Recently, GPU cloud computing company Lambda announced the launch of its new 1-Click cluster service, allowing customers to access Nvidia H100GPUs and Quantum2InfiniBand clusters on demand. This innovative service enables businesses to obtain computational power only when needed, particularly suitable for companies that do not require 24/7 GPU usage.
Image source note: The image is AI-generated, provided by the image licensing service Midjourney.
Robert Brooks, co-founder and vice president of Lambda, stated in an interview that this service is designed for companies building their own models. This allows them to quickly obtain the necessary GPU computing power only when needed, avoiding the hassle of complex hardware and software coordination in the past. He mentioned that starting an AI company requires significant capital and time, especially when configuring a large GPU cluster.
Unlike traditional GPU rental models, Lambda's service eliminates the complex process of negotiating usage times with GPU suppliers. Brooks pointed out that Lambda has found a solution to make Nvidia H100GPUs and Infini networks available at any time for new model training projects.
Many companies prefer to use pre-trained AI models and fine-tune them with their own data management policies due to the high cost of building and training foundational models. Moreover, Nvidia GPUs are expensive, and many companies do not have the capability to establish their own data centers to house these GPUs. For small AI enterprises, they typically have to choose between renting GPU space or partnering with cloud service providers to obtain the processing power needed for model training.
Although cloud service providers like AWS and Microsoft Azure offer access to AI and Nvidia GPUs, Brooks emphasized that their focus is on the inference phase of models, not the training phase. Traditional GPU cluster contracts often require signing for a year or longer.
In contrast, the 1-Click cluster is specifically designed for companies that only need short-term models, especially those that do not want to sign long-term contracts. Customers can book the required nodes as needed, with a minimum booking period of two weeks. The specific pricing varies depending on the number of nodes and the required time.
Lambda was founded in 2012 and successfully raised $320 million in funding this February, with the company's valuation reaching $1.5 billion.
Key Points:
- 💻 Lambda introduces the 1-Click cluster service, allowing users to access Nvidia H100GPUs on demand, breaking the constraints of long-term contracts.
- ⏳ This service is designed for enterprises needing short-term model training, with a minimum booking period of two weeks.
- 💰 Lambda raised $320 million this year, with a valuation of $1.5 billion, indicating strong market demand for its services.