Lol-To The Moon - Tiger Brokers Online Investing Community

Like
Comment
Favorite

Boris628
2021-02-12

Lol

Baidu was revealed to establish a chip company! Why has the Internet entered the game to make cores?

市值突破1000亿美元大关后，百度最新进展也随之曝光——成立一家专门的AI芯片公司，独立发展的那种，像智能汽车模式的那种。而百度自研芯片昆仑2，也将于今年实现量产。消息一出，百度盘前大涨5%。不得不说

Baidu was revealed to establish a chip company! Why has the Internet entered the game to make cores?

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Report

Comment

Top
Latest

No comments yet

Most Discussed

{"i18n":{"language":"en_US"},"data":{"magic":2,"id":386082708,"tweetId":"386082708","gmtCreate":1613112621104,"gmtModify":1704878516230,"author":{"id":3568930674369958,"idStr":"3568930674369958","authorId":3568930674369958,"authorIdStr":"3568930674369958","name":"Boris628","avatar":"https://static.tigerbbs.com/a8dfcfdfd3fab52796ba7c3129a89fe8","vip":1,"userType":1,"introduction":"","boolIsFan":false,"boolIsHead":false,"crmLevel":11,"crmLevelSwitch":0,"individualDisplayBadges":[],"fanSize":0,"starInvestorFlag":false},"themes":[],"images":[],"coverImages":[],"extraTitle":"","html":"<html><head></head><body>Lol</body></html>","htmlText":"<html><head></head><body>Lol</body></html>","text":"Lol","highlighted":1,"essential":1,"paper":1,"likeSize":0,"commentSize":0,"repostSize":0,"favoriteSize":0,"link":"https://ttm.financial/post/386082708","repostId":1109604105,"repostType":2,"repost":{"id":"1109604105","kind":"news","pubTimestamp":1613107738,"share":"https://ttm.financial/m/news/1109604105?lang=en_US&edition=fundamental","pubTime":"2021-02-12 13:28","market":"us","language":"zh","title":"Baidu was revealed to establish a chip company! Why has the Internet entered the game to make cores?","url":"https://stock-news.laohu8.com/highlight/detail?id=1109604105","media":"量子位 ","summary":"市值突破1000亿美元大关后，百度最新进展也随之曝光——成立一家专门的AI芯片公司，独立发展的那种，像智能汽车模式的那种。而百度自研芯片昆仑2，也将于今年实现量产。消息一出，百度盘前大涨5%。不得不说","content":"After the market value exceeded the $100 billion mark,<a href=\"https://laohu8.com/S/BIDU\">Baidu</a>The latest progress has also been exposed-Establish a dedicatedAI chip companies, the kind of independent development, like<a href=\"https://laohu8.com/S/5RE.SI\">smart</a>The car mode kind.Baidu's self-developed chip Kunlun 2 will also be mass-produced this year.As soon as the news came out, Baidu rose 5% before the market opened.It has to be said that leading technology companies have been surprisingly consistent in their actions in recent years:Whether at home or abroad, it seems that they have set their sights on the field of \"AI chips\".<img src=\"https://static.tigerbbs.com/90a41f0b41a5ad062ea6d81e8f3937c5\" tg-width=\"1000\" tg-height=\"667\" referrerpolicy=\"no-referrer\">First, mobile phone manufacturers such as<a href=\"https://laohu8.com/S/AAPL\">Apple</a>, Huawei, and OPPO have gone overseas to develop their own AI chips.In 2017, Apple A11 Bonic came out, using a self-developed GPU for the first time and equipped with a neural network engine for the first time, ushering in the era of smartphone AI;The release of Huawei's \"Ascend\" series chips marks Huawei's official entry into the AI processor industry;OPPO has also clearly entered semiconductors and built self-developed chips.Meanwhile, Internet companies such as Google,<a href=\"https://laohu8.com/S/AMZN\">Amazon</a>, BAT, also began to take action.Foreign countries such as Google began designing server chips for data centers in 2014, and set up a chip team in India in 2019.One of Amazon's largest sources of profit, cloud computing company AWS, also disclosed its first self-developed cloud server CPUGraviton at the end of 2018.Domestically, such as Ali, established \"Pingtouge Semiconductor Co., Ltd.\" in 2018 and officially entered the AI chip market;As for<a href=\"https://laohu8.com/S/00700\">Tencent</a>, has also bet on Suiyuan Technology, focusing on the commercial implementation of AI chips.This time, Baidu announced the independent spin-off of AI chips, and even directly smashed the blockbuster news that self-developed chips will be mass-produced this year.BAT's layout in core manufacturing is now clear.Why are giants entering the game to make cores?There are two reasons why Internet giants have entered the game to make cores.