Dell Technologies Says It’s First To Bring Intel’s Gaudi 3 PCIe To Market

Dell is first in bringing Intel’s latest AI accelerators to market inside a server that can house eight Gaudi 3 PCIe, which cost less and draw about half as much power as the market leading GPUs.

Dell is first in bringing Intel’s latest AI accelerators to market inside a server that can house eight Gaudi 3 PCIe, which cost less and draw about half as much power as the market-leading GPUs.

Dell said the Gaudi allows for fast inferencing, silicon diversity, and capital expense management for enterprises leveraging AI workloads. The Gaudi accelerators have a thermal design power rating of 600 watts, according to Intel’s spec sheets compared to Nvidia’s Blackwells, which are rated at 1,200 watts, according to that manufacturer.

The Round Rock, Texas-based Dell said the Gaudi 3 can come installed on the Dell PowerEdge XE7740 rack server, a 4U device that is capable of fine-tuning AI models for specific workflows, and run high-performance inferencing, Dell stated in a press release.

The XE7740 can accommodate up to eight Gaudi 3 PCIe accelerators in either a double-wide, or a four-way bridged configurations. Dell said accelerator-to-accelerator bridging is critical for large scale AI models and larger memory requirements in order to maintain workflow scalability and flexibility.

The server can also deliver flexible networking with one network interface card or NIC per accelerators, and Open Compute Project networking that can provide frictionless integration into existing infrastructure, Dell said.

The XE7740 uses Dell “smart cooling” technology and is designed for air-cooled racks while providing enough horsepower to run Llama4, Llama3, Deepseek, Phi4, Qwen3, Falcon3, and other models. The Gaudi 3 PCIe accelerators integrate with PyTorch and platforms like Hugging Face, enabling organizations to fine-tune models or infer complex AI models efficiently, Dell said.

In an interview with CRN last year, Michael Green, who became the head of Intel’s new North America partner scale group in November 2024, said the company’s channel partners are going to be key to the company winning market share as the company made the Gaudi 3 available to Dell, Hewlett Packard Enterprise, SuperMicro, and its making its own server.

“The channel plays a massive role in how do we take Gaudi 3 out to the masses,” he added. “But if you go from Gaudi 2 where we had one OEM partner to now having four, it's massive progress. So then how do we take that and make it more applicable to the channel partners? In our strategy, it will just take a little bit of time.”

Dell is the worldwide market leader in selling servers and during its most recent quarter, it delivered record server and networking revenue of $12.9 billion, up 69 percent, year over year.

“We’ve now shipped $10 billion of AI solutions in the first half of FY26, surpassing all shipments in FY25. This helped deliver another record revenue quarter in our Servers and Networking business, which grew 69 percent,” said Jeff Clarke, vice chairman and chief operating officer, Dell Technologies, in prepared remarks ahead of the company’s August 29 earnings call. “Demand for our AI solutions continues to be exceptional, and we’re raising our AI server shipment guidance for FY26 to $20 billion dollars.”

Since the AI goldrush kicked off, Dell has touted its close relationship with AI GPU kingpin Nvidia. Similar to its announcement on Wednesday, Dell was also first to market when it delivered a server with Nvidia’s GB300 NVL 72 to AI hyperscaler CoreWeave in July.