“An Opportunity for Samsung and SK Hynix” Korean Chipmakers Smile as Intel Challenges Nvidia [Chip Talk]

CEO Tan: "Hiring a Design Expert"...A Direct Challenge to Nvidia
Diversifying Demand Sources to Boost Samsung and SK hynix's Bargaining Power
Securing a Software Ecosystem Remains a Key Challenge

Intel, the U.S. semiconductor company, has formally declared that it will strengthen its graphics processing unit (GPU) business, which is at the core of artificial intelligence (AI) semiconductors. Its strategy is to break the dominance of Nvidia in the GPU market. Experts believe Intel’s entry into the GPU market will create an opportunity for Korean memory chipmakers such as Samsung Electronics and SK Hynix to increase their bargaining power.

Ripbu Tan, Chief Executive Officer (CEO) of Intel.

Ripbu Tan, Chief Executive Officer (CEO) of Intel, officially announced Intel’s re-entry into the GPU market at the “AI Summit” held in San Francisco, United States, on the 3rd (local time), saying, “We have hired a highly capable head of GPU design.” The person Intel is believed to have recruited is Eric Demers, a GPU expert and senior vice president who previously worked at Qualcomm. Demers recently revealed via LinkedIn that he is joining Intel as a senior vice president.

CEO Tan stated that the GPU development project will be overseen by Kevork Kechichian, Executive Vice President (EVP), who was recruited last year from Arm. He said, “GPUs are closely related to data centers,” adding, “We will work with our customers to define what it is they need.”

Intel’s renewed attempt to enter the GPU market is seen as stemming from the judgment that it will be difficult to regain market leadership without GPUs, which are central to AI computation. Until now, Intel has introduced its own AI accelerators under the “Gaudi” series, but it has struggled against the barrier posed by Nvidia’s powerful software ecosystem, CUDA. However, Intel is now seeking to enter the server GPU market in tandem with the full-scale launch of its 1.4-nanometer (14A) foundry business in the second half of this year. CEO Tan explained, “Several customers are actively working with Intel Foundry.” Analysts say this move is aimed at major technology companies such as Amazon, Meta, and Microsoft, which are currently looking for alternatives to reduce their dependence on Nvidia GPUs.

Experts expect Intel’s re-entry into the GPU market to have a positive impact on Korean memory semiconductor companies. High Bandwidth Memory (HBM) is essential for manufacturing GPUs and AI accelerators, and as the list of buyers expands beyond Nvidia to include Intel, AMD, and others, the bargaining power of memory manufacturers is expected to increase.

Kim Joungho, Professor at the School of Electrical Engineering at the Korea Advanced Institute of Science and Technology (KAIST), said, “Whether it is Nvidia, AMD, or Intel, if they want to make GPUs, they absolutely need HBM,” adding, “From the standpoint of Korean companies, the increase in demand sources strengthens their pricing power and creates a favorable environment in which they can control supply volumes and secure market leadership.”

However, there are many hurdles Intel must clear before it can succeed in the GPU market. Professor Kim pointed out, “The biggest challenges are the lack of a software ecosystem and the absence of experience in operating data centers,” and added, “It is also questionable whether competitors will readily adopt Intel’s GPUs, given that Intel has long been a dominant force in the CPU market.”