Intel launched its latest Xeon server processors on Tuesday, aiming to reclaim its share of the data center market. The company also announced that its Gaudi 3 artificial intelligence accelerator chips would be priced significantly lower than those of its competitors.
The release of the sixth-generation Xeon chips is critical for Intel (INTC.O), which has faced challenges in retaining data center market share against rivals like Advanced Micro Devices (AMD) (AMD.O). According to data from Mercury Research, Intel’s share of the x86 data center market has declined by 5.6 percentage points over the past year to 76.4%, with AMD now holding 23.6%.
Intel’s setbacks in its manufacturing process have allowed AMD to gain traction, as the latter utilizes Taiwan Semiconductor Manufacturing Co (TSMC) (2330.TW) for chip fabrication.
The Xeon 6 server processors come in two main variants: a larger, high-performance version, and an “efficiency” model positioned as a replacement for older-generation chips. The efficiency model aims to reduce server rack requirements by about 67% compared to second-generation chips, catering to media, website hosting, and database calculations.
“Simply put, performance up, power down,” stated Intel CEO Pat Gelsinger at the Computex trade fair in Taipei, emphasizing the efficiency gains of the new processors.
The high-performance model is tailored for complex AI computations and tasks requiring increased processing power. The efficiency Xeon chip is available for purchase immediately, with the performance model slated for release in the third quarter. Intel plans to introduce additional variations next year, following a one-year delay due to changes in manufacturing processes.
During a briefing with reporters, Intel disclosed that a Gaudi 3 accelerator kit, featuring eight AI chips, is priced at approximately $125,000, while the earlier Gaudi 2 carries a list price of $65,000. Gelsinger remarked on the competitive pricing, stating, “In other words, it crushes the competition.”
In comparison, pricing details for AMD and Nvidia (NVDA.O) chips remain undisclosed. However, custom server vendor Thinkmate estimates that a comparable HGX server system with eight Nvidia H100 AI chips can cost over $300,000.
Intel also unveiled details about its next-generation laptop chip, Lunar Lake, which boasts a 40% reduction in power consumption and a more powerful AI processor. The chip is set to ship in the third quarter, with Gelsinger expressing enthusiasm for its adoption in numerous new PCs.
Looking ahead, Gelsinger outlined Intel’s roadmap, with Arrow Lake slated for introduction next year and Panther Lake planned for 2025. He also acknowledged the contribution of TSMC in facilitating core technologies for Lunar Lake, underscoring Intel’s commitment to advancing AI in PCs.
