Chinese chip companies broke through the encirclement
Over the years, NVIDIA has held a monopoly in the field of GPUs and AI chips, and its CUDA ecosystem has become the de facto standard in the industry. In recent years, Chinese chip companies have finally made breakthroughs in this field through independent research and development.
China's local chip companies, such as Huawei, Horizon Robotics, Cambrian, etc., have successfully developed domestic GPUs and AI chips with excellent performance through continuous R&D investment. These chips not only benchmark or surpass NVIDIA's similar products in key indicators such as computing power and energy efficiency, but also have more competitive prices due to the advantages of localization.
Taking Huawei's Ascend AI chip as an example, its single-precision floating-point computing capability is as high as 768TFLOPS, surpassing NVIDIA's flagship A100 in all respects. In terms of inference performance, the Ascend AI processor is also more than twice that of NVIDIA. What's even more exciting is that the price of the Ascend chip is only about half that of the A100.
The Journey series GPU launched by Horizon is the first fully autonomous and controllable general-purpose GPU in China. The journey GPU adopts the 7nm process and has powerful parallel computing capabilities, which can be widely used in artificial intelligence, high-performance computing and other fields. Compared with NVIDIA's RTX 30 series, the journey GPU has better performance in application scenarios, but the price is only about half of the latter.
The breakthroughs of these Chinese chip companies indicate that the mainland is breaking the foreign monopoly in the field of GPU and AI chips and moving towards independent and controllable.
Nvidia's CUDA ecological conundrum
Although Chinese chip companies have made great strides in GPU and AI chip hardware, in order to truly break the monopoly of NVIDIA, they must overcome the "soft underbelly" of the CUDA ecosystem.
CUDA is a GPU-accelerated computing platform launched by NVIDIA that has become the de facto standard for AI software development. The vast majority of AI frameworks, development tools, and application software are built on top of the CUDA ecosystem. This makes it difficult to directly run existing AI software even with domestic GPU/AI chips.
In order to solve this problem, Chinese chip companies are working hard to achieve CUDA compatibility. For example, Horizon's Journey GPU claims to be able to seamlessly support the CUDA ecosystem. Nvidia has made it clear that it will prohibit any chip from being compatible with CUDA.
This is undoubtedly a "green barrier" deliberately set up by Nvidia, intended to maintain its monopoly position in the AI chip market. Faced with NVIDIA's "ecological lock-in", Chinese companies have no choice but to independently develop open-source alternatives compatible with CUDA.
The U.S. government is also "escorting" Nvidia. Last year, the U.S. Department of Commerce banned chip companies such as Nvidia from exporting their latest AI-accelerated chips to China, cutting off access to Nvidia's high-end chips for Chinese companies.
This will undoubtedly further accelerate China's independent and controllable process in the field of GPU and AI chips, prompt domestic enterprises to increase R&D investment, and break through the shackles of the CUDA ecosystem as soon as possible.
U.S. chip giants join forces to fight monopoly
What Nvidia did not expect was that in addition to the challenge of Chinese chip companies, its monopoly position in the AI chip market was also jointly counterattacked by local chip giants in the United States.
Intel, AMD, Google, Qualcomm and other companies established the UXL Alliance last year with the aim of building an open and diverse AI chip ecosystem and reducing dependence on NVIDIA CUDA.
These companies believe that Nvidia's "dominance" in the field of AI chips has seriously distorted the market order and hindered industry innovation. They hope to promote the openness of the AI chip ecosystem through the UXL Alliance, eliminate vendor lock-in, and allow users to freely choose chips and software from different vendors.
One of the goals of the UXL Alliance is to develop a unified instruction set and programming model for AI chips to replace NVIDIA's CUDA. Once the UXL ecosystem takes shape, it will directly shake NVIDIA's dominance in the AI chip market.
In the future, NVIDIA's "unique seedling" position in the field of AI chips will no longer exist. It will have to face a double squeeze from Chinese chip companies and local competitors in the United States.
Nvidia's supremacy is under threat
Looking at the latest developments in the GPU and AI chip industries, NVIDIA's long-standing "one dominance" situation is being broken. For the first time, its monopoly position in this field has been challenged by China and the United States.
On the one hand, Chinese chip companies have made breakthroughs in GPU and AI chip hardware through independent innovation, and their product performance is not inferior to or even surpasses NVIDIA. On the other hand, local chip giants in the United States have also joined forces with anti-monopoly to try to share the cake at the level of software ecology.
Under this internal and external attack, NVIDIA's AI chip supremacy is being severely tested. The GPU and AI chip markets will present a diversified pattern, and NVIDIA will have to face fierce market competition.
As an old giant in the industry, NVIDIA's technology accumulation and ecological advantages in the field of GPU and AI chips should not be underestimated. But if you rest on your laurels and follow the monopoly, you will be abandoned by the times.
Only by actively embracing openness, joining the construction of the industry ecosystem, and actively participating in fair competition can NVIDIA remain invincible in the new round of GPU/AI chip revolution. Otherwise, its hegemony will not be guaranteed.