Ji Dong

Author | Cheng Qian

Editor | Heart

Zhidong reported on 27 on 0, according to foreign media The Information,Nvidia plans to acquire Lepton AI, a startup from AI giant and former vice president of Alibaba Cloud, for hundreds of millions of dollars。

Lepton was founded in 2023 years by Jia Yangqing, a big star in the AI field known as the "father of Caffe", after leaving Alibaba, and its main business isRent out NVIDIA GPU servers，開發軟體幫助創企在雲中構建和管理自己的應用。該公司於2023年5月完成了1100萬美元（摺合人民幣約7900萬元）天使輪融資。

Foreign media speculated that the reason for Nvidia's acquisition of Lepton may be,Trying to enter the cloud and enterprise software market, competing with major cloud service providers such as AWS and Google。 AWS, Google Cloud, and others, as Nvidia's largest customers, are currently trying to weaken Nvidia by developing and leasing replacement chips at low prices.

Since last year, Nvidia has been revealed to have been acquiredFour start-ups, including Israeli GPU startup Run:AI, Israeli AI startup Deci, and Seattle AI startup OctoAI, all of which are in the business of reducing the cost of developing or running AI models powered by NVIDIA chips. Recently, Nvidia was revealed to have acquired Gretel, an American synthetic data company that helps customers create and use AI-generated data to develop and test AI applications.

1. The main role is "GPU dealer", with only 20 employees

Lepton gets its name from a basic substance particle lepton, and its main business is to rent out NVIDIA GPU servers, and the company does not manage data centers or servers itself, but rents them from cloud providers and then subleases them to other customers. At present, the number of employees is approximately20 peopleAround.

Lepton is known as a GPU reseller, and its direct competitors are startups like Together AI, which develop software to help other startups build and manage their own AI applications in the cloud.

據投資者人透露，Together的年化收入已經超過1.5億美元，每月達到1250萬美元。不過與Together相比，Lepton的規模較小，可能不會產生太多收入。

Lepton has revealed in a blog post that its clients include gaming startup Latitude.io and scientific research startup SciSpace, which uses Lepton's platform to run AI models for services used by hundreds of thousands of monthly active users.

The co-founders of Lepton are Jia Yangqing, Bai Junjie, and Li Xiang.

Yangqing Jia graduated from Tsinghua University and received his Ph.D. in Computer Science from the University of California, Berkeley. During his Ph.D., Jia Yangqing developed the deep learning framework Caffe (Convolutional Architecture for Fast Feature Embedding), which became one of the most popular open-source deep learning frameworks in the world. With outstanding results, Jia Yangqing entered Google Brain as an intern after graduation and participated in the development of the TensorFlow framework.

From 3 to 0 years, Jia Yangqing served as the president of the Intelligent Computing Platform Business Unit of Alibaba Cloud, mainly responsible for big data computing and artificial intelligence platforms. In 0/0, Jia Yangqing announced that he would leave Alibaba Cloud and subsequently founded Lepton.

Among the other two founders of Lepton, Bai Junjie was the former director of AI platform of Alibaba Cloud and the MetaAI Research Institute, and Li Xiang is the former director of engineering of Alibaba Cloud and worked at Meta and Uber.

二、英偉達雲業務發展受挫，曾預計雲業務收入超1500億美元

Nvidia's intention to acquire Lepton may be related to the development of its cloud and software business.

Nvidia's cloud and software business is in its infancy, and its current main business includes renting servers powered by NVIDIA GPUs directly to enterprises, providing software to help companies develop AI models and applications, and providing a cluster of graphics processing units to manage and train AI.

英偉達此前曾告訴投資人，從長遠來看，英偉達將從雲服務和軟體業務中獲得1500億美元的收入 ，比英偉達、AWS目前的年收入還要多。

However, during Nvidia's final quarterly earnings call this month, a spokesperson avoided the topic entirely.

In the past few years, cloud service vendors have purchased about half of Nvidia's AI server chips. But Nvidia CEO Jensen Huang believes that over time, he expects other types of companies to buy chips and sales will be much larger than cloud service providers.

As a result, he suggested that companies such as automakers could one day buy AI chips directly instead of renting them from cloud vendors.

Conclusion: Nvidia has increased its cloud service layout

In fact, Nvidia has been acquiring small AI and cloud startups to reduce the cost for developers to run AI models with its chips.

Nvidia's chips have previously suffered from high prices and high demand, which has made it difficult for many companies to develop AI applications at lower value. This undoubtedly proves that Nvidia is increasing its layout in the cloud and software fields.