Dell Technologies began selling the Dell Pro Max with GB300 desktop workstation on August 20, 2026. The platform is designed to run and develop artificial intelligence models locally. The workstation is based on the NVIDIA Grace Blackwell Ultra GB300 Superchip and delivers up to 2 petaflops of FP4 compute performance, according to the information included in the announcement.
Dell is targeting this platform at model creation and fine-tuning, data science, and molecular simulation, in addition to running artificial intelligence applications at the edge. This places the device in the category of workstations that seek to reduce reliance on cloud environments or data centers during development and testing, while retaining the ability to scale to larger infrastructures when needed.
Hardware Designed for AI Workloads
The workstation includes 748 gigabytes of total unified memory, divided between 252 gigabytes of HBM3e and 496 gigabytes of LPDDR5X. Its storage architecture supports installing up to four Gen4 or Gen5 M.2 NVMe SSDs, with total capacity of up to 16 terabytes.
Dell relies on MaxCool technology for thermal management. It combines an intelligent cooling-plate design with custom cooling components and additional thermal paths. The system also uses a liquid-pumping system to help dissipate heat, in an attempt to balance high performance with noise and power consumption; however, the source does not provide detailed figures for these aspects.
High-Speed Connectivity and Resource Partitioning
With the NVIDIA ConnectX-8 SuperNIC network card, the workstation supports connectivity at speeds of up to 800 gigabits per second. Two workstations can be linked to increase performance and expand models that exceed the capacity of a single device.
The platform also includes a GPU resource-partitioning function through MIG technology, allowing separate resources to be allocated to up to seven users on a single device. This could benefit development teams that need to share the workstation among multiple tasks or users, but the effectiveness of this approach will depend in practice on model size and the nature of concurrent workloads.
What Changes in Practice?
The Dell Pro Max with GB300 provides a desktop option for workloads that were typically associated with AI servers or cloud infrastructure. Its primary value lies in combining large memory, high FP4 performance, and fast network connectivity, rather than being merely an upgrade to the specifications of a traditional workstation.
On the software side, the workstation ships with Ubuntu and a development stack that includes CUDA-X libraries. It can also be integrated with Dell AI Factory with NVIDIA to move from testing models in an initial environment to a larger operating infrastructure in a data center or cloud. The source does not specify the device’s price, geographic availability, or detailed power and cooling requirements, all of which are important considerations before evaluating it as a practical option for organizations.