The industry consensus that deploying digital coworkers requires a public cloud provider just hit a significant wall at .conf26 in Denver. Cisco and Splunk have officially broken that mold, announcing the availability of on-premises AI infrastructure designed specifically to host agentic workloads. This move signals a major pivot for enterprises that have felt forced to choose between the agility of agentic AI and the strict requirements of their own data centers.
The Turnkey Stack for Local AI
The announcement introduces the Cisco AI POD for Splunk, a turnkey stack combining Cisco UCS compute, NVIDIA GPU acceleration, Red Hat OpenShift, and Splunk’s AI runtime software. It is a direct play to bring the capabilities of the Splunk AI Assistant — which allows for ad-hoc agentic investigations — and the upcoming Agent Launchpad into the enterprise data center. While Agent Launchpad has been available on the Splunk Cloud Platform since July 15, 2026, this move marks the first time Splunk has extended its accelerated computing capabilities to on-premises hardware. Enterprises now have a binary choice: adopt the turnkey Cisco POD or build their own infrastructure to run these agents locally.
Governance as a Location Decision
This launch formalizes a growing divide in the enterprise agent landscape. As Cisco SVP Jeff Schultz noted at Cisco Live in June 2026, we are transitioning from the era of chatbots to the age of agentic AI, where these tools function as digital coworkers rather than simple productivity aids. However, the location of these coworkers is no longer just an IT preference; it is a governance decision.
With the EU AI Act Article 10 becoming enforceable on August 2, 2026, organizations in regulated industries are under immense pressure to ensure data governance for high-risk AI. The sovereign cloud market, which reached $80B in 2026 with 35.6% year-over-year growth, proves that many firms are unwilling to trust public clouds with their most sensitive reasoning processes. For agents, sovereignty is not just about where data is stored; it is about where the reasoning, tool calls, and derived outputs actually execute. By moving the infrastructure on-prem, Cisco and Splunk are offering a path for companies that view cloud-only agent models as a compliance liability.
The Operational Burden Shifts In-House
For the IT and security teams tasked with managing these systems, the shift is profound. In a cloud-native model, the vendor handles the underlying plumbing. When you bring that infrastructure inside your own perimeter, the burden of trust, routing, and telemetry shifts entirely to the enterprise. As Schultz pointed out, every action an agent takes is a combination of a routing challenge, a trust decision, and a telemetry event. When the agent lives in your data center, your team is responsible for auditing those decisions and ensuring the telemetry is captured correctly. While Agent Launchpad is designed to inherit existing RBAC permissions and record every run with a full evidence trace, the operational reality is that IT teams are now managing a complex, high-stakes software stack that they must maintain, patch, and secure themselves.
Navigating Measurement and Governance Gaps
This development sits at the intersection of two major challenges currently facing the industry. First, it complicates the Agent Measurement Problem. As enterprises split their agent deployments between cloud and on-prem, standardizing performance metrics across these heterogeneous environments becomes significantly harder. Second, it highlights the ongoing formation of the Agent Governance Stack. By providing a way to run agents on-prem, Cisco and Splunk are attempting to bridge the gap between the agility of no-code agent building and the strict requirements of enterprise security.
The Future of Hosting Strategy
The primary counterargument to this shift is that on-prem infrastructure may remain a niche response to regulatory pressure rather than a broader market trend. For non-regulated enterprises, the convenience and rapid innovation cycles of cloud-native platforms — Salesforce’s Agentforce, SpaceXAI’s GrokBot — remain highly attractive. This leaves us with several unanswered questions: Can a unified governance stack truly bridge the gap between cloud-native and on-prem environments? What will be the long-term operational cost for IT teams that choose to maintain this infrastructure? As the industry moves toward agentic workflows, the decision of where to host these digital coworkers will define the next phase of enterprise IT strategy.