Nvidia Advances AI Infrastructure With Cloverleaf Partnership and New Harness Research

Nvidia announced a strategic partnership with Cloverleaf Infrastructure, a data center site development company founded in 2024 that connects utility companies with data centers, providing critical infrastructure like power. While the terms of the deal were not disclosed, reports suggest Nvidia's investment could reach hundreds of millions of dollars for a minority stake.

This partnership continues Nvidia's broader strategy of directing profits to fund and develop data centers that ultimately purchase its AI systems. Earlier this week, Nvidia invested $1.5 billion in an SB Energy data center project in Ohio linked to OpenAI.

Nvidia published research showing that the software shell around an AI model plays a larger role in achieving long-horizon task performance than the model itself. The researchers used a custom shell called Agentic Variation Operators, which includes a supervisor agent component that redirects the system when it goes off track. Results showed that Claude Opus 5 achieved a perfect score on the ARC-AGI-3 benchmark with the shell, compared to only 30% without it.

Adel El Hallak, Vice President of AI Products at Nvidia, stated that agents should be understood as the combination of the model and its surrounding scaffolding, tools, and runtime, not just the model's interface.

Nvidia's findings echo conclusions from OpenAI and Databricks, both of which found that shell design significantly impacts model accuracy and operational costs. Databricks CEO Ali Ghodsi noted that the same model can produce vastly different costs depending on the shell choice.

El Hallak said Nvidia's broader goal is to demonstrate that open, controllable shells across infrastructure and runtime layers are essential for safely advancing the agent ecosystem.