Intel publicly disclosed the 8-chip Gaudi 3 platform pricing, breaking industry norms of secrecy.

Intel is breaking industry tradition by publicly disclosing the list prices of its upcoming Gaudi 3 AI accelerator chip platforms, a move that marks a significant departure from the secrecy that typically surrounds data center accelerator chip pricing. At Computex 2024 in Taiwan, Intel announced that its universal baseboard with eight Gaudi 3 OAM modules will carry a list price of $125,000 for server vendors including Dell Technologies, Hewlett Packard Enterprise, Lenovo, and Supermicro, who plan to support the chip when it launches in the third quarter. The company also revealed a list price of $65,000 for its Gaudi 2 platform, which uses eight Gaudi 2 chips that debuted in 2022.

Intel's pricing strategy positions the Gaudi 3 platform at two-thirds the estimated cost of Nvidia's HGX H100 platform, which Intel calculates at roughly $187,000, according to Anil Nanduri, vice president and head of Intel's AI acceleration office. Beyond cost alone, Gaudi 3 delivers significant performance advantages—providing 2.3 times greater performance-per-dollar than the H100 for inference throughput and 90 percent better performance-per-dollar for training throughput, according to Intel testing. The Gaudi 2 platform sits at one-third the estimated cost of Nvidia's HGX H100 platform. Intel notes that Nvidia does not disclose pricing for its data center GPUs or platforms, and that pricing can vary based on volume and how a server is configured. The Gaudi 3 chips will debut in the third quarter in both PCIe and OAM versions, with Intel declining to provide pricing for individual Gaudi 3 chips.

Nanduri explained that Intel decided to disclose pricing to help customers—particularly startups who typically lack visibility into vendor pricing and purchasing processes—plan their AI computing investments more effectively. "A lot of the innovation is happening from startups. They're not familiar with the ecosystems [and] how the purchasing, procurement [and request for quotes] work, so it helps them get a framing of what to expect," he said. While public pricing for CPUs has been standard in the data center industry, accelerator chips have traditionally remained opaque, with vendors like Nvidia and AMD keeping pricing secret. "We've historically done that with CPUs. When it comes to GPUs, the norm was pricing was always kind of a secret sauce," Nanduri noted.

The decision to disclose pricing reflects changing market dynamics and customer priorities. Gaudi 3 has secured support not only from major partners like Dell, HPE, Lenovo, and Supermicro, but also from additional server vendors including Asus, Gigabyte, and QCT, all of whom announced plans to back the processor with their own systems at Computex. Customers are increasingly focused on customizing AI models in ways that are "computationally sustainable from a cost perspective," according to Nanduri, with particular sensitivity to inference costs. "When customers look at the tokens throughput per dollar or the cost of a million tokens, you see a huge disparity in terms of the cost, and a lot of it is fundamentally driven by either the model size [or] the cost of the infrastructure," he said.