Tech Giants Dash to Include DeepSeek’s AI Into Their Offerings

In a rapidly evolving technological landscape, U.S. and Chinese tech giants are racing to integrate the large language models developed by DeepSeek, a Chinese firm, into their products. This indicates the significant impact DeepSeek's emergence has had on the global AI sector. As of Sunday, leading companies such as Nvidia Corp., Microsoft Corp., Amazon Web Services Inc. (AWS), along with the cloud-computing divisions of Huawei Technologies Co. Ltd. and Tencent Holdings Ltd., have incorporated DeepSeek’s AI models to enhance their offerings and attract a larger developer and subscriber base. [para. 1][para. 2]

Microsoft spearheaded the advancements by announcing the integration of DeepSeek-R1 into its Azure AI model library and embedding this model into its Copilot chatbot. Following this, AWS integrated R1 into its machine learning platforms for developers. Nvidia also followed suit by adding R1 to its Nvidia NIM service, underscoring its role in DeepSeek’s development and emphasizing its GPU's scalability and the efficiency required for handling AI workloads. [para. 3][para. 4]

Moreover, Huawei Cloud has partnered with Beijing-based AI startup SiliconFlow to provide access to DeepSeek’s models via its Ascend cloud service. Meanwhile, Tencent Cloud has enabled developers to deploy DeepSeek-R1 through its HAI app-making platform efficiently, marking a significant move in the tech domain. [para. 5]

DeepSeek, a company from Hangzhou, disrupted the market by launching its open-source AI model R1 on January 20, which stands in competition with OpenAI and Meta Platform Inc.'s models. Interestingly, the training of DeepSeek-V3, the predecessor of R1, required significantly fewer resources—2048 Nvidia H800 chips and an investment of $5.58 million—compared to what OpenAI and Meta expended on similar models. [para. 6][para. 7]

The rapid adoption of DeepSeek's AI chatbot app, which topped download charts in a very short time, further highlights the startup's swift ascent. From January 10 to 24, the app saw 1 million downloads across major app stores, surpassing notable competitors like OpenAI’s ChatGPT and ByteDance Ltd.’s Doubao AI chatbot. This unexpected rise impacted major U.S. tech companies, as reflected in the plunging stock prices and Nvidia losing around $590 billion in market capitalization. [para. 8][para. 9]

DeepSeek’s remarkable performance at a significantly lower cost is seen as a testament to the evolving maturity of Chinese tech firms. Founded by Liang Wenfeng, DeepSeek has not only demonstrated technological proficiency but also highlighted Chinese firms' resilience against U.S. restrictions. [para. 10]

Nevertheless, the company is under scrutiny as there are speculations it might have developed its model by breaching rules. OpenAI has raised concerns about the possibility of DeepSeek using distillation—a method to replicate larger models using smaller, more efficient ones—potentially violating OpenAI's terms of service. Adding to this scrutiny, U.S. officials are investigating whether DeepSeek circumvented U.S. sales restrictions on advanced AI chips by acquiring Nvidia semiconductors through third parties in Singapore. [para. 11][para. 12]

Industry insiders attribute DeepSeek’s cost-effectiveness to a fast-follower advantage. They argue that later entrants in the AI model development space benefit from the trial and error of predecessors, significantly reducing training costs while yielding comparable results. This strategy allows firms like DeepSeek to emerge as formidable competitors despite limited resources. [para. 13][para. 14]