DeepSeek-R1 AI Model Hosted on Huawei’s Platform, Sparking Speculation About Training Hardware
The recent announcement about the DeepSeek-R1 artificial intelligence (AI) model being hosted on Huawei’s ModelArts Studio platform has ignited speculation. People are curious about the hardware used to train the model. Huawei has not officially disclosed the specific chipsets powering the AI. However, a tipster claims that Ascend 910C GPUs are being used. This has fueled discussions about the possibility that DeepSeek also trained its models on Huawei’s hardware. This development comes amidst allegations from OpenAI that DeepSeek used its proprietary models for training.
Tipster Alexander Doria (@Dorialexander) shared a promotional image on X (formerly Twitter) from Huawei. The image highlights the “Ascend-adapted” nature of its ModelArts Studio platform. Huawei’s description suggests the platform’s data centers utilize its Ascend series chipsets. However, the company has remained tight-lipped about the precise GPUs employed. Doria claims that the DeepSeek LLM’s inference is powered by the Ascend 910C chips. These chips are considered a competitor to Nvidia’s H800. This claim has added fuel to the fire.
The timing of DeepSeek-R1’s release, just a week prior to this announcement, raises intriguing questions. AI models are typically optimized for the same hardware used during training. While it’s possible to optimize for other GPUs, this is often a lengthy process. Huawei can run the model’s inference on its Ascend platform. This ability suggests the possibility that the same infrastructure was used for training, although it does not confirm it. Currently, no definitive proof exists to establish this connection.
The “black box” nature of DeepSeek-R1’s release has contributed to the swirling speculation. Despite its open-source release, DeepSeek only provided the model weights, withholding details about the datasets and training process. Claims by DeepSeek about the model’s low training cost (approximately $6 million) have drawn scrutiny from industry experts. These experts have raised concerns about the methodologies employed. Adding to the complexity, OpenAI has alleged that DeepSeek used its proprietary models to train DeepSeek-R1. This allegation remains under investigation. The combination of these factors has created an atmosphere of uncertainty surrounding the model’s development and training. The connection between DeepSeek-R1, Huawei’s hardware, and OpenAI’s allegations remains a developing story.
Leave a Reply