Nvidia, renowned for its dominance in AI and GPU technologies, is now venturing into the ASIC (Application-Specific Integrated Circuit) market to address evolving trends in AI chip design and intensifying competition.
The surge in generative AI and large language models (LLMs) has spiked global GPU demand. Nvidia CEO Jensen Huang announced plans to hire 1,000 engineers in Taiwan in 2024, and the company has now established a dedicated ASIC department, actively recruiting talent, according to Taiwan’s Commercial Times.
While Nvidia’s H series GPUs remain widely adopted for training AI models, the market is increasingly shifting toward inference chips, or ASICs. These chips are optimized for real-world AI applications, offering better efficiency for tasks like deploying LLMs and generative AI, as well as cryptocurrency mining.
The market potential is immense. Verified Market Research predicts that the inference AI chip market, valued at $15.8 billion in 2023, will grow to $90.6 billion by 2030. Industry players like Google have already introduced custom ASIC solutions, such as its “Trillium” chip, which became generally available in December 2024.
This shift has heightened competition among semiconductor companies. Firms like Broadcom and Marvell have gained traction, collaborating with cloud providers to create custom data center chips. Nvidia’s new ASIC division is poised to leverage expertise from Taiwan’s semiconductor talent pool, including professionals from MediaTek, to stay competitive.