<img src=\"https://static.tigerbbs.com/d4cfa0e5f3e56944faf3d2485f097c7a\" tg-width=\"1080\" tg-height=\"406\" referrerpolicy=\"no-referrer\">On the one hand, as far as the international situation is concerned, domestic technology companies have been \"stuck\" with foreign technologies on chips for a long time.In 2018, the U.S. Department of Commerce suddenly announced that it would ban U.S. companies from submitting to<a href=\"https://laohu8.com/S/00763\">ZTE Corporation</a>Selling parts.In March 2019,<a href=\"https://laohu8.com/S/000063\">ZTE Corporation</a>Announced a loss of 6.983 billion yuan in 2018, a difference of more than 10 billion yuan compared with a profit of 4.568 billion yuan in 2017.Previously, ZTE relied on American chip companies for a large number of chip supplies. This incident is equivalent to a blow to domestic companies.Before and after this, there were domestic companies such as DJI, Huawei, Xiaomi, etc., all of which were \"blocked\" by the United States to varying degrees.If they want to be immune to changes in the industry, Chinese enterprises must make corresponding choices.It can be said that self-developed chips are a path that domestic companies must take.<img src=\"https://static.tigerbbs.com/48e8525be202d4f02f4386a2bc6ea57c\" tg-width=\"1080\" tg-height=\"810\" referrerpolicy=\"no-referrer\">On the other hand, the needs and opportunities brought by artificial intelligence.Previously, chips such as personal computers and mobile phones, although basically<a href=\"https://laohu8.com/S/INTC\">Intel</a>、<a href=\"https://laohu8.com/S/AMD\">AMD</a>Waiting for companies to monopolize, but the rise of artificial intelligence technology is tantamount to bringing a rare window period to the industry.With the rapid development of artificial intelligence algorithms, industries such as finance, medical care, energy, manufacturing, government affairs, transportation, and home furnishing are realizing further \"AI + industrialization\".While all walks of life are rapidly becoming intelligent, the arrival of the epidemic has further accelerated the demand for artificial intelligence algorithms.Whether it is the demand for intelligent services such as \"unmanned\" or the demand for intelligent voice such as \"contactless\", waves of industrial waves are constantly forcing the application of AI technology.The rapid development of software is bound to drive the innovation of hardware industry.Faced with the special requirements of AI algorithms for computing power and task processing performance, the previous generation of processors, including GPUs, have shown a weak trend, and the actual task operation efficiency is very low.<img src=\"https://static.tigerbbs.com/12b3a713c066634f25cfab7c9b4f97c9\" tg-width=\"1000\" tg-height=\"615\" referrerpolicy=\"no-referrer\">And the software redefines the hardware, and the AI algorithm redefines the chip, which has come to the breaking point.In the chip industry, under the impact of AI, a gap has been opened, and hardware demand continues to emerge.This is undoubtedly the best time for a chip to enter-If domestic technology companies can seize this opportunity and develop their own AI chips while developing AI technology, the layout of integrated software and hardware will inevitably have an impact on foreign technology industries.And Catherine Wood, the \"Queen of the Bull Market\" who once became a god due to her heavy position in Musk and Bitcoin, boldly predicted in the 2021 ARK report that the future expenditure on AI chips will increase by more than 4 times:From the current $5 billion, it will grow to $22 billion in 2025.If a powerful player doesn't do it at this time, when will he wait?If you don't do it yourself, who will you treat?Baidu's core manufacturing actually takes ten years to grind a swordBut this time Baidu's AI chip business is spun off independently, is it just chasing the trend?What is the confidence behind it?Judging from the pace of domestic technology giants betting on chips, Baidu seems to be the latest.However, if you trace back to the source or are familiar with Baidu's technology bets, you will find that Baidu's layout in chips started as early as more than ten years ago.As the earliest technology company in China to focus on AI, in 2010, in order to serve its own business, Baidu began to use FPGA to research and develop AI architecture, and launched a small-scale deployment the following year.By 2017, Baidu had completed the deployment of more than 10,000 FPGAs.On this basis, in 2018, Baidu officially released its self-developed AI chip-Baidu Kunlun。