On August 2nd, Patrick Gelsinger, CEO of Intel, answered several questions about AI PC and servers at the Q2 2024 earnings conference, stating that the AI PC based on Lunar Lake will ship this quarter, and the new generation of AI chip Gaudi 3 will also be launched in Q3.
Patrick Gelsinger said that the market share of AI PC will grow from less than 10% currently to over 50% in 2026. The shipment of Intel Core Ultra doubled in Q2 and provided AI function for over 300 applications and 500 AI models. This further proves the strong ecosystem cultivated by Intel through 40 years of continuous investment. Since its launch in December, Intel has shipped more than 15 million Windows AI PCs, more than all competitors combined. Intel is still expected to ship more than 40 million AI PCs by the end of the year, with a cumulative shipment of over 0.1 billion by the end of 2025.
The next generation of AI PC Lunar Lake has been mass-produced ahead of schedule in July and will become a catalyst for the next industry-wide device update. Compared with the previous generation of products, Lunar Lake has superior performance, half the power consumption, 50% graphics performance improvement, and 40% energy efficiency improvement. Lunar Lake achieves 3 times the TOPS enhancement with enhanced MTU. Microsoft has certified Lunar Lake to run on more than 80 new Copilot+ PCs of Lenovo and other 20+ OEMs this quarter, which will begin shipping this quarter.
Regarding the global hot AI server and other computing infrastructure settings, Patrick Gelsinger said that we have a solid foundation, including more than 0.13 billion Xeon processors currently supporting data centers worldwide. Our roadmap aims to provide higher performance and efficiency on this fast installation basis, achieve open, flexible and scalable AI solutions, and reduce customer's total ownership costs. This quarter, we took some important steps, first of which is to launch Xeon 6 with e-core processor. It can improve performance, reduce power consumption and greatly integrate the rack. Compared with competitive solutions, the performance per watt has increased by 25%. Next, the P-core Xeon 6 codenamed Granite Rapids, which provides higher performance for the most demanding workloads, will begin shipping this quarter.
Intel is focused on open models, open developer frameworks, and reference designs that combine Xeon with accelerators through OPEA (Enterprise AI Open Platform), which are gaining considerable market attraction. Gaudi 3 will be launched in Q3, which will take Intel's accelerator performance to a new level and reduce the cost of competing products by two-thirds. Specifically, Intel expects Gaudi 3 to increase performance per dollar in inference and training by about 2 times that of H100.
Patrick Gelsinger further stated that Gaudi 3 has strong ecosystem support, including Lenovo, Supermicro, Foxconn, Gigabyte, Inventek, and others.