which vendor built which piece, and Salesforce just demonstrated it is willing to hand a rival meaningful control over its own infrastructure to make that happen. Whether AWS and Microsoft answer with their own multicloud concessions, or instead double down on vertical integration within their own ecosystems, will shape how the rest of the enterprise software market approaches cloud hosting decisions through 2027.
Frequently Asked Questions
What is Hyperforce and why does it matter for this announcement?
Hyperforce is Salesforce’s infrastructure architecture for running its CRM and AI platform on public cloud providers instead of Salesforce-owned data centers. It originally launched on AWS, which has remained its primary host. The September 2026 announcement adds Google Cloud as a second native host, not just an integration partner.
When does Hyperforce on Google Cloud become generally available?
General availability in North America is targeted for November 2026. Select US customers begin migrating in the fourth quarter of 2026, and Hyperforce on Google Cloud was already handling live production traffic as of the September 15 announcement. Germany is the next named region, expected in 2027.
Is Salesforce leaving AWS?
No. AWS remains Hyperforce’s original and still-primary cloud host, and Salesforce is simultaneously deepening its AWS integrations alongside the new Google Cloud partnership. This is an addition of a second infrastructure option, not a migration away from AWS.
What is Zero Copy data sharing between Data Cloud and BigQuery?
Zero Copy lets Salesforce Data Cloud (Data 360) and Google BigQuery query each other’s data directly without duplicating it into a separate copy. This removes the need for sync pipelines between the two systems and reduces the storage and compute cost of maintaining duplicate datasets.
Can Agentforce agents use Google’s Gemini models?
Yes. The expanded partnership makes Gemini models available as a model option inside Agentforce, alongside Salesforce’s own models and other partner LLMs. Separately, Gemini Enterprise agents can call Salesforce actions and workflows directly through the headless architecture connection.
Does this affect Salesforce’s relationship with Microsoft Azure?
Azure is not part of the Hyperforce hosting expansion. Salesforce’s relationship with Microsoft has historically centered on Slack, Teams, and Office 365 integrations rather than hosting Salesforce’s core infrastructure, and nothing in the September 2026 announcement changes that scope.
What is the financial value of the Salesforce-Google Cloud deal?
Neither company has disclosed a contract value, minimum spend commitment, or term length. Public statements describe it as an expanded strategic partnership focused on infrastructure, data, and agent interoperability rather than a single quantified deal.
How does this compare to OpenAI’s GPT-6 Astra multicloud rollout?
Both reflect the same trend of major software and AI vendors refusing single-cloud exclusivity. GPT-6 Astra reached general availability on Amazon Bedrock on September 8, 2026, joining its existing availability through OpenAI’s own API and Microsoft Foundry, just a week before Salesforce announced its own move to a second hyperscaler host.