<img src=\"https://static.tigerbbs.com/ded62bbc1daec92d2dcb15c4797a3bc3\" tg-width=\"1080\" tg-height=\"647\" referrerpolicy=\"no-referrer\">Kunlun's 1st-generation processor is built based on FPGA architecture, using 14nm process, 16GB HBM memory and 2.5 D packaging solution, and can provide 512GB/s memory bandwidth.Moreover, at a power lower than 150W, Kunlun 1st generation can achieve an INT8 processing capacity of 256 TOPS.In other words, on the basis of satisfying AI reasoning and conventional training, and supporting general artificial intelligence algorithms, Kunlun chips can further reduce usage costs.To sum up simply, Kunlun has the following characteristics:<ul><li>Versatility, supporting all AI applications and scenarios;</li><li>Easy to program, can build a powerful software ecological moat;</li><li>High performance, which can support deeper exploration of AI applications;</li><li>Autonomous and controllable, while reducing the cost of use, it can alleviate the problem of being \"stuck\".</li></ul>Up to now, Kunlun 1 has mass-produced more than 20,000 pieces.So, where are such chips used? Is there a fancy shelf?In fact, Kunlun 1 has been widely deployed in both the familiar Baidu search engine and the Baidu intelligent cloud of To B.<img src=\"https://static.tigerbbs.com/765ac4e6ae4bf2f15ef887c3c8f49830\" tg-width=\"1080\" tg-height=\"706\" referrerpolicy=\"no-referrer\">The most advanced open source deep learning framework in the domestic ecologyFlying paddle(PaddlePaddle), which is also deeply integrated with Kunlun chips, forming a software and hardware integrated combination of \"self-developed chips + independent AI framework\".This combination of software and hardware makes Baidu's chips face smart finance, smart medical care,<a href=\"https://laohu8.com/S/600869\">Smart Energy</a>, smart manufacturing, smart government affairs, smart transportation and other digital and intelligent scenarios, it can be more targeted, reduce costs and increase efficiency.In addition, in addition to natively supporting Baidu's own Flying Paddle, Baidu Machine Learning Platform (BML), and various vertical AI capability engines, Kunlun also supports global mainstream CPUs, operating systems, deep learning frameworks such as PyTorch and TensorFlow. It has also cooperated with domestic manufacturers to support domestic CPUs such as Feiteng, Shenwei and Haiguang, and domestic operating systems such as Kirin, Deepin and Communications.Baidu also revealed that Kunlun 2 will be mass-produced in the first half of 2021. This chip uses a 7nm process, and its performance will be three times higher than that of the first generation.In addition to Kunlun, which focuses more on the industry side, Baidu also has finished products for consumers, that isBaidu Honghu voice chip。Honghu voice chip was officially launched in July 2019. Compared with Kunlun, it targets more vertical scenarios:Specially designed for smart home voice interaction, smart car voice interaction, and smart IoT scenarios, it is a voice interaction chip focusing on ultra-low power consumption.In terms of chip architecture, Honghu has an embedded end-to-end integrated modeling algorithm, which can complete a series of voice interaction functions such as far-field pickup, wake-up, and positioning, and directly transmit the extracted features to the cloud for high-precision recognition in the cloud.This also enables manufacturers to further reduce the cost of intelligent hardware without losing voice experience.In addition, Honghu chips also use a dual-core HiFi4 architecture custom instruction set to improve computing power.<img src=\"https://static.tigerbbs.com/232eb4c57a4991efefee5841c57d8adf\" tg-width=\"1080\" tg-height=\"1008\" referrerpolicy=\"no-referrer\">The infrared version of the Xiaodu Smart Speaker 2 launched in 2020 is already equipped with this chip, and has thus achieved three major performance improvements:<ul><li>The speech recognition error rate is reduced by an average of 30%;</li><li>The first wake-up rate under high noise increases by more than 10%, meeting the ultra-low false alarm requirements for home use;</li><li>The average operating power consumption is only about 100mw, and the standby power drops by 90%.