infrastructureScore 85/100Watch
Jensen Huang: Agentic systems require fine-grained controllability and sandboxing; OpenClaw, Hermes critical infrastructure for enterprise AI
“how can we have very very specific fine-grained control? You know, if not for rags, if not for conditional inputs, if not for our all of our prompts um directly into output was was too coarse. And so the fact that we can condition, the fact that we can control the agents um all the way down... I think that that level of control and that level of collaboration with agents will be game changing. We don't need the the agents to be 100% accurate, 100% high quality in order for us to use it... And so I I think controllability is probably the single biggest breakthrough that we need for agents at every single level.”
Why it matters
Huang identifies controllability—not raw model capability—as the critical infrastructure gap for enterprise agentic systems. This signals NVIDIA's strategic focus on sandboxing, orchestration middleware, and model grounding tools (OpenClaw, Hermes) as non-negotiable for production deployment, implying capex toward these layers.
Investment implication
Companies providing agent orchestration, sandbox/isolation layers, prompt engineering platforms, and fine-tuning infrastructure (including LangChain, deep agent alternatives) will compete for prominence in NVIDIA's ecosystem. The emphasis on local, controllable AI vs. cloud monoliths suggests increased demand for edge inference engines and model serving infrastructure tuned for determinism and auditability.