Tech Giants Support DeepSeek Ecosystem Rise!

Recently, DeepSeek, an AI technology company from China, has gained significant attention in the global tech industry with its AI models. Initially met with skepticism, leading tech giants have now recognized DeepSeek's AI capabilities, offering their support for the models.

DeepSeek's V3 and R1 open-source AI models have disrupted industry perspectives on "AI computational power requirements." Specifically, the DeepSeek-R1 inference model not only competes with OpenAI's O1 model in performance but also requires only a fraction of the training costs—approximately 1/20th of OpenAI's. For instance, V3 used just 2048 H800 GPUs and completed training in two months for a cost of about $5.58 million, while its API pricing is roughly 1/28th that of OpenAI, reducing usage costs by nearly 97%.

In summary, DeepSeek has achieved results comparable to leading US AI technology companies, such as OpenAI, by utilizing less advanced AI chips with lower computational requirements and reduced costs. This achievement is seen as a challenge to the United States' AI leadership and has caused concern among companies like OpenAI, Microsoft, Meta, and Google, as well as a revaluation of AI chip companies like NVIDIA.

OpenAI and Microsoft publicly announced investigations into potential inappropriate "Knowledge Distillation" practices by DeepSeek. Distillation is a technique that transfers knowledge from a larger model to a smaller one. OpenAI clarified that while distillation does not expose the model's internal workings, it may improve applications. However, OpenAI's terms prohibit the use of distilled data to develop competing AI products.

Public opinion soon shifted. OpenAI CEO Sam Altman publicly acknowledged that "DeepSeek is indeed a very good model," and stated that OpenAI would develop better models, but that it would no longer maintain such a significant lead over competitors. Altman also admitted a misstep on OpenAI’s part regarding open-source strategy discussions, noting that OpenAI needed to adopt a new approach, as not all within the company shared the same view.

Meanwhile, major cloud service providers, including Amazon AWS, Microsoft, Alibaba Cloud, and others, launched cloud services based on the DeepSeek-R1 model. AI chipmakers like NVIDIA, Intel, and AMD also announced their support for DeepSeek models.

On January 30, Amazon AWS announced that users could deploy DeepSeek-R1 on Amazon Bedrock and Amazon SageMaker, offering superior service and support. Soon after, Microsoft also made DeepSeek-R1 available on Azure AI Foundry and GitHub, allowing developers to test and build cloud-based applications.

On February 1, Huawei Cloud announced a partnership with Silicon Flow, launching DeepSeek R1/V3 inference services based on Ascend Cloud, with performance comparable to global high-end GPU models. Tencent Cloud followed suit on February 2, offering one-click deployment of the DeepSeek-R1 model on its HAI platform, enabling developers to access it in just three minutes. Baidu AI Cloud also joined the trend on February 3, integrating DeepSeek-R1 and V3 models on its Qianfan platform with affordable pricing options and limited-time free services.

In the same period, ByteDance's Volcano Engine and JD Cloud & AI also began supporting the DeepSeek models, offering both public and private cloud deployment options for fast access. On February 5, China Unicom unveiled a comprehensive cloud service, providing DeepSeek-R1 model deployment through its Xingluo platform, ensuring full operational support and flexibility for users.

In the chip sector, leading manufacturers quickly adapted their products to support DeepSeek's models. On January 31, NVIDIA announced DeepSeek-R1's integration into NIM, achieving impressive processing speeds on a single NVIDIA HGX H200 system. AMD, a key competitor, followed by integrating DeepSeek-V3 into its Instinct MI300X GPUs optimized for AI inference.

Intel also announced support for DeepSeek models on its Core Ultra 200H AI PCs and Gaudi 2D AI accelerators, aimed at reducing costs and enhancing efficiency for developers. Additionally, Sugon, a leading Chinese tech company, confirmed that its Hygon team had completed integration of DeepSeek's models on Sugon's DCU platform.

Several other chipmakers, including Moore Threads and Iluvatar CoreX, have also completed integration with DeepSeek models, showcasing the broad industry adoption.

The ecosystem surrounding DeepSeek's AI models continues to expand rapidly. On February 5, Hubei Century Network Technology announced it was testing DeepSeek models for integration with its social products, aiming to lower computational costs. Wondershare Technology also revealed that it had integrated DeepSeek-R1 into its software products, such as Filmora and MindMaster, enhancing them with AI capabilities.

Various other companies, including eclicktech and Arcvideo Technologies, have incorporated DeepSeek's AI models into their services, advancing innovation in multiple industries, from media to AI-driven marketing. Furthermore, 360, a Chinese technology giant, confirmed that some of its products had integrated DeepSeek models under the MIT open-source protocol, marking a further milestone in the growing adoption of DeepSeek's AI technology.