7nm Ai Chip Is Ready To Be Sent To Hair

At the present stage, Xilinx has not focused on improving profits. Instead, it has invested more research and development resources to accelerate the development of new product line-up. In the emerging data center artificial intelligence (AI) operation, se is actively launching its own 7 nanometers Everest chip, which goes beyond the Intel (Intel) that is still delayed by the current 10 nanometers. If it can integrate the current leading field programmable gate array (FPGA) technology, it will help drive the future larger. Create potential.

According to The Motley Fool, Allied Market Research forecast that the global AI chip demand in the next 5 years is expected to grow nearly 50% in the next 1 years, showing that the demand for AI chips is expected to burst in the next few years, and Se se seems to be fully aware of this potential market opportunity, in addition to its own FPGA technology in the military AI field, and has continued to strengthen the research in the past few years. Investment, such as the proportion of 1/4 invested in R & D in 2017, increased by 20% annually. In fact, the growth rate of Xilinx R & D expenditure has been growing faster than total revenue growth in 3 years. The growth trend of the R & D expenditure is expected to continue in the 2019 year of the xilinsi plan, which will increase operating expenses by nearly 10%.

Currently, most of xilinsi chips are based on 28 and 20 nm process technology, but xilinsi is planning to move towards more efficient 16 and 7 nm process nodes. Once xilinsi will begin to produce more advanced process chips in the future, it will help further reduce the cost of chip manufacturing, and thus create greater profitability.

On the other hand, to make FPGA chip more advanced production production, it should also have a new AI chip market opportunity for snatling to eat up the emerging market of the new AI chip market, and SILs also looks at the business opportunity of the global data center cloud AI computing load market for the growing demand for AI chips. This is also why SILs actively promotes the 7 nano Everest chip and claims 7 nano E. The verest chip calculates AI load at a speed of at least 20 times faster than the current 16 nm chip. Xilinsi predicts that the composite growth rate (CAGR) of global data center chips will reach 67% in the next 5 years, and the market size will reach US $4 billion 600 million.

At present, SILs has started testing 7 nanometers Everest chips with some selected customers and plans to start shipping in 2019. This allows sailis to have a stronger competition with Intel on a feeding data center chip. After all, Intel's 10 nanometer scale production progress is still postponed and uncertain when it will be seen in 2019.

FPGA has a more flexible architecture that allows developers to reprogram specific tasks based on specific tasks, more energy-efficient than graphics chips (GPU), and is more suitable for deployment in a server like large-scale computing requirement facility, so several server vendors have begun deploying FPGA in their cloud, such as Amason (Amazon) in 2017. At the end of the alliance with sainling, FPGA provides cloud computing execution entities with FPGA, and other Baidu and Alibaba, such as mainland China, also choose to use FPGA to speed up its own cloud service.

Allied Market Research predicts that in the future, FPGA will become the fastest growing product category in the AI chip industry, and even surpass GPU products such as NVIDIA.