Dell expands AMD-based AI infrastructure with new servers and rack systems
Dell adds the air-cooled PowerEdge M8825, modular MI350P platforms and a 2.9-exaflop Helios rack with 72 MI455X GPUs, widening AMD-based options across every AI deployment scale.

Why it matters
- Dell expanded its AI Factory with AMD-based systems on November 17, 2026, covering servers, modular platforms and rack-scale systems.
- The PowerEdge M8825 is an air-cooled AMD EPYC server offering up to 512 cores and 1,024 threads per node.
- The AMD Helios Rack-scale Solution from Dell delivers up to 2.9 exaflops with 72 AMD Instinct MI455X GPUs, per Dell.
- Modular Dell AI Platform configurations use PowerEdge XE7745 and R7725 servers with AMD Instinct MI350P PCIe GPUs.
- The platform adds validated support for AMD Enterprise AI Suite, Red Hat OpenShift and Red Hat AI Enterprise.
Dell Technologies has expanded its Dell AI Factory with new AMD-based infrastructure spanning servers, modular platforms and rack-scale systems, the company announced on November 17, 2026. The update adds the Dell PowerEdge M8825 server, new configurations for the Dell AI Platform with AMD, and the AMD Helios Rack-scale Solution from Dell. The additions give enterprise buyers another vendor-backed route to AI compute built on AMD silicon, at a moment when large hardware suppliers are racing to cover every deployment scale from single nodes to multi-exaflop racks.
What is Dell actually shipping?
The portfolio update covers three tiers. At the node level sits the PowerEdge M8825, an air-cooled server based on AMD EPYC processors. In the middle sit modular configurations of the Dell AI Platform with AMD, built on PowerEdge XE7745 and R7725 servers. At the top sits the AMD Helios Rack-scale Solution from Dell, rated by Dell at up to 2.9 exaflops of performance with 72 AMD Instinct MI455X GPUs. Dell frames the three tiers as a single-vendor path from entry-level experimentation to frontier-scale deployment.
The stakes for buyers are straightforward: the update widens the choice of processor and GPU suppliers inside one vendor's management, support and integration stack. That matters for enterprises wary of locking AI infrastructure to a single silicon vendor.
How much compute fits in one air-cooled box?
The PowerEdge M8825 is aimed at organisations running agentic AI (AI systems that autonomously plan and execute multi-step tasks) and other compute-intensive workloads. It offers up to 512 cores and 1,024 threads per node, according to Dell.
The design goal is explicit: expand AMD-based AI compute capacity while staying on conventional air cooling. That distinguishes it from the liquid-cooled rack-scale systems now common at the top of the AI infrastructure market, where thermal constraints increasingly dictate architecture. Dell also highlighted thermal and power delivery features it said would support operational use in enterprise settings. The company did not publish independent performance figures or pricing, so buyers will need to validate the claims against their own workloads. The takeaway: Dell is betting that a large slice of enterprise AI demand still runs best on air-cooled hardware.
What does the modular platform change?
Dell has expanded its Dell AI Platform with AMD by adding modular configurations built on the PowerEdge XE7745 and R7725 servers. These systems use AMD Instinct MI350P PCIe GPUs — accelerator cards that slot into standard server slots rather than bespoke rack architectures.
The platform bundles several Dell components into one validated stack:
- Dell PowerEdge servers
- PowerSwitch networking
- PowerScale storage
- The Dell AI Data Platform with software and services
According to Dell, the modular approach lets customers start with a single compute node and add GPUs, nodes, storage or networking as demand grows. The expanded platform will include validated support for AMD Enterprise AI Suite, Red Hat OpenShift Container Platform (a Kubernetes-based container management platform) and Red Hat AI Enterprise.
Dell positions this as a pre-integrated option for customers moving from smaller deployments to production systems without replacing earlier infrastructure. That upgrade-without-rip-and-replace claim is the platform's core selling point for enterprises that have already invested in PCIe-based GPU servers. The caveat: "validated support" covers a defined software list, and Dell has not detailed which workloads the validations target. In plain words, the pitch is growth without starting over.
How big is the rack-scale system?
At the top end, Dell introduced the AMD Helios Rack-scale Solution from Dell for organisations with larger compute requirements. The system can deliver up to 2.9 exaflops of performance with 72 AMD Instinct MI455X GPUs, according to Dell.
The rack-scale setup is paired with Dell PowerRack for networking and storage. Dell said the package is backed by deployment and lifecycle services, giving customers a single supplier for integrated AI infrastructure. That single-throat-to-choke support model has become a competitive battleground as rack-scale AI systems grow more complex to deploy and operate.
The 2.9-exaflop figure is Dell's own rating, and the company has not specified the benchmark or precision at which it was measured — a detail that materially affects comparability with rival systems. What is clear is the market signal: Dell intends to compete for the largest AI training and inference installations, not just departmental deployments.
Why does supplier choice matter here?
The announcement shows how major hardware suppliers are broadening AI product lines to cover a wider range of deployment models, from entry-level systems to large rack-scale installations. For enterprise buyers, the emphasis is increasingly on offering a choice of processor and GPU suppliers while keeping management, support and integration under one vendor.
Dell framed the expansion as part of a broader push to give customers more flexibility in silicon, cooling, rack design and scale. It said the additions create a single-vendor route from compute infrastructure to production deployment — from an air-cooled EPYC node, through a modular MI350P platform, to a 72-GPU Helios rack.
The company put its positioning plainly in a statement: "Dell and AMD are giving customers more ways to deploy AI with confidence, whether they are starting small, scaling fast, or going all-in on frontier infrastructure."
That quote carries the voice of the announcement, and it also contains its limit. "Confidence" here rests on Dell's integration and services layer, not on published third-party benchmarks, disclosed pricing, or delivery timelines — none of which appear in the announcement. Buyers weighing the AMD path against Nvidia-based alternatives will want those details before committing. What the expansion does establish is that Dell now offers an AMD-based answer at every tier of the AI infrastructure market, and that AMD's Instinct lineup has secured a place in a top-tier vendor's mainstream enterprise portfolio.
Source: itbrief.com.au
More from James Calloway
Show full bio
News editor covering people, launches and day-to-day practice in artificial intelligence and machine learning at AI In Context.
722 articles
Related articles
- Dell expands AI Factory with NVIDIA for agentic AI workloads
- AMD Ships 256-Core Zen 6 EPYC 'Venice' Chips as AI Agents Drive CPU Demand
- AMD launches 256-core EPYC 9006 CPUs to fight Intel and Nvidia
- ASUS Ramps Vera Rubin NVL72 Shipments as Rack Race Shifts to Execution
- Lenovo Bets on Liquid Cooling and Services to Solve AI's Power Problem