Logpresso and FuriosaAI have signed a memorandum of understanding to jointly develop an NPU-based reference architecture for agentic Security Operations Centers (SOCs), a move that targets real-time, autonomous threat detection for enterprise security teams. The collaboration combines Logpresso's integrated security platform with FuriosaAI's high-performance AI inference processor, RNGD (Renegade), to deliver a hardware-optimized alternative to cloud-dependent security analytics.
What an agentic SOC actually does
An agentic SOC is a framework where AI agents autonomously detect, analyze, and respond to cyber threats. The approach demands high-performance AI inference to process massive volumes of logs and security events in real time, plus a stable operating infrastructure. That heavy AI workload is precisely what makes a dedicated neural processing unit (NPU) like FuriosaAI's Renegade more efficient than general-purpose GPUs or CPUs. The shift toward autonomous security operations - sometimes called AI Agents & Automation in broader enterprise contexts - requires infrastructure that can run these models continuously without latency spikes.
Hardware that matches the workload
Under the agreement, the two companies will create a reference architecture that combines Logpresso's platform with the Renegade NPU. They'll also pursue joint AI demonstration projects and market development. The architecture is designed as a practical model for organizations that want to deploy AI-driven security operations without re-engineering their entire stack. The partners said the collaboration has already moved into the proof-of-concept stage. FuriosaAI's Renegade processor is being used in the 2026 Integrated Security Model Development Pilot Project led by Logpresso and organized by the Korea Internet & Security Agency (KISA), as well as in an AX Sprint consortium project.
Sovereign AI and the next market push
The two companies plan to expand the technology model validated through these pilots into markets with strong demand for sovereign AI - countries and industries that require data and AI processing to remain within jurisdictional boundaries. That positioning could make the NPU-based SOC approach attractive for government agencies, critical infrastructure operators, and enterprises subject to strict data residency rules. For these buyers, a reference architecture that runs entirely on local hardware removes a major compliance hurdle.
Why this matters for Operations
For operations teams, the announcement signals that hardware-specific AI stacks are moving from lab projects to reference architectures that can be evaluated and deployed. Rather than relying on black-box cloud AI that may not meet latency, privacy, or sovereignty requirements, teams get a model they can adapt to on-premise or private cloud environments. The skills needed to assess and integrate such systems are also evolving; professionals who want to lead these deployments may benefit from an AI Learning Path for Cybersecurity Analysts that covers SOC optimization and threat detection with AI. The shift reduces dependence on external AI vendors and gives operations teams more control over the security AI pipeline from inference hardware to response playbooks.
Your membership also unlocks:
