On theCUBE Pod: Nvidia steamrolls expectations as Mythos shakes up cybersecurity

Nvidia’s latest quarterly earnings exceeded revenue expectations, and the company is using its windfall to expand its AI footprint through new investments and partnerships. Among the moves detailed were an expanded collaboration with Cisco Systems Inc. on AI infrastructure, a reported $12.9 billion deal to acquire Hugging Face, and another investment in Perplexity. According to analysts on theCUBE Pod, the AI build-out phase has shifted into a focus on execution and control, with tokens becoming the primary measure of profit. The scope of Nvidia’s influence is now larger than just graphics processing units, touching areas such as networking, inference, agents, robotics, software economics, security, and political backlash.

The Cisco-Nvidia partnership is centered on building rack-scale AI factories that combine Nvidia’s accelerated computing and Spectrum-X architecture with Cisco’s networking and enterprise operations. Dave Vellante, chief analyst for theCUBE Research, noted that this creates a bridge from general-purpose enterprise IT infrastructure to accelerated computing. He pointed out that trillions of dollars in installed general-purpose computing value will eventually be absorbed by accelerated computing, making Cisco highly relevant again. This deal reflects a broader shift from networking that merely connects computers to networking that effectively creates the computer by orchestrating thousands of GPUs. John Furrier, research executive analyst for theCUBE Research, predicts Nvidia will guide the expansion of distributed AI systems as the technology moves from enterprises to the edge.

Despite concerns on Wall Street about a so-called SaaSpocalypse, Salesforce Inc. beat earnings expectations, with its stock rising nearly 23% in a single trading day. TheCUBE analysts say this shows software-as-a-service companies are not disappearing but rather evolving in their roles. Furrier stated that the simplistic rationale behind the SaaSpocalypse narrative is dead, noting that infrastructure demand remains strong while incumbent enterprise software companies show that AI can expand their businesses rather than destroy them. He added that bad software will always fail, but the pressure is forcing software to become AI-native while simultaneously creating a new infrastructure stack.

AI model vendors are in urgent need of more data, which software companies can supply. These partnerships could also create what Furrier and Vellante call a system of intelligence to help organize the emerging AI enterprise. Vellante suggested that frontier model developers will likely acquire or partner with companies like Salesforce. He believes they intend to use a Trojan horse strategy to capture value, although he acknowledged that leaders like Marc Benioff and Larry Ellison are not easily outmaneuvered. He called this a first step toward simplifying AI.

Anthropic PBC’s release of the Claude Mythos AI model has created significant disruption in security operations by demonstrating how quickly advanced AI can identify and exploit vulnerabilities. This disruption has benefited CrowdStrike Holdings Inc., which reported $333 million in net new annual recurring revenue, coming in $45 million above the high end of its guidance. Vellante said CrowdStrike not only held its ground but expanded its territory, using its Falcon Flex program to drive new customers. The timing aligned with the company’s sales kickoff earlier this year and was further aided by the Microsoft event, which helped companies like CrowdStrike consolidate their positions.

Data from Qualitate, an AI-native primary data intelligence platform, supports the view that Mythos represents an inflection point in cybersecurity. Furrier described the model’s release as a ChatGPT moment for the security industry. In its aftermath, companies are scrambling to prepare for a new wave of cyberattacks, which could become even more dangerous when AI agents are involved. Furrier called it a call to arms, noting that the cybersecurity landscape is not as healthy as many believed, and that agents, which function like drones, could potentially go off the rails.

More cybersecurity news from TechManNews.