On Tuesday, Nvidia Corp.’s shares tumbled after Intel presented its Gaudi 3 artificial intelligence (AI) chip, which tops the NVDA H100.
The stock of the Silicon Valley-headquartered company slumped more than 5.00% before settling 2.04% lower at $853.54 per share on April 09. It also shed 0.74% after-hours, with analysts anticipating a 1.34% decline to $842.10 apiece in the coming trading day.
Nvidia has grown immensely since OpenAI Inc. released ChatGPT in November 2022, which sparked a generative AI (GenAI) arms race. The firm’s stock has risen by 209.00% in the past 12 months, including a 72.00% year-to-date (YTD) gain.
However, other tech companies have caught up in large language model (LLM) technologies nearly a year and a half later. Advanced Micro Devices (AMD) Inc. released its MI300 accelerators in December, while Intel Corp. introduced the AI PC as an alternative to GPUs for low-end GenAI in October.
Moreover, Nvidia’s biggest clients are aiming to switch to in-house AI chips within the next two years. Google LLC plans to deploy Axion, its new Arm Holdings plc-based CPU that outperforms x86 chips, across its cloud services.
According to industry watchers, the overdue correction signals that the firm’s AI bubble is about to burst. Nvidia has lost 5.54% in April alone as tech firms continue to provide updates regarding their respective AI chip projects.
Intel Gaudi 3 Beats Nvidia H100 in LLM Training
During Tuesday’s Vision event, Intel Vice President of Strategy and Product Management Jeni Barovian said Gaudi 3 can process LLMs 50.00% faster than H100. In addition, the former can accelerate the inferencing of GenAI systems quicker than the latter in most Intel systems.
The chipmaker built the Gaudi 3 using Taiwan Semiconductor Manufacturing Co.’s (TSMC) 5nm process, which Apple Inc., Huawei Technologies Co. Ltd., and Qualcomm Inc. also utilize. The next-generation AI chip uses two fused main processors and is twice as fast as the Gaudi 2.
Barovian stressed that Gaudi 3’s superior head-to-head performance provides consumers with a choice, which has been lacking in the advanced AI chip market. Last year, Nvidia controlled almost 83.00% of the global data center chip market, leading to severe scarcity.
