Cisco warns enterprises: Without tapping machine data, your AI strategy is incomplete

Cisco executives argue that the line between product and model companies is fading, and tapping into the 55% of enterprise data growth that current AI overlooks will be the key to success.

In a recent interview with Jeetu Patel, Cisco’s President and Chief Product Officer, and DJ Sampath, Senior Vice President of AI Software and Platform, VentureBeat gained insight into their shared thesis. They believe that every thriving product company must evolve into an AI model company to thrive in the coming years.

With shrinking product lifecycles and the benefits of digital twin technology, the shift to becoming an AI model company seems inevitable.

The discussion highlighted the reasons behind this transformation, supported by solid data points. The team asserts that 55% of data growth consists of machine data that current AI models do not handle. Greg Brockman from OpenAI estimates that we would need 10 billion GPUs to equip every individual with the necessary AI agents, and Cisco’s open-source security model, Foundation-Sec-8B, has already garnered 200,000 downloads on Hugging Face.

Transitioning to a Model-Driven Approach

VentureBeat: You’ve mentioned that every product company will eventually become a model company. Why is this shift inevitable rather than just a potential path?

Jeetu Patel: In the future, there won’t be a distinction between model companies and product companies. Successful product companies will essentially be model companies. The integration between model and product forms a closed loop. Improving the model enhances the product, rather than just modifying the user interface.

New companies that are merely a thin layer on top of a model are facing obsolescence. The true competitive advantage lies in the model that drives product functionality. This requires proficiency in two areas: developing robust models in data-rich domains and creating exceptional product experiences driven by these models in an iterative process where the models adapt and evolve based on product enhancements.

DJ Sampath: This becomes even more crucial as we move towards agent-driven systems. Agents will be governed by these models. The strength of your model will determine your competitive edge.

Unlocking the Potential of Machine Data

VentureBeat: You mentioned that 55% of data growth is machine data, yet current models are not trained on it. Why is this an immense opportunity?

Patel: Up to now, models have been primarily trained on publicly available, human-generated data accessible on the internet. However, we have exhausted the public data that can be crawled. The next frontier is enterprise data, which is largely untapped.

Machine data accounts for 55% of data growth, yet existing models do not leverage this data. Most companies claim that their data is their competitive advantage, but many lack a systematic approach to prepare this data for AI training, limiting their ability to fully utilize AI.

Imagine the vast amount of log data generated when agents operate around the clock, with each individual having multiple agents. Greg Brockman from OpenAI suggested that we are three orders of magnitude away from the required 10 billion GPUs if each person were to have a GPU. Failing to train models effectively with machine data hinders the full potential of AI utilization.

Sampath: The majority of models are trained on public data, while enterprise data primarily consists of machine data. We are unlocking this machine data, providing each enterprise with a foundational model. This model serves as a starting point, allowing enterprises to develop applications and agents customized to their proprietary data. While we are transitioning into a model company, we are also simplifying the process for every enterprise to construct their models using the infrastructure we offer.

The Strategic Advantage of Hardware Companies

VentureBeat: Hardware is often viewed as a drawback in the software and AI era. However, you argue otherwise. Why?

Patel: Many underestimate the value of hardware. I believe that hardware is a significant asset because when you combine exceptional hardware, software, and AI models, the synergy creates magic.

Consider the possibilities of correlating machine data from logs with our time series model. A slight change in your switch or router temperature could predict system failure in three days, a correlation that was previously undetectable. By identifying the change, rerouting traffic, and preempting issues, you can enhance predictive maintenance and infrastructure stability significantly.

Cisco plays a vital role as the infrastructure company for AI. This revolutionizes the level of reliability we can achieve for our infrastructure. In industries like manufacturing, which generate vast amounts of data daily, the integration of agentic AI and accumulated metadata reshapes competitiveness. With sufficient data, these industries can navigate disruptions such as tariffs and supply chain fluctuations, avoiding price and availability challenges.

Cisco’s Commitment to Open Source

VentureBeat: Why release your security models as open source, potentially compromising your competitive edge?

Sampath: The reality is that attackers also have access to open source models. The focus should shift to empowering as many defenders as possible with robust models to strengthen defense capabilities. This was the motivation behind our launch of the open-source model, Foundation-Sec-8B, at RSAC 2025.

Funding for open-source projects has stagnated, leading to increased strain in the open-source community and a need for sustainable, collaborative funding sources. It is a corporate responsibility to provide these models while also granting communities access to AI defense tools.

We have integrated ClamAV, a widely used open-source antivirus tool, with Hugging Face, which hosts over 2 million models. Every model undergoes malware scanning to secure the AI supply chain. We are at the forefront of safeguarding the AI ecosystem.

Patel: We introduced not only an open-source security model but also one for Splunk for time series data. These models combine time series and security incident data to yield valuable insights. With 200,000 downloads on Hugging Face, we are witnessing resellers developing applications based on these models.

Customer Response to Cisco Live Product Launches

VentureBeat: How have customers reacted to the recent product launches at Cisco Live?

Patel: Customers fall into three categories. Firstly, there are those who are thrilled: ‘This is exactly what we needed. Thank you.’ Secondly, there are those willing to explore: after a demo by DJ, they are amazed by the product during a POC, exceeding their expectations from the stage presentation.

The third group consists of skeptics who scrutinize every announcement. This group has diminished over the years. As this group has shrunk, we have seen significant improvements in our financial performance and market perception.

We refrain from discussing plans too far into the future, focusing on a six-month timeframe. Our roadmap is so extensive that we have enough to engage with customers for six months. Our primary challenge is keeping customers abreast of the rapid pace of innovation.

Customer-Centric Approach Over Hardware Focus

VentureBeat: How are you transitioning your hardware-centric customer base without causing disruption?

Patel: Rather than fixating on ‘hardware versus software,’ we start by understanding the customer’s current position. The strategy can no longer rely solely on perimeter-based firewalls for network security, given the evolving market dynamics. We offer a fully revamped firewall lineup.

Our approach involves starting with firewalls, transitioning to Multicloud Defense, integrating Hypershield enforcement points with Cilium for observability, and introducing Smart Switches, all managed through Security Cloud Control. This allows customers to progress at their own pace without overwhelming complexity, thanks to our platform advantage.

Our message to customers is to begin where they are and guide them through the transformation journey, rather than demanding an abrupt shift to new solutions. By offering a seamless transition, we ensure a smooth evolution for our customers.

Empowering Global Partners for AI Revenue Growth

The interview concluded with insights into Cisco’s Partner Summit in San Diego, scheduled for November. Patel emphasized the importance of sustained emphasis to mobilize the partner ecosystem. VentureBeat recognizes the critical role of a robust global partner network in realizing the long-term AI vision of any cybersecurity company.