CrowdStrike, CoreWeave partner to secure AI at machine speed

CrowdStrike and CoreWeave have teamed up to embed AI-native security directly into high-performance computing infrastructure, aiming to close the gap between attacker breakout times and defensive response. The collaboration, revealed in a joint announcement, positions CoreWeave’s cloud as a potential backbone for CrowdStrike’s real-time threat detection and response, particularly for workloads that may not tolerate latency.

The move follows CrowdStrike’s latest threat report, which identified the fastest observed eCrime breakout time in 2025 at 27 seconds. That acceleration has likely prompted some security teams to reconsider architectures that were not designed to operate at such speeds. Some traditional security tools may rely on periodic scans or cloud-based analysis, which could introduce delays. CoreWeave’s infrastructure, built for AI training and inference, is said to offer sub-second processing capabilities, which CrowdStrike may now be leveraging to run its platform directly on the same hardware hosting customer workloads.

The partnership may not be just another cloud collaboration. CoreWeave has positioned itself as a neutral, high-performance alternative to hyperscalers, particularly for model training and inference. Its hardware is already used for some of the most demanding AI workloads, including those for enterprises and startups deploying agentic systems. CrowdStrike’s move could suggest that security is being treated as a co-located priority alongside compute, rather than a separate layer. The announcement does not detail specific customers or deployment timelines, but the approach implies that for some organizations running AI at scale, security might need to be a built-in feature of the infrastructure.

The deal also reflects a broader trend in the AI security market. Over the past month, StartupReader has tracked funding rounds—Island’s $400 million, Reco’s $55 million, Rig Security’s $12 million, and Kontext’s $4 million—all targeting the challenge of securing AI agents as they operate inside enterprise systems. These startups focus on identity, runtime monitoring, and policy enforcement, often treating AI agents as a new class of non-human users. CrowdStrike and CoreWeave, however, seem to be approaching the problem from a different angle. Rather than treating AI agents as a security challenge, they may be positioning AI-native security as a performance consideration—one that could operate without adding overhead.

The distinction between these approaches may be relevant. The startups raising capital are largely working to adapt existing security models for a world where AI agents can spin up, execute tasks, and disappear in seconds. CrowdStrike and CoreWeave, meanwhile, appear to be exploring whether the most effective way to secure AI at scale is to integrate security into the compute layer itself. This could influence how the rest of the market evolves, depending on enterprise adoption patterns.

The partnership’s impact remains to be seen. Some enterprises might prefer the modular, agent-focused approaches from startups like Island or Kontext, while others could opt for this integrated model. The outcome may depend on whether CrowdStrike and CoreWeave can show measurable improvements in detection or response times. For now, the collaboration suggests that securing AI could involve more than just policy or identity—it might also require infrastructure-level adaptations.

Sources: siliconangle.com