South Korea’s high net portal large Naver Corp. has changed its principal chip provider of its synthetic intelligence server for its map service, Naver Place, from Nvidia Corp. to Intel Corp.
Naver has up to now used Nvidia’s graphic processing unit (GPU)-based server to run its AI-powered location data provision service however lately changed it with Intel’s central processing unit (CPU)-based server, individuals accustomed to the matter stated on Monday.
Naver’s AI server change comes as world data know-how companies are more and more disgruntled with Nvidia’s GPU worth hikes and a world scarcity of its GPUs.
The Korean tech firm makes use of an AI mannequin to differentiate false data from helpful knowledge when customers seek for data resembling “close by eating places” on the Naver Map app for the Naver Place service.
This course of entails knowledge processing, inference and verification, and Naver has used Nvidia’s GPU chips to energy such processes.
SHORT SUPPLY OF AI GPUs
Nevertheless, it has develop into more and more exhausting to safe GPU accelerators, together with Nvidia’s H100, which use a number of chips for knowledge processing and computing, as a consequence of rising demand to be used in generative AI units.
The value of an AI accelerator utilizing Nvidia’s GPUs has risen to as excessive as 80 million gained ($59,193) from about 40 million gained earlier this 12 months, making it costlier for AI server customers to improve their methods, business officers stated.
To make issues worse, the lead time, or the time to get Nvidia’s GPU-based accelerators after inserting an order, has grown to 52 weeks, making it even more durable for a system improve, they stated.
Nvidia produces GPUs utilized in generative AI units resembling ChatGPT, a chatbot developed by Microsoft Corp.-backed OpenAI, which requires semiconductors that may course of giant knowledge rapidly.
The US chip designer has been closely counting on Taiwan Semiconductor Manufacturing Co. (TSMC) for the chip packaging of its GPUs and accelerators. With TSMC’s packaging course of strains virtually totally booked, nonetheless, Nvidia is popping to different foundry gamers resembling Samsung Electronics Co.
INTEL IMPROVES CPU CAPABILITIES
Trade officers stated GPU servers normally deal with AI inference processes 10 instances quicker than CPU servers.
Nevertheless, Intel significantly improved its AI inference course of time for these purchasers looking for alternate options to Nvidia’s GPUs, business watchers stated.
Naver has gone by means of a month of take a look at runs of its Intel CPU-powered AI server earlier than placing it into service, sources stated.
The Korean firm will possible proceed to make use of Intel’s next-generation CPUs, such because the Sapphire Rapids CPU, for its system upgrades to supply higher location and different data search providers, they stated.
Analysts stated the Naver-Intel tie-up could result in an extra diminishing of Nvidia’s affect within the AI processor market.
Microsoft Corp. has reportedly teamed up with fabless chip designer Superior Micro Units Inc. (AMD) to assist bolster the US chipmaker’s growth into AI processors.
The 2 firms are becoming a member of forces to compete towards Nvidia, which controls an estimated 80% market share within the world AI processor market, media reviews stated.
Write to Jeong-Soo Hwang at hjs@hankyung.com
In-Soo Nam edited this text.
