IT之家 News on November 11: NVIDIA's lead in the CUDA ecosystem may be facing an increasingly large challenge. A slide from Anthropic's IPO roadshow materials shows that as high-performance AI models gradually gain the ability to autonomously adapt to different computing platforms, GPUs are becoming more and more like interchangeable, general-purpose computing resources.
IT之家 has noted that Anthropic's IPO roadshow materials revealed a piece of rather eye-catching information: its Claude Mythos Preview model took only about 3 days to figure out how to get another model running on AMD MI355X GPUs. By comparison, under traditional circumstances, the adaptation work needed to run inference tasks on a new computing platform typically takes 6 to 12 months.
This means Anthropic can now treat GPUs from different vendors as relatively general-purpose computing resources, and the technical barrier to switching between platforms has been greatly reduced. Precisely because of this, some are beginning to believe that NVIDIA's long-prided CUDA ecosystem moat may in fact have lost its original meaning.
However, it may still be too early to declare the end of NVIDIA's CUDA ecosystem advantage. CUDA is not merely a set of software tools; behind it lies a massive developer ecosystem, rich libraries and development tools, as well as technical accumulation refined through years of continuous optimization. A large number of developers still build and maintain all kinds of applications around CUDA.
Over the past few years, NVIDIA has also been taking various measures to consolidate CUDA's competitive advantage. For example, the company provides GPUs and training resources to universities and researchers for free, attracting more developers into the CUDA ecosystem.
In addition, NVIDIA has deeply integrated NVIDIA Collective Communications Library (NCCL), a communication toolkit used to connect multiple GPUs, with proprietary hardware technologies such as NVLink and InfiniBand, further improving the efficiency of multi-GPU collaborative computing while also increasing the cost for users to migrate to other platforms.