</li></ul>Baidu core manufacturing, the key lies in ecologySumming up this path, in fact, the differentiated competitiveness of Baidu Core Manufacturing has been further revealed.The key is stillsceneAndEcology。Independent research and development of AI chips requires not only hardware technology, but also an in-depth understanding of AI algorithms and the ability to combine software and hardware.As the earliest technology company in China to start deploying AI, Baidu now has the AI infrastructure \"Baidu Brain\" that covers all aspects from hardware to software, from the bottom of technology to application scenarios. Whether it is maps, search, unmanned vehicles, or smart speakers, Baidu's technology accumulation and business practice for many years are integrated into it.<img src=\"https://static.tigerbbs.com/8fa976eff2584fc9474bf93af989fc77\" tg-width=\"1080\" tg-height=\"432\" referrerpolicy=\"no-referrer\">Baidu Brain 6.0, released in 2020, has opened more than 270 leading AI capabilities, gathered 2.3 million developers, and the number of models released by enterprises exceeds 310,000. It is the most comprehensive AI open platform with the largest service scale in the industry.Backed by such an ecosystem, Baidu Core Manufacturing has more opportunities for implementation.Taking Honghu chips as an example, it is conceivable that Xiaodu speakers are only the first stop. In the future, such voice chips may be further verified in Apollo unmanned vehicles, and even home and industrial scenarios.Standing at the current time point, grasping the new opportunities in the chip industry in the AI era and further exploring from independent controllability of software to independent controllability of hardware is of far-reaching significance not only to Baidu itself, but also to the entire Chinese technology industry.Therefore, after Xiaodu was valued at 20 billion, Apollo was also benchmarked by research institutions against Cruise, which was valued at 30 billion US dollars...How does Baidu's newly hatched AI chip company need to be evaluated?From the recent rise in Baidu's stock price, you may smell a hint of wind direction.<img src=\"https://static.tigerbbs.com/1cff424f30f5f185dea340f865a40d1a\" tg-width=\"840\" tg-height=\"470\" referrerpolicy=\"no-referrer\">Let's wait and see.","source":"lsy1569318700471","collect":0,"html":"<!DOCTYPE html>\n<html>\n<head>\n<meta http-equiv=\"Content-Type\" content=\"text/html; charset=utf-8\" />\n<meta name=\"viewport\" content=\"width=device-width,initial-scale=1.0,minimum-scale=1.0,maximum-scale=1.0,user-scalable=no\"/>\n<meta name=\"format-detection\" content=\"telephone=no,email=no,address=no\" />\n<title>Baidu was revealed to establish a chip company! Why has the Internet entered the game to make cores?</title>\n<style type=\"text/css\">\na,abbr,acronym,address,applet,article,aside,audio,b,big,blockquote,body,canvas,caption,center,cite,code,dd,del,details,dfn,div,dl,dt,\nem,embed,fieldset,figcaption,figure,footer,form,h1,h2,h3,h4,h5,h6,header,hgroup,html,i,iframe,img,ins,kbd,label,legend,li,mark,menu,nav,\nobject,ol,output,p,pre,q,ruby,s,samp,section,small,span,strike,strong,sub,summary,sup,table,tbody,td,tfoot,th,thead,time,tr,tt,u,ul,var,video{ font:inherit;margin:0;padding:0;vertical-align:baseline;border:0 }\nbody{ font-size:16px; line-height:1.5; color:#999; background:transparent; }\n.wrapper{ overflow:hidden;word-break:break-all;padding:10px; }\nh1,h2{ font-weight:normal; line-height:1.35; margin-bottom:.6em; }\nh3,h4,h5,h6{ line-height:1.35; margin-bottom:1em; }\nh1{ font-size:24px; }\nh2{ font-size:20px; }\nh3{ font-size:18px; }\nh4{ font-size:16px; }\nh5{ font-size:14px; }\nh6{ font-size:12px; }\np,ul,ol,blockquote,dl,table{ margin:1.2em 0; }\nul,ol{ margin-left:2em; }\nul{ list-style:disc; }\nol{ list-style:decimal; }\nli,li p{ margin:10px 0;}\nimg{ max-width:100%;display:block;margin:0 auto 1em; }\nblockquote{ color:#B5B2B1; border-left:3px solid #aaa; padding:1em; }\nstrong,b{font-weight:bold;}\nem,i{font-style:italic;}\ntable{ width:100%;border-collapse:collapse;border-spacing:1px;margin:1em 0;font-size:.9em; }\nth,td{ padding:5px;text-align:left;border:1px solid #aaa; }\nth{ font-weight:bold;background:#5d5d5d; }\n.symbol-link{font-weight:bold;}\n/* header{ border-bottom:1px solid #494756; } */\n.title{ margin:0 0 8px;line-height:1.3;color:#ddd; }\n.meta {color:#5e5c6d;font-size:13px;margin:0 