ASICs are emerging as a more efficient option for inference tasks compared to GPUs, a trend reminiscent of the cryptocurrency mining industry. The market for Inference AI chips is predicted to witness significant growth by the end of the decade, with tech giants like Google already making strategic moves in this direction.

Nvidia, a prominent player in AI and GPU technologies, is venturing into the Application-Specific Integrated Circuit (ASIC) market to tackle the evolving landscape of AI semiconductor design and rising competition. The demand for GPUs has surged due to the widespread adoption of generative AI and large language models (LLMs), prompting Nvidia to hire 1000 engineers in Taiwan by 2024. Recent reports from Taiwan’s Commercial Times reveal that Nvidia has formed a new ASIC department and is actively recruiting talent to capitalize on this market shift.

While Nvidia’s H series GPUs have been favored for AI model training, the industry is shifting towards ASICs for inference tasks, driven by the need for chips optimized for real-world AI applications such as large language models and generative AI. Unlike GPUs, ASICs offer enhanced efficiency for inference tasks and even cryptocurrency mining. Market projections by Verified Market Research indicate a substantial growth in the inference AI chip market, from $15.8 billion in 2023 to $90.6 billion by 2030. Leading tech companies like Google have already incorporated custom ASIC designs in products like the AI chip “Trillium”, marking a significant industry shift.

This transition to custom AI chips has intensified competition among semiconductor giants, with companies like Broadcom and Marvell gaining traction through collaborations with cloud service providers to develop specialized chips for data centers. To stay competitive, Nvidia’s new ASIC department is focusing on recruiting top talent with local expertise from companies like MediaTek.