A major shift in power is brewing in the artificial intelligence industry. According to fresh data from the analytics agency Fubon Research, Google could surpass its main rival, Nvidia, in terms of AI accelerator sales as early as 2028.
Ambitious Forecasts and Figures
The search giant is demonstrating confidence in its capabilities, planning to sell between 12 and 15 million of its own chips. For comparison, current forecasts for Nvidia show sales volumes of 8.2 million accelerators this year. It is expected that by 2028, this figure will grow to 12.4 million units. If Google realizes its plans, the company will take the lead in the number of chips sold.
The discussion concerns Tensor Processing Unit (TPU) chips of the ninth generation. It is these specialized processors that will serve as the foundation for performing heavy computations in neural networks.
Specifics of Competition
It is important to understand that a direct comparison of sales volumes has its nuances. The architecture of Google and Nvidia chips differs radically, and the companies' business models are not identical. Nvidia traditionally sells its solutions to a wide range of clients, whereas Google often uses its accelerators within its own cloud service infrastructure, although it does offer them to external customers.
Production Challenges
To achieve the stated figures, Google will need to solve a serious production task. Currently, chips for the company are manufactured exclusively by the Taiwanese TSMC. Analysts note that to ensure the production of 12–15 million units, Google will likely have to involve another partner in production — Intel. Without expanding production capacity, the realization of such ambitious plans could be at risk.