0 .5em; }\na{text-decoration:none; color:#2a4b87;}\n.meta .head { display: inline-block; overflow: hidden}\n.head .h-thumb { width: 30px; height: 30px; margin: 0; padding: 0; border-radius: 50%; float: left;}\n.head .h-content { margin: 0; padding: 0 0 0 9px; float: left;}\n.head .h-name {font-size: 13px; color: #eee; margin: 0;}\n.head .h-time {font-size: 12.5px; color: #7E829C; margin: 0;}\n.small {font-size: 12.5px; display: inline-block; transform: scale(0.9); -webkit-transform: scale(0.9); transform-origin: left; -webkit-transform-origin: left;}\n.smaller {font-size: 12.5px; display: inline-block; transform: scale(0.8); -webkit-transform: scale(0.8); transform-origin: left; -webkit-transform-origin: left;}\n.bt-text {font-size: 12px;margin: 1.5em 0 0 0}\n.bt-text p {margin: 0}\n</style>\n</head>\n<body>\n<div class=\"wrapper\">\n<header>\n<h2 class=\"title\">\nBaidu was revealed to establish a chip company! Why has the Internet entered the game to make cores?\n</h2>\n<h4 class=\"meta\">\n\n量子位 2021-02-12 13:28\n\n</h4>\n</header>\n<article>\nAfter the market value exceeded the $100 billion mark,<a href=\"https://laohu8.com/S/BIDU\">Baidu</a>The latest progress has also been exposed-Establish a dedicatedAI chip companies, the kind of independent development, like<a href=\"https://laohu8.com/S/5RE.SI\">smart</a>The car mode kind.Baidu's self-developed chip Kunlun 2 will also be mass-produced this year.As soon as the news came out, Baidu rose 5% before the market opened.It has to be said that leading technology companies have been surprisingly consistent in their actions in recent years:Whether at home or abroad, it seems that they have set their sights on the field of \"AI chips\".<img src=\"https://static.tigerbbs.com/90a41f0b41a5ad062ea6d81e8f3937c5\" tg-width=\"1000\" tg-height=\"667\" referrerpolicy=\"no-referrer\">First, mobile phone manufacturers such as<a href=\"https://laohu8.com/S/AAPL\">Apple</a>, Huawei, and OPPO have gone overseas to develop their own AI chips.In 2017, Apple A11 Bonic came out, using a self-developed GPU for the first time and equipped with a neural network engine for the first time, ushering in the era of smartphone AI;The release of Huawei's \"Ascend\" series chips marks Huawei's official entry into the AI processor industry;OPPO has also clearly entered semiconductors and built self-developed chips.Meanwhile, Internet companies such as Google,<a href=\"https://laohu8.com/S/AMZN\">Amazon</a>, BAT, also began to take action.Foreign countries such as Google began designing server chips for data centers in 2014, and set up a chip team in India in 2019.One of Amazon's largest sources of profit, cloud computing company AWS, also disclosed its first self-developed cloud server CPUGraviton at the end of 2018.Domestically, such as Ali, established \"Pingtouge Semiconductor Co., Ltd.\" in 2018 and officially entered the AI chip market;As for<a href=\"https://laohu8.com/S/00700\">Tencent</a>, has also bet on Suiyuan Technology, focusing on the commercial implementation of AI chips.This time, Baidu announced the independent spin-off of AI chips, and even directly smashed the blockbuster news that self-developed chips will be mass-produced this year.BAT's layout in core manufacturing is now clear.Why are giants entering the game to make cores?There are two reasons why Internet giants have entered the game to make cores.<img src=\"https://static.tigerbbs.com/d4cfa0e5f3e56944faf3d2485f097c7a\" tg-width=\"1080\" tg-height=\"406\" referrerpolicy=\"no-referrer\">On the one hand, as far as the international situation is concerned, domestic technology companies have been \"stuck\" with foreign technologies on chips for a long time.In 2018, the U.S. Department of Commerce suddenly announced that it would ban U.S. companies from submitting to<a href=\"https://laohu8.com/S/00763\">ZTE Corporation</a>Selling parts.In March 2019,<a href=\"https://laohu8.com/S/000063\">ZTE Corporation</a>Announced a loss of 6.983 billion yuan in 2018, a difference of more than 10 billion yuan compared with a profit of 4.568 billion yuan in 2017.Previously, ZTE relied on American chip companies for a large number of chip supplies. This incident is equivalent to a blow to domestic companies.Before and after this, there were domestic companies such as DJI, Huawei, Xiaomi, etc., all of which were \"blocked\" by the United States to varying degrees.If they want to be immune to changes in the industry, Chinese enterprises must make corresponding choices.It can be said that self-developed chips are a path that domestic companies must take.