Machine data: The next frontier in AI

This week I’m thrilled to be kicking off the Splunk .conf25 user conference in Boston. This is the second opportunity for me to join a Splunk community that is so passionate about what they do to keep their IT environments and applications up, secure, and performing at a high level. It doesn’t take you long to realize that Splunk customers really love Splunk, and I am honored to help lead this organization in a moment where our customers need us.

As I wrote way back at the beginning of the summer, we’re firmly in the era of agentic AI. It’s truly exciting, but to keep the pace of adoption and innovation cooking we have to tackle some major obstacles.

AI places unprecedented demand on infrastructure. It’s hungry for power, compute, and network bandwidth. It presents a whole new set of security threats, creating a trust deficit for users and enterprises alike. And increasingly, there’s an emerging data gap, where we’re struggling to apply AI to all the different data types and sources in our organizations.

At Cisco, we’re addressing these challenges head-on. We provide the critical infrastructure for the AI era – including high-bandwidth low-latency networking, AI safety and security, and a data platform that AI-first organizations need to thrive.

And it’s this last area – DATA – that’s the focus this week at .conf25, and it’s central to how Splunk continues to be critical to our strategy as a company.

A data platform for the AI era

Data is the essential fuel for AI. While the industry has done well training AI models on human-generated data like text, the same has yet to be done with machine-generated data like metrics, events, logs, traces, and other telemetry. Every company has massive volumes of this machine data, but it’s been largely left out of AI for a few reasons: LLMs don’t speak the language of machine data, the information is spread across disparate silos, and the expertise and costs involved can be prohibitive. As a result, we’ve only begun to scratch the surface of what we can do with AI.

Today we announced the Cisco Data Fabric with the ambition to make it as easy as possible to leverage proprietary machine data for training AI models. Here’s what’s under the hood:

Splunk at scale with an open API architecture, adaptability for multi-cloud and hybrid environments, and federation so you can work with your distributed information stores without moving your data. Whether your data is in Snowflake, S3, or anywhere else, you can leverage it for AI.
A new Time Series Foundation Model that we’ve trained and will be open sourcing on Hugging Face. The model is pre-trained for tasks like anomaly detection, forecasting, and automation, but because we’re open sourcing it, anyone can fine tune the model with their own proprietary data. I firmly believe open source will play a major role in the development of AI and at Cisco, with this model and our Foundation Security model, which we open sourced at RSA, we’re all in.
A new Splunk Machine Data Lake that provides a persistent, AI-ready foundation for analytics and training AI models.
AI-Native tools and experiences from the jump, featuring capabilities like Cisco AI Canvas, which reimagines how teams of humans and AI agents can collaborate in real-time on complex issues.

We’re beyond excited for what Cisco Data Fabric with do for our customers. Splunk revolutionized how enterprises understood systems through machine data and that accelerated the cloud revolution. It’s time to do the same for AI.

But the bigger question is one for all of you…

What will you do with your very own MachineGPT?

Machine data is messy, massive, and mission critical. But it’s also the heartbeat of business in nearly every industry. It could be sensor readings in vehicles or industrial equipment, manufacturing lines, retail checkout streams, hospital equipment, or financial transactions, for example.

Ultimately, whatever the types of machine data are in your world, it’s an incredible source of competitive advantage. Ask yourself: “What could you accomplish if you could harness this kind of data for AI?” Maybe you could solve problems you didn’t even know existed? Maybe an AI could predict scenarios you never imagined? Or maybe an AI could find insights and make connections that would be impossible at human scale?