Thanks to the surge in generative AI applications, tensor processing unit (TPU) technology is experiencing substantial growth, heralding as a transformative force with lasting implications.
Despite big players currently reigning supreme in the market, startups are emerging as formidable challengers, as indicated by insights from GlobalData's innovation intelligence platform, Technology Foresights.
Initially crafted by Google to power its TensorFlow framework, TPUs have evolved into specialized hardware designed to expedite the processing of AI and machine learning models. Beyond their original scope, TPUs now encompass a broader spectrum of AI accelerator chips utilized for executing neural networks and diverse machine learning tasks, spanning from autonomous driving to drug discovery.
Read More: Chips 2.0: Meeting Escalating Demands Amid Global Turbulence
US Leads AI Accelerator Chip Innovation
At the forefront of AI accelerator chip innovation is the United States, with China and South Korea closely trailing behind. In China, Huawei leads the charge with its Ascend series of AI accelerator chips, closely followed by Alibaba, Baidu, and other major players.
Additionally, Meta Platforms, a frontrunner in AI accelerator chips, recently unveiled its own chip after making strides in advancing its large language model Llama for generative AI applications. On another front, Samsung secured a significant $750 million contract from Naver Corp. to supply edge AI accelerator chips, further solidifying its position in the market.
According to GlobalData's Technology Foresights, the landscape of AI accelerator chips is witnessing dynamic shifts, with 17 new companies entering the market and filing patents in the past year alone. Notably, innovation is pivoting towards enhancing AI processes on edge devices, with 24% of all patents emphasizing edge computing applications, followed by autonomous driving and other solutions.
Sourabh Nyalkalkar, Practice Head of Innovation Products at GlobalData, highlighted the evolving dynamics of the market, stating, “While the market is dominated by large players like Google, Nvidia, and Intel, there are clear signs of increasing disruption by startups in this space.”
Read More: From Innovation to Responsibility: AI Deployment
Startup Domination
Currently, startups have significantly contributed to AI accelerator chip technology, accounting for 20% of all patents filed, indicating a growing disruption in the market. Among the innovative startups are Hailo, an Israeli company specializing in AI accelerator chips for edge devices; Cortica, a leader in developing AI processors for autonomous driving applications; and Femtosense, recognized for crafting specialized AI chips for small, energy-efficient electronic devices.
OpenAI, a prominent generative AI company, is also reportedly exploring potential acquisitions to diversify its reliance away from Nvidia.
Nyalkalkar emphasized the robust growth opportunity for AI accelerator chips, citing nearly USD 4 billion raised by specialized startups between 2021 and 2023, coupled with significant scale-up efforts by industry giants.
In light of this, industry stakeholders are advised to remain vigilant, monitoring the landscape for potential mergers and acquisition opportunities in the near term.