<img src=\"https://static.tigerbbs.com/48e8525be202d4f02f4386a2bc6ea57c\" tg-width=\"1080\" tg-height=\"810\" referrerpolicy=\"no-referrer\">On the other hand, the needs and opportunities brought by artificial intelligence.Previously, chips such as personal computers and mobile phones, although basically<a href=\"https://laohu8.com/S/INTC\">Intel</a>、<a href=\"https://laohu8.com/S/AMD\">AMD</a>Waiting for companies to monopolize, but the rise of artificial intelligence technology is tantamount to bringing a rare window period to the industry.With the rapid development of artificial intelligence algorithms, industries such as finance, medical care, energy, manufacturing, government affairs, transportation, and home furnishing are realizing further \"AI + industrialization\".While all walks of life are rapidly becoming intelligent, the arrival of the epidemic has further accelerated the demand for artificial intelligence algorithms.Whether it is the demand for intelligent services such as \"unmanned\" or the demand for intelligent voice such as \"contactless\", waves of industrial waves are constantly forcing the application of AI technology.The rapid development of software is bound to drive the innovation of hardware industry.Faced with the special requirements of AI algorithms for computing power and task processing performance, the previous generation of processors, including GPUs, have shown a weak trend, and the actual task operation efficiency is very low.<img src=\"https://static.tigerbbs.com/12b3a713c066634f25cfab7c9b4f97c9\" tg-width=\"1000\" tg-height=\"615\" referrerpolicy=\"no-referrer\">And the software redefines the hardware, and the AI algorithm redefines the chip, which has come to the breaking point.In the chip industry, under the impact of AI, a gap has been opened, and hardware demand continues to emerge.This is undoubtedly the best time for a chip to enter-If domestic technology companies can seize this opportunity and develop their own AI chips while developing AI technology, the layout of integrated software and hardware will inevitably have an impact on foreign technology industries.And Catherine Wood, the \"Queen of the Bull Market\" who once became a god due to her heavy position in Musk and Bitcoin, boldly predicted in the 2021 ARK report that the future expenditure on AI chips will increase by more than 4 times:From the current $5 billion, it will grow to $22 billion in 2025.If a powerful player doesn't do it at this time, when will he wait?If you don't do it yourself, who will you treat?Baidu's core manufacturing actually takes ten years to grind a swordBut this time Baidu's AI chip business is spun off independently, is it just chasing the trend?What is the confidence behind it?Judging from the pace of domestic technology giants betting on chips, Baidu seems to be the latest.However, if you trace back to the source or are familiar with Baidu's technology bets, you will find that Baidu's layout in chips started as early as more than ten years ago.As the earliest technology company in China to focus on AI, in 2010, in order to serve its own business, Baidu began to use FPGA to research and develop AI architecture, and launched a small-scale deployment the following year.By 2017, Baidu had completed the deployment of more than 10,000 FPGAs.On this basis, in 2018, Baidu officially released its self-developed AI chip-Baidu Kunlun。<img src=\"https://static.tigerbbs.com/ded62bbc1daec92d2dcb15c4797a3bc3\" tg-width=\"1080\" tg-height=\"647\" referrerpolicy=\"no-referrer\">Kunlun's 1st-generation processor is built based on FPGA architecture, using 14nm process, 16GB HBM memory and 2.5 D packaging solution, and can provide 512GB/s memory bandwidth.Moreover, at a power lower than 150W, Kunlun 1st generation can achieve an INT8 processing capacity of 256 TOPS.In other words, on the basis of satisfying AI reasoning and conventional training, and supporting general artificial intelligence algorithms, Kunlun chips can further reduce usage costs.To sum up simply, Kunlun has the following characteristics:<ul><li>Versatility, supporting all AI applications and scenarios;</li><li>Easy to program, can build a powerful software ecological moat;</li><li>High performance, which can support deeper exploration of AI applications;</li><li>Autonomous and controllable, while reducing the cost of use, it can alleviate the problem of being \"stuck\".