Nvidia's dominance in the realm of cutting-edge AI chips is facing potential challenges as tech giants, including OpenAI, unveil their custom-designed semiconductors, experts report to CNBC.
On Tuesday, OpenAI revealed the Jalapeño chip, claiming it delivers "industry-leading speed and efficiency." In addition to OpenAI, major players like Google, AWS, and Meta are also in the process of architecting their own AI chips.
Nvidia's stock price has soared in response to a surge in demand for its chips, driven by the ongoing expansion of data centers essential for both model training and the operational tasks of AI systems.
However, major cloud service providers and AI firms are rapidly advancing their own chip development to support AI functionalities. The Jalapeño chip, particularly focused on inference tasks, demonstrates that chips designed by hyperscalers can rival—or even surpass—Nvidia's Blackwell-class GPUs in efficiency for inference, according to Adrien Sanchez, a technology analyst from Yole Group.
Sanchez noted that while Nvidia maintains a significant lead in AI computing and benefits from the established user base of its software platform, CUDA, OpenAI's new chip poses a challenge to Nvidia's profit margins in inference, which is currently the fastest-growing segment.
OpenAI has confirmed they are collaborating with Broadcom on this initiative, aiming to integrate the Jalapeño into their computing architecture by the year's end. They also disclosed their ongoing efforts in the development of the subsequent generations of this semiconductor.
