RAVEL Launches Project Quicksilver With Sandia Labs

The research collaboration will study how AI workloads can be orchestrated across distributed sites to manage power, cooling and cost.

Updated on Nov. 19, 2026 in Data Centers

Isometric editorial illustration of three sealed cooling tanks linked by a copper loop on a bare concrete floor, evoking distributed computing infrastructure.
RAVEL and Sandia National Laboratories have launched Project Quicksilver to study how AI workloads can be orchestrated across distributed sites to manage power, cooling, and cost. AI Illustration. Upload story photo >

RAVEL has begun a research collaboration with Sandia National Laboratories on Project Quicksilver, using RAVEL Orchestrate AI to study power-aware orchestration of AI workloads across distributed sites. The initial phase evaluates air-cooled and immersion-cooled computing environments across three Sandia locations linked by a wide-area network.

Why it matters

Operators face pressure to expand AI computing while managing electricity supply, infrastructure spending and regulatory limits. The research aims to produce evidence on whether orchestration policies can reduce waste and improve economic outcomes in distributed AI estates across the United States.

Phase one compares air-cooled and immersion-cooled environments at three Sandia sites, measuring energy consumption and operating costs for identical workloads. Network bandwidth and speed are varied using Box-Behnken and Plackett-Burman experimental designs, and results for those measurements have not yet been reported.

The players

RAVEL

RAVEL is a company that develops RAVEL Orchestrate AI, a federated orchestration layer for AI workloads across heterogeneous infrastructure.

Sandia National Laboratories

Sandia National Laboratories is a U.S. Department of Energy national laboratory that conducts research in national security, energy and advanced computing.

Dave Martinez

Dave Martinez is the Engineering Program and Project Lead for Infrastructure Computing Services at Sandia National Laboratories.

The details

Workloads are scheduled and moved across sites as one coordinated system, informed by available power, cost and performance requirements. Phase one also analyzes capital and operating costs related to power, cooling, networking and operations, and examines how scheduling and placement decisions affect AI workload execution time.

Timeline

  1. Phase one: initial evaluation of cooling environments across Sandia locations.

  2. Multi-phase program: ongoing research using RAVEL Orchestrate AI.

Seasonal Patterns

The project extends the work of the U.S. Department of Energy national laboratory network into AI infrastructure economics, testing how orchestration choices affect energy and cost. The partners plan to expand the initiative to additional DOE national laboratories.

The research targets how AI services are powered and priced at the infrastructure level, which could shape operating costs for organizations that rely on large-scale computing. Results on cooling and orchestration may eventually inform how future data centers are designed and run.

The takeaway

Efficient placement of AI workloads, not just faster chips, may determine how much electricity and money large computing estates consume. Tracking the Quicksilver results could show whether orchestration policies deliver measurable savings.

Further reading

Explore more on Data Centers coverage for context on how AI infrastructure is being built and powered.

Source note: This article includes information reported by DataCenterNews US.