</li></ul>Up to now, Kunlun 1 has mass-produced more than 20,000 pieces.So, where are such chips used? Is there a fancy shelf?In fact, Kunlun 1 has been widely deployed in both the familiar Baidu search engine and the Baidu intelligent cloud of To B.<img src=\"https://static.tigerbbs.com/765ac4e6ae4bf2f15ef887c3c8f49830\" tg-width=\"1080\" tg-height=\"706\" referrerpolicy=\"no-referrer\">The most advanced open source deep learning framework in the domestic ecologyFlying paddle(PaddlePaddle), which is also deeply integrated with Kunlun chips, forming a software and hardware integrated combination of \"self-developed chips + independent AI framework\".This combination of software and hardware makes Baidu's chips face smart finance, smart medical care,<a href=\"https://laohu8.com/S/600869\">Smart Energy</a>, smart manufacturing, smart government affairs, smart transportation and other digital and intelligent scenarios, it can be more targeted, reduce costs and increase efficiency.In addition, in addition to natively supporting Baidu's own Flying Paddle, Baidu Machine Learning Platform (BML), and various vertical AI capability engines, Kunlun also supports global mainstream CPUs, operating systems, deep learning frameworks such as PyTorch and TensorFlow. It has also cooperated with domestic manufacturers to support domestic CPUs such as Feiteng, Shenwei and Haiguang, and domestic operating systems such as Kirin, Deepin and Communications.Baidu also revealed that Kunlun 2 will be mass-produced in the first half of 2021. This chip uses a 7nm process, and its performance will be three times higher than that of the first generation.In addition to Kunlun, which focuses more on the industry side, Baidu also has finished products for consumers, that isBaidu Honghu voice chip。Honghu voice chip was officially launched in July 2019. Compared with Kunlun, it targets more vertical scenarios:Specially designed for smart home voice interaction, smart car voice interaction, and smart IoT scenarios, it is a voice interaction chip focusing on ultra-low power consumption.In terms of chip architecture, Honghu has an embedded end-to-end integrated modeling algorithm, which can complete a series of voice interaction functions such as far-field pickup, wake-up, and positioning, and directly transmit the extracted features to the cloud for high-precision recognition in the cloud.This also enables manufacturers to further reduce the cost of intelligent hardware without losing voice experience.In addition, Honghu chips also use a dual-core HiFi4 architecture custom instruction set to improve computing power.<img src=\"https://static.tigerbbs.com/232eb4c57a4991efefee5841c57d8adf\" tg-width=\"1080\" tg-height=\"1008\" referrerpolicy=\"no-referrer\">The infrared version of the Xiaodu Smart Speaker 2 launched in 2020 is already equipped with this chip, and has thus achieved three major performance improvements:<ul><li>The speech recognition error rate is reduced by an average of 30%;</li><li>The first wake-up rate under high noise increases by more than 10%, meeting the ultra-low false alarm requirements for home use;</li><li>The average operating power consumption is only about 100mw, and the standby power drops by 90%.</li></ul>Baidu core manufacturing, the key lies in ecologySumming up this path, in fact, the differentiated competitiveness of Baidu Core Manufacturing has been further revealed.The key is stillsceneAndEcology。Independent research and development of AI chips requires not only hardware technology, but also an in-depth understanding of AI algorithms and the ability to combine software and hardware.As the earliest technology company in China to start deploying AI, Baidu now has the AI infrastructure \"Baidu Brain\" that covers all aspects from hardware to software, from the bottom of technology to application scenarios. Whether it is maps, search, unmanned vehicles, or smart speakers, Baidu's technology accumulation and business practice for many years are integrated into it.