Integrated systems, new hardware, and Silicon Valley investment. These are the topics of this article. The article also features an interview with two Stanford professors who were involved in starting a startup in SambaNova. They discuss the challenges and advantages of starting a company in Silicon Valley.
Stanford professors found startup
Several Stanford professors are partners with leading venture capital firms. This article discusses the reasons why these professors decided to invest in startups.
Stanford has the highest number of venture-backed startups of any university in the country. The school’s endowment will invest in student-founded companies that raise at least $500k in outside investors.
Stanford’s Office of Technology Licensing has issued more than 3,600 licenses, generated more than $1.7 billion in royalties, and created more than 11,000 inventions. Stanford has been featured in The New Yorker and CNN. The school is also listed as the world’s best college for entrepreneurial thinking.
Stanford’s computer-science department may be a way station for talent. Big companies have limited access to domain knowledge and talent. Stanford’s faculty is encouraged to start their own companies.
Several Stanford professors have invested in student companies. In fact, more Stanford professors have invested in student companies than any other university in the world.
A Stanford student partnered with a Stanford professor to build an Internet search engine. This project was rejected by several major search engines. The two decided to turn the project into a startup business.
Silicon Valley investment
During the last decade, Silicon Valley focused primarily on software, but in recent years, it has shifted its focus to backing semiconductors. During 2020, Silicon Valley investors invested more than $7.4 billion in technology, up from $3.5 billion in 2018. That’s a lot of money for a relatively small region.
One startup that has been making waves in the world of silicon is SambaNova Systems Inc. The company aims to take on Nvidia Corp. and Cerebras Systems in the AI arena. Founded by Stanford professors Kunle Olukotun and Chris Re, SambaNova has developed a software-defined-hardware system that is set to deliver world-class performance and efficiency across applications.
The company’s flagship product is the Dataflow-as-a-Service, a subscription-based, extensible AI services platform that allows enterprises to tap into SambaNova’s AI system without the cost of maintaining it. Its AI-powered system is powered by input from TensorFlow, PyTorch and other open source algorithms, and uses a novel, single core processor.
New hardware
Touted as an AI superpower, SambaNova is one of the new breed of silicon chipsets slated for the future of computing. They aren’t exactly cheap, but the company has managed to pull off a coup by securing a few notable VC investors. They closed a series of notable funding rounds, including the aforementioned $150 million series B round led by Intel Capital and BlackRock. It’s also managed to attract an impressive number of high-ranking tech execs to its ranks. With a hefty capital stack in tow, SambaNova is staking out a piece of the AI pie based on its own proprietary software, and the latest fad in AI, cloud computing. Its biggest customer, the Department of Energy, has a sizable presence in the SambaNova fold. It’s also partnered with Cerebras Systems and TSMC to deliver AI systems of all shapes and sizes to the masses.
The SambaNova AI Platform boasts several scalability options, including the aforementioned DataScale SN10-8R courtesy of AMD and eight Cardinal chipsets. It’s also got some serious memory with 12 terabytes of DDR4 memory.
Integrated systems
Among the many companies chasing AI opportunities, SambaNova Systems is one of them. In addition to building artificial intelligence chips, the Silicon Valley startup has launched a service to help businesses access its AI system. The company has already partnered with a number of Fortune 500 companies and has been able to set world records in AI accuracy and scale.
SambaNova Systems aims to develop enterprise-level AI solutions for new industries. Its approach to this challenge involves selling software-defined hardware systems that can dramatically improve the performance of AI applications. The company has developed a software platform called DataScale, which is designed to help businesses run applications faster and more efficiently. It also offers Dataflow-as-a-Service, which gives businesses access to its AI system without the need to maintain it. This service is also free to use for research institutions.
Conclusion
In addition to building custom board-based hardware, SambaNova uses algorithmic optimizations to improve the performance of its AI systems. The company claims that its systems can improve the performance of AI apps by five times compared to other GPUs running the same models.
