Dell PowerEdge AI servers – training and inference in your rack
Purpose-built PowerEdge XE systems with NVIDIA, AMD and Intel accelerators: from 2U air-cooled servers to liquid-cooled platforms with 8 GPUs. Delivery, rack integration and commissioning in Serbia.
-
Up to 8 GPUs per server – NVIDIA HGX or AMD Instinct
-
Air and direct liquid cooling (DLC)
-
Rack audit, integration and commissioning at your request
What PowerEdge AI is
Servers designed around accelerators
PowerEdge XE is Dell’s series of rack servers designed around accelerators: power, airflow and chassis are sized for training and inference under continuous full load – not for occasional GPU jobs.
The offer follows the manufacturer’s current line: from the 2U R760xa and 4U XE8640 for smaller GPU fleets, through the 6U XE9680 and flexible XE7740 systems, to the 10U XE9780 and the liquid-cooled XE9780L, XE9785L and XE9680L platforms with 8 NVIDIA HGX or AMD Instinct accelerators.
You pick the GPU platform by the task: NVIDIA HGX for large-model training, AMD Instinct for more cost-effective inference, Intel Gaudi 3 or flexible PCIe cards for mixed workloads – in the same rack format and with the same integration service.
We deliver the servers as a complete system: a rack and power audit, installation, iDRAC setup, a load test and IT team training – and we stay with the equipment through warranty service.
Why PowerEdge AI
Why companies choose PowerEdge XE for AI
Four reasons the XE series ends up in AI labs and data centers instead of a generic rack server.
Purpose-built for AI workloads
The XE series is not a regular server with a GPU: Titanium power supplies, deep chassis and airflow are sized for 24/7 full accelerator load.
A choice of GPU platform
NVIDIA HGX (H100/H200/B200/B300), AMD Instinct MI300X/MI355X, Intel Gaudi 3 or flexible PCIe cards – the scenario chooses the GPU, not the other way round.
Air or direct liquid cooling
From 2U air-cooled systems to DLC platforms in IR7000/IR5000 racks – we pick the scale your room and power can realistically take.
Management and security
iDRAC with a Redfish API, cryptographically signed firmware and Silicon Root of Trust – the GPU fleet stays under your IT team’s control.
Models
PowerEdge XE models on offer
From a 2U rendering server to liquid-cooled platforms with 8 accelerators – every card is a “Request a quote” inquiry, with no prices. The manufacturer’s current range changes – the concrete model and configuration are confirmed in the specification.
Dell PowerEdge XE7740
- 8× PCIe GPUs up to 600 W or 16× up to 75 W (RTX PRO 6000 SE, H200 NVL, L40S, L4, Gaudi 3)
- 2× Intel Xeon 6, up to 144 cores
- up to 8 TB of DDR5 memory
- up to 122 TB NVMe
- 4U, air cooling
Dell PowerEdge R760xa
- 4× GPUs up to 400 W or 12× up to 75 W
- 2× Intel Xeon Scalable
- up to 8 TB of DDR5 memory
- up to 122 TB NVMe
- 2U, air cooling
Dell PowerEdge XE9780
- 8× NVIDIA HGX B300 or B200 (NVLink, SXM6)
- 2× Intel Xeon 6, up to 86 cores
- up to 4 TB of DDR5 memory
- up to 245 TB NVMe (16× E3.S)
- 10U, air cooling, 12× 3200 W Titanium
Dell PowerEdge XE9680
- 8× NVIDIA H100 or H200, 8× AMD MI300X or 8× Intel Gaudi 3
- 2× Intel Xeon Scalable, up to 64 cores
- up to 4 TB of DDR5 memory
- up to 122 TB NVMe
- 6U, air cooling
Dell PowerEdge XE8640
- 4× NVIDIA H100
- 2× Intel Xeon Scalable
- up to 4 TB of DDR5 memory
- up to 122 TB NVMe
- 4U, air cooling with GPU LAAC assist
Dell PowerEdge XE9780L
- 8× NVIDIA HGX B300 NVL8, 270 GB SXM6 (or HGX B200)
- 2× Intel Xeon 6, up to 86 cores
- up to 4 TB of DDR5 memory
- up to 122 TB NVMe (E1.S or U.2)
- 3OU node in an IR7000 rack, direct liquid cooling
Dell PowerEdge XE9785L
- 8× AMD Instinct MI355X, 288 GB OAM – or 8× NVIDIA HGX B300
- 2× AMD EPYC 9005, up to 192 cores
- up to 6 TB of DDR5 memory
- up to 122 TB NVMe
- 3OU node in an IR7000 rack, direct liquid cooling
Dell PowerEdge XE9680L
- 8× NVIDIA HGX B200, 180 GB SXM6, NVLink
- 2× Intel Xeon Scalable, up to 64 cores
- up to 4 TB of DDR5 memory
- up to 122 TB NVMe (8× 2.5-inch)
- 4U, liquid cooling, IR5000 rack
FAQ
Frequently asked questions
Request a specification and a rack check
Describe the job the server should cover (training, fine-tuning, inference, rendering) and how many GPUs you are considering – we propose the model and cooling, and check what your rack and installation can realistically take before the order.
AI infrastructure
More from the AI range
NVIDIA RTX PRO Server
A GPU server with RTX PRO Blackwell Server Edition cards – up to 8 GPUs and vGPU/MIG sharing for multiple users.
NVIDIA DGX Station
A deskside AI supercomputer for the team: GB300 Grace Blackwell Ultra, up to 748 GB of coherent memory.
Dell AI Switches
Dell PowerSwitch and NVIDIA switches for GPU clusters: 400G and 800G Ethernet, InfiniBand and Enterprise SONiC.
Servers & storage
Standard rack servers, all-flash and hybrid storage from Dell, HPE, Lenovo and Supermicro – for virtualization and business applications.