[Financial News] Intel Corporation, best known for its Central Processing Units (CPUs) and for operating semiconductor fabs in the United States, is set to begin full-scale production of high-performance Graphics Processing Units (GPUs) for the artificial intelligence (AI) industry. If the plan proceeds smoothly, Intel could reclaim chip production orders that US companies currently place with Taiwan Semiconductor Manufacturing Company Limited (TSMC), and potentially move toward developing its own AI semiconductors. According to US tech outlet TechCrunch, Intel CEO Lip-Bu Tan announced on the 3rd (local time) at the "AI Summit" hosted by Cisco Systems in San Francisco that the company will build GPUs optimized for AI. Both CPUs and GPUs are semiconductors that process information, but they operate differently. CPUs mainly handle information serially and are responsible for running operating systems and performing complex logical operations. GPUs, by contrast, process data in parallel. They were once used primarily to render graphics for games and video production, but as they proved more suitable than CPUs for AI training and inference, they entered a new era. Nvidia Corporation, which manufactures GPUs, has grown into the world’s most valuable company by market capitalization as major global tech firms rushed to order GPUs for AI development. Products commonly referred to as "AI semiconductors" are typically specialized GPUs for AI, combined with high-bandwidth memory (HBM) chips and other components. Intel also produces GPUs, but they are largely limited to integrated graphics embedded in CPUs. This contrasts with rival Advanced Micro Devices (AMD), which has long pursued high-performance standalone GPUs. As a result, Intel has been watching from the sidelines while competitors ship massive volumes of GPUs to AI data centers. At the event on the 3rd, Lip-Bu Tan said, "We recently hired a very capable head of GPU design." He did not specify who this person is. However, foreign media outlets have speculated that he was referring to GPU expert Eric Demers, whom Intel hired last month from Qualcomm. Demers recently announced on LinkedIn that he would be joining Intel as senior vice president. Tan also noted that Kevork Kechichian, an executive vice president who joined last year from UK chip designer Arm to lead Intel’s data center division, will head the new GPU production business. "GPUs are closely tied to data centers," Tan said, adding, "We will work with our customers to define exactly what they need." Because Intel operates contract semiconductor manufacturing (foundry) plants in the United States, it could also produce GPUs on behalf of competitors such as Nvidia. Until now, Nvidia has mainly relied on Taiwanese foundry giant TSMC for manufacturing. Tan explained, "Several customers are actively working with Intel’s foundry business," and said that much of this interest is focused on the company’s cutting-edge 1.4-nanometer (nm) process technology. Tan also pointed out that Chinese chipmakers are managing to produce advanced semiconductors even without access to state-of-the-art manufacturing equipment due to sanctions by the United States and others. "I was shocked to learn that Huawei has hired some of the best semiconductor design experts," he said. "China’s semiconductor industry is trying to lead in AI through a kind of self-reliance model, despite lacking key equipment," he assessed. At the same time, Tan warned that if AI industry growth slows in the future, the bottleneck will likely be in memory chips, highlighting the risk of a memory shortage. pjw@fnnews.com Park Jong-won Reporter