During the Chinese Entrepreneurs Forum 2023, Liu Qingfeng, the visionary founder of HKUST Xunfei, a prominent Chinese AI firm, disclosed that his company has collaborated with Huawei to create an AI Graphics Processing Unit (GPU) that rivals the performance of NVIDIA’s renowned A100 GPU. This revelation marks a crucial advancement in China’s AI industry, potentially altering the dynamics of AI hardware market share. Also, Liu’s announcement hinted at a forthcoming AI language model, one that will be capable of rivalling the much-acclaimed GPT-4.
Pioneering Chinese collaboration leads to AI GPU on par with NVIDIA’s A100
Liu Qingfeng, in his address to the forum attendees, emphasized the collaborative efforts between HKUST Xunfei and various other Chinese AI solution providers, notably Huawei. This partnership has culminated in the development of an AI GPU that boasts AI performance capabilities comparable to NVIDIA’s A100 GPU. Although specific technical details regarding the GPU’s specifications were not disclosed, insiders have indicated that it possesses impressive capabilities. Particularly, the GPU has demonstrated its prowess in efficiently handling the computational demands of large language models, such as GPT-3 and the anticipated GPT-4.
As of now, NVIDIA’s A100 GPU stands as one of the leading AI GPUs available, with only the Hopper H100 surpassing it in performance, albeit amidst high demand. The A100 is known for its capacity to deliver 624 Tera Operations Per Second (TOPs) of Integer 8-bit (INT8) compute power. In comparison, the Hopper H100 boasts a remarkable 2000 TOPs. Should the claims regarding Huawei’s AI GPU hold true, it could potentially influence the distribution of AI hardware demand within the Chinese market. While NVIDIA currently dominates the Chinese AI GPU market, fueled by robust demand, the emergence of a domestic alternative might present an intriguing shift.
Liu Qingfeng, the Founder and Chairman of HKUST Xunfei, conveyed his excitement during a recent address, noting that Huawei’s GPU capability now stands at par with that of Nvidia’s A100. He highlighted the significance of this achievement, revealing that Ren Zhengfei, Huawei’s prominent figure, has placed considerable importance on this advancement. Liu also pointed out that three directors from Huawei, who were once part of the specialized class at iFlytek within HKUST, have played a pivotal role in attaining the same level of performance as Nvidia’s A100.
Elevating the AI language model game by developing a GPT-4 competitor
Beyond the groundbreaking AI hardware announcement, HKUST Xunfei is gearing up to introduce a novel and versatile Large Language Model (LLM) by October 24th. This LLM is being positioned as a direct contender to the popular ChatGPT model and is set to launch in both Chinese and English versions. While the company acknowledges that it might not initially match ChatGPT’s performance upon launch, it remains steadfast in its commitment to enhancing its AI software. The ultimate goal is to develop an AI language model that can stand shoulder-to-shoulder with GPT-4, a feat they anticipate achieving by the first half of 2024.
Huawei’s commitment to AI innovation is further underscored by its recent partnership with iFlytek to develop All-in-One (AIO) machines tailored for AI applications. These machines are aimed at catering to the growing demand for general-purpose AI solutions. In addition, Huawei and iFlytek are integral members of the China Mobile Metaverse Industry alliance, a consortium that includes other industry giants, collectively driving advancements in the AI and tech landscape.
As the Chinese AI industry continues to evolve, these developments hold immense potential for both domestic and international markets. The emergence of an AI GPU on par with NVIDIA’s A100 and the promise of a GPT-4 competitor signal China’s escalating prowess in AI hardware and software, setting the stage for a dynamic and competitive future.
From Zero to Web3 Pro: Your 90-Day Career Launch Plan
