Earlier this week, Sambanova Systems, a cloud and data warehousing platform, announced its new funding round. This includes a new machine learning and big data platform, as well as Dataflow-as-a-Service offering.
Machine learning and big data platform
Using a software-defined hardware platform, SambaNova has built a machine learning and big data platform that has a little bit of everything. This includes a reconfigurable “dataflow” architecture, an open source stack called SambaFlow, and a next-generation processor. Using this combination, SambaNova can provide high-performance data processing, model training capabilities, and efficiency across applications. In addition to the best-in-class AI performance, SambaNova’s systems also offer superior power consumption and cost-effectiveness when compared to its competitors.
SambaNova’s dataflow-as-a-service is the industry’s most powerful integrated AI software-defined hardware platform. This offering includes pre-trained deep learning models, which enables enterprises to tap into SambaNova’s AI system. This allows organizations to focus on business objectives, while accelerating the data center. SambaNova’s flagship DataScale system is available for purchase, rent, or leasing. Among its features are customized seven-nanometer chips that more closely align with machine learning processes.
SambaNova’s Dataflow-as-a-Service is a cloud-based solution that accelerates existing data centers and enables organizations to capitalize on their artificial intelligence (AI) capabilities. This is done through a subscription-based service that focuses on applications. The company claims that this is the most efficient way to implement AI-powered solutions. As a result, the cost of the AI solutions is less than one-tenth of what customers pay for conventional systems. The benefits of the service include enhanced AI productivity, a 6x performance advantage, and a greater yield on massive exploration investments.
Dataflow-as-a-Service offering
Founded in December 2015, SambaNova Systems is a startup that is building artificial intelligence hardware and software. Their initial focus was on chip solutions for machine learning. However, they soon moved into dataflow-as-a-service. This new model allows them to offer a full stack solution that includes their hardware, software, and APIs. This means they can charge customers based on how much they use the system.
SambaNova’s flagship DataScale system uses an open source, reconfigurable “dataflow” architecture. This eliminates the need to interface with RAM, which is a bottleneck for some companies. It also runs an open source software stack called SambaFlow. The benefits include speed, scalability, and a reduction in complexity. The company has already sold its solution to several biopharma companies.
SambaNova’s tensor streaming processors are designed to handle AI workloads. They have a simpler architecture than CPUs and provide six petaops of performance. They consume 3.3kW of power and are available on demand through the Nimbix Cloud. The system is capable of analyzing billions of data points and providing results in just a few minutes.
SambaNova has a strong enterprise business as well. They compete against companies like Cerebras Systems and Nvidia. They recently closed a $676 million round of funding. Their plan is to sell software-based AI systems to enterprises. They hope this will increase their revenue.
New funding round
Founded by Stanford Professors in December of 2017, SambaNova Systems is a startup that is developing semiconductors for AI workloads and building software to help enterprises use those chips. The company plans to build a new generation of AI hardware and software that will break the computational limits of current hardware and software.
SambaNova’s dataflow architecture automatically extracts the optimal dataflow graph for models. It also offers a development interface that developers can use to create models more easily. SambaNova’s DataScale hardware platform features next-generation processors and optimized algorithms, allowing organizations to deliver high-performance computing to AI applications. The hardware can be purchased or rented through SambaNova’s Dataflow-as-a-Service service, which is a subscription-based, cloud-based solution that accelerates existing data centers.
SambaNova’s flagship chip uses 40 billion transistors and is built on Taiwan Semiconductor Manufacturing Company’s N7 process. It is designed for training and accelerating deep learning workloads. Its tensor streaming processors offer six petaops of performance, consuming 3.3kW of power. SambaNova is currently testing its AI system with the Department of Energy.
Conclusion
SambaNova is targeting the enterprise market, with its flagship offering being Dataflow-as-a-Service. It has already signed several biopharma companies and is in phase one trials for solid tumors. The company’s workforce has grown to more than 300 and the company plans to grow to more than 500 by the end of the year.