Cisco and NVIDIA are moving AI factory deployments beyond GPU procurement toward complete production systems spanning compute, networking, storage, cooling, and software. The shift reflects growing demand from neoclouds, enterprises, and sovereign AI programs for infrastructure that can deliver inference capacity reliably and efficiently at scale.
CISA added CVE-2026-21962, a critical flaw affecting Oracle HTTP Server and Oracle WebLogic Server, to its Known Exploited Vulnerabilities catalog after confirming active attacks. The CVSS 10.0 vulnerability can let an unauthenticated attacker with network access over HTTP reach sensitive data, making rapid remediation a priority for exposed deployments.
Threat actors have compromised over 270 Zimbra Collaboration Suite instances in an ongoing campaign exploiting a high-severity remote code execution vulnerability. Organizations running Zimbra should review exposure and indicators of compromise, since the scale of observed attacks shows active and repeatable exploitation.
OpenAI's Jalapeño chip is designed for fast inference at scale and posted strong results on SemiAnalysis's InferenceX benchmark. The system delivered more tokens per user and higher throughput per kilowatt than currently available leading alternatives, pointing to a focus on both serving performance and data-center efficiency.
Perplexity has introduced Portable Computer, a version of its agentic Computer platform that runs entirely on hardware users already own, beginning with NVIDIA systems. Local execution removes per-token cloud costs and offers a different deployment model for users who want agent capabilities closer to their data and infrastructure.