TL;DR — Key Takeaways
- Splunk AI can now run on Cisco Secure AI Factory with NVIDIA, extending Cisco’s enterprise AI infrastructure strategy.
- The platform combines compute, security, observability and automation in an on-premises environment.
- Cisco says the architecture is tightly integrated but loosely coupled, allowing enterprises to swap components where needed.
- Cost controls include tools for optimizing AI token use and support for less expensive open-weight models.
- Cisco expects AI agents to place major new demands on compute, networks and security, forcing many enterprises to rethink legacy infrastructure.
The Splunk arm of Cisco this week revealed that its suite of artificial intelligence (AI) and machine learning (ML) tools that enable teams to analyze machine data, automate workflows, and resolve security or IT issues faster is now available on an on-premises IT platform for running AI workloads.
Announced at the .conf26 event hosted by Splunk, the Splunk AI portfolio can now be deployed on Cisco Secure AI Factory with NVIDIA, a set of integrated IT infrastructure based on graphical processor units (GPUs) from NVIDIA.
Jeetu Patel, president and chief product officer for Cisco, told conference attendees that making Splunk AI available on the Cisco Secure AI Factory with NVIDIA advances a vertical integrated platform strategy that Cisco is pursuing in the AI era. The goal is to provide IT teams with everything from custom silicon and photonics up through the entire software stack needed to safely deploy AI workloads, he added.
“We are living in a post-Mythos world,” said Patel. “All our assumptions around security and infrastructure must change.”
However, the Cisco approach is designed to be simultaneously tightly integrated but loosely coupled in the sense that IT teams can swap out elements of the software stack as they might prefer, noted Patel.
Fundamentally, Cisco is working toward providing IT teams with more control with the ability to deploy AI workloads at scale across a hybrid IT environment, including all the observability tools and security infrastructure needed to safely deploy AI agents at scale, he added.
Cisco will also provide tools to reduce costs by both providing access to tools to optimize consumption of AI tokens and access to open-weight AI models that can be deployed as alternatives to more expensive AI frontier models, said Patel. Those open-weight models may not be as powerful as the latest frontier models from Anthropic and OpenAI, but in many use cases they are more than adequate for the needs of a typical enterprise, he added.
Finally, Cisco will provide the security tools and platforms needed to have defenses capable of operating at machine speed, said Patel.
The overall goal is to make it sustainable for organizations to deploy what will become trillions of AI agents, said Patel. The AI Agents already consume 5X more tokens than humans and will generate 450% more network traffic than human end users, he noted.
Each IT team will need to determine to what degree their underlying infrastructure will need to be redesigned for the AI era but it’s already apparent that legacy systems are not going to be able to stand the strain. The challenge is that as those AI workloads are deployed the rate at which they consume IT infrastructure resources rapidly outpaces budget allocations. A Futurum Group survey finds nearly half (47%) of organizations are over budget on AI spending, which is requiring them to reallocate resources from other IT initiatives.
Hopefully, IT teams as they head into 2027 will have a better understanding of the total cost of running AI workloads. Less clear, however, is to what degree the average enterprise has the financial resources needed to actually fund those initiatives.
