Beijing on March 16 morning news, Google Alphabet parent company under a venture capital division called GV, but never GV investment manufacture of artificial intelligence (AI) processor start-ups. This practice has been broken recently - GV took the lead and invested $56 million in smart hardware company SambaNovaSystems, a startup that produces computer processors and AI and data analysis software. CEO Rodrigo Liang has worked for Oracle.
Earlier, Google CEO Sandal Pichai and other executives often said that Google is an "AI-up" company. The company sells AI services to development companies and uses its features for consumer products. In the business overview section of the 2016 and 2017 annual reports, Alphabet mentions artificial intelligence.
However, in the past few years, GV (formerly known as Google Ventures) did not invest in AI hardware startups, GV general partner DaveMunichiello told CNBC.
“I think, according to our standards, SambaNova is at the right time, because the other companies we have seen are only in their infancy,†he said, and the products produced by those start-ups were not recognized in the market.
Mnichelo said that GV's existing products may not be able to adapt to future AI needs, and GV has unique insights. Google uses NVIDIA's graphics card to train the AI ​​model. At the same time, Google itself has developed a custom AI chip, the tensor processing unit (TPU). Another partner of Google, Intel Corporation, has also developed chips for AI following the acquisition of smart chip startup Nervana in 2016.
SambaNova was founded in November last year and currently employs more than 50 people from different locations. CEO RodrigoLiang has managed a team of nearly a thousand chip designers at Oracle. Others are from Stanford University, including Professor Christopher Re and Professor Kunle Olukotun. Professor Olukotun worked with Rodrigo Liang on the chip company AfaraWebsystems, which was acquired by Sun Microsystems in 2002.
Olukotun said that SambaNova's focus is not only on training AI models: its infrastructure will be able to accommodate other types of computing, including quick query databases.
“The various methods have been tried, but they are all subject to the basic restrictions of Moore's Law,†Olukotun said. “To further improve performance, we need to consider a more efficient calculation method.â€
RodrigoLiang said that the startup has not yet built the first generation of chips, but has already negotiated with potential customers and public cloud providers.
In this round of financing, Walden International and GV led the investment, Atlantic Bridge Ventures (AtlanticBridge Ventures) and Red Line Capital all participated.
Customized Table Clamp,Customized Table Clamp For Screen,Fixture Table Clamps,Customized Table Clamp For Rod
MGX Metalworks Co., Ltd. , https://www.mgxmetal.com