Mark III Systems Secured Nvidia AI Factory Contract
The deployment for a Fortune 100 enterprise marks an expansion in internal hosting of open-source AI models.
Updated on Sept. 30, 2026 in Data Centers
Live Poll
Is now a good time for large companies to move AI infrastructure to private data centers?
Mark III Systems has signed an agreement to deploy an Nvidia AI factory utilizing Vera Rubin NVL72 hardware for a Fortune 100 company. The deal, announced during a TD Synnex earnings call, underscores the growing enterprise shift toward managing internal AI infrastructure.
Why it matters
Enterprises are increasingly seeking to host open models in their own data centers to control costs and budget predictability, moving away from reliance on expensive, external frontier models. This shift suggests that internal infrastructure management is becoming a critical competitive advantage for large corporations.
The Vera Rubin NVL72 architecture links 72 Nvidia GPUs with 36 Nvidia Vera CPUs into a unified cluster. This hardware is currently seeing a rapid product ramp, following the initial release of the architecture in the summer of 2026.
The players
Mark III Systems
A Houston-based systems integrator that specializes in designing, deploying, and managing enterprise AI factory infrastructure.
Nvidia
A semiconductor company that designs high-performance GPUs and CPUs and creates the architecture for large-scale AI data centers.
TD Synnex
A Fremont-based IT distributor and solutions aggregator that facilitates the supply chain for complex enterprise technology deployments.
The details
Mark III Systems provides a full-stack integration for these AI factories, covering hardware deployment, supply chain logistics, and Day 2 co-admin operations—ongoing maintenance and management after the initial setup. The Vera Rubin platform is designed to consolidate high-performance computing tasks, allowing enterprises to run large open-source models locally. This allows companies to bypass the unpredictable costs often associated with cloud-based frontier models, which are estimated to serve only 10 to 20 percent of typical enterprise use cases.
Timeline
Mark III Systems began delivering AI server clusters with TD Synnex in 2023.
Nvidia began shipping Vera Rubin platforms during the summer of 2026.
TD Synnex revealed the new agreement during a third-quarter earnings call in September 2026.
The Tech Race
This deployment marks an evolution in the firm's technical capability by moving beyond the Nvidia DGX SuperPod designs they have delivered since 2023 to the newer Vera Rubin platform. It reflects a broader market trend where enterprise integrators are racing to standardize high-density, open-model infrastructure to displace cloud-reliant AI services.
This deal signals that internal high-performance computing is becoming accessible to more large enterprises that previously relied on external AI providers. Organizations will likely see more options for purchasing turn-key AI factory solutions that replace the need to develop proprietary infrastructure from scratch.
The takeaway
The rapid adoption of the Vera Rubin platform indicates that infrastructure demand is pivoting toward hardware designed specifically for local, cost-managed AI training. Watch for the official performance metrics of the first fully operational Vera Rubin AI factory as a benchmark for this transition.
Further reading
For more on how companies are building dedicated AI infrastructure, visit the Data Centers section.
Live Poll
Is now a good time for large companies to move AI infrastructure to private data centers?