<img src=\"https://static.tigerbbs.com/8fa976eff2584fc9474bf93af989fc77\" tg-width=\"1080\" tg-height=\"432\" referrerpolicy=\"no-referrer\">Baidu Brain 6.0, released in 2020, has opened more than 270 leading AI capabilities, gathered 2.3 million developers, and the number of models released by enterprises exceeds 310,000. It is the most comprehensive AI open platform with the largest service scale in the industry.Backed by such an ecosystem, Baidu Core Manufacturing has more opportunities for implementation.Taking Honghu chips as an example, it is conceivable that Xiaodu speakers are only the first stop. In the future, such voice chips may be further verified in Apollo unmanned vehicles, and even home and industrial scenarios.Standing at the current time point, grasping the new opportunities in the chip industry in the AI era and further exploring from independent controllability of software to independent controllability of hardware is of far-reaching significance not only to Baidu itself, but also to the entire Chinese technology industry.Therefore, after Xiaodu was valued at 20 billion, Apollo was also benchmarked by research institutions against Cruise, which was valued at 30 billion US dollars...How does Baidu's newly hatched AI chip company need to be evaluated?From the recent rise in Baidu's stock price, you may smell a hint of wind direction.<img src=\"https://static.tigerbbs.com/1cff424f30f5f185dea340f865a40d1a\" tg-width=\"840\" tg-height=\"470\" referrerpolicy=\"no-referrer\">Let's wait and see.\n<div class=\"bt-text\">\n\n\n source：<a href=\"https://mp.weixin.qq.com/s/9O3PvYr0Sp5sHxrkScIExQ\">量子位 </a>\n\n\n</div>\n</article>\n</div>\n</body>\n</html>\n","type":0,"thumbnail":"https://static.tigerbbs.com/765ac4e6ae4bf2f15ef887c3c8f49830","relate_stocks":{"BIDU":"百度"},"source_url":"https://mp.weixin.qq.com/s/9O3PvYr0Sp5sHxrkScIExQ","is_english":false,"share_image_url":"https://static.laohu8.com/e9f99090a1c2ed51c021029395664489","article_id":"1109604105","content_text":"市值突破1000亿美元大关后，百度最新进展也随之曝光——成立一家专门的AI芯片公司，独立发展的那种，像智能汽车模式的那种。而百度自研芯片昆仑2，也将于今年实现量产。消息一出，百度盘前大涨5%。不得不说，头部科技公司，近年来行动还真是出奇地一致：无论国内外，似乎都将目光放在了“AI芯片”这一领域。先是手机厂商如苹果、华为、OPPO，纷纷下海自研AI芯片。2017年，苹果A11 Bonic问世，首次采用自研GPU，首次搭载神经网络引擎，开启智能手机AI时代；华为“昇腾”系列芯片出炉，标志着华为正式进军AI处理器行业；OPPO也明确进军半导体，打造自研芯片。同时互联网公司如谷歌、亚马逊、BAT，也开始有所行动。国外如谷歌，2014年就开始为数据中心设计服务器芯片，2019年更是在印度组建芯片团队。而亚马逊最大盈利来源之一、云计算公司AWS，也在2018年底披露了首款自研云服务器CPUGraviton。国内如阿里，2018年成立“平头哥半导体有限公司”，正式入局AI芯片；至于腾讯，也已押注燧原科技，专注于AI芯片的商用落地。此番，百度又宣布AI芯片独立分拆，甚至直接砸出自研芯片今年量产落地的重磅消息。BAT在造芯方面的布局，如今已然明朗。为何巨头纷纷入局造芯？互联网巨头入局造芯的原因有两大方面。一方面，就国际形势而言，国内科技公司在芯片上，遭遇国外技术“卡脖子”已久。2018年，美国商务部突然宣布，7年内禁止美国企业向中兴通讯销售零件。2019年3月，中兴通讯公布2018年度亏损69.83亿元人民币，与2017年度盈利45.68亿元比较，相差超过100亿元。此前，中兴大量芯片供应依赖于美国芯片企业，这次事件，相当于给国内企业当头一击。在这前后，还有大疆、华为、小米……等等国内企业，均遭遇了美国不同程度的“封杀”。要想在行业中不受局势变化的影响，中国企业必须做出对应的抉择。可以说，自研芯片是国内企业必须走出的一条路。另一方面，人工智能带来的需求和机会。此前，个人电脑和手机等芯片，虽然基本已经被英特尔、AMD等公司垄断，但人工智能技术的兴起，无异于给行业带来了一个难得的窗口期。随着人工智能算法的飞速发展，金融、医疗、能源、制造、政务、交通、家居等行业，正在实现进一步的“AI+产业化”。各行各业飞速智能化的同时，疫情的到来，又进一步加速了人工智能算法的需求。无论是“无人化”等智能服务需求、还是“无接触”等智能语音需求，一波波产业浪潮，正在不断倒逼AI技术的落地应用。软件的飞速发展，势必带动硬件行业的革新。面对AI算法对算力、任务处理性能的特殊需求，包括GPU在内的上一代处理器，已经呈现出乏力之势，实际任务运行效率非常低下。并且软件重新定义硬件，AI算法重新定义芯片，来到了破局点。芯片行业，在AI的冲击之下，打开了一个缺口，硬件需求源源不断地涌现。这无疑是一个入局芯片的最好时机——如果国内科技公司能抓住这一机遇，在发展AI技术的同时自研AI芯片，软硬件一体的布局，势必也能对国外技术产业造成冲击。而曾经因重仓马斯克和比特币封神的“牛市女皇”Catherine Wood，则更是在2021年ARK报告中大胆预测，AI芯片未来的支出将增长4倍以上：从当前的50亿美元，增长到2025年的220亿美元。有实力的玩家，此时不做，更待何时？自己不做，更待何人？百度造芯，其实「十年磨一剑」但此番百度AI芯片业务独立分拆，只是在追逐风口吗？背后的底气何在？从国内科技巨头押注芯片的节奏上来看，百度似乎落子最迟。然而，如果寻根溯源，或是对百度技术押注熟悉，就会发现，原来百度在芯片方面的布局，早在十几年前就已展开。作为国内最早重注AI的科技公司，2010年，为了服务于自身业务，百度就已经开始使用FPGA进行AI架构的研发，并于次年就开展了小规模部署上线。到了2017年，百度已经完成超过1万片FPGA的部署。在这样的基础之上，2018年，百度正式发布自主研发的AI芯片——百度昆仑。昆仑1代处理器基于FPGA架构打造，采用14nm工艺，16GB HBM内存和2.5D封装解决方案，能够提供512GB/s的内存带宽。并且，在低于150W的功率下，昆仑1代能实现256 TOPS的INT8处理能力。也就是说，在满足AI推理和常规训练、支持通用人工智能算法的基础之上，昆仑芯片还能进一步降低使用成本。简单总结起来，昆仑具有以下几点特征：通用性，支持全部AI应用和场景；易编程，可以构建强大的软件生态护城河；高性能，可以支持更深入的AI应用探索；自主可控，在降低使用成本的同时，缓解被“卡脖子”的问题。截至目前，昆仑1已经量产超过2万片。所以，这样的芯片都用在了哪里？是否空有花架子？事实上，无论是在大众所熟悉的百度搜索引擎，还是在To B的百度智能云中，昆仑1都已经被广泛部署。国内生态最领先的开源深度学习框架飞桨（PaddlePaddle），也同昆仑芯片深度融合，形成了“自研芯片+自主AI框架”的软硬一体组合。而这样软硬件结合的模式，也就使得百度的芯片在面对智慧金融、智慧医疗、智慧能源、智能制造、智慧政务、智能交通等等数字化、智能化场景时，能做到更加有的放矢、降本增效。另外，除了原生支持百度自家的飞桨、百度机器学习平台（BML）以及各垂类的AI能力引擎，昆仑也支持全球主流CPU、操作系统、PyTorch和TensorFlow等深度学习框架。并与国内厂商展开合作，实现了对飞腾、申威和海光等国产CPU，麒麟、深度和通信等国产操作系统的支持。百度方面还透露，昆仑2将于2021年上半年实现量产，这款芯片采用7nm工艺，相比于第一代性能会提高3倍。而除了更侧重于行业端的昆仑，百度面向消费者，同样已有成品，就是百度鸿鹄语音芯片。鸿鹄语音芯片于2019年7月正式推出，相比于昆仑，它面向的场景更加垂直：专门为智能家居语音交互、智能车载语音交互，以及智能物联等场景设计，是一款主打超低功耗的语音交互芯片。在芯片架构上，鸿鹄内嵌端到端一体化建模算法，能完成远场拾音、唤醒、定位等一系列语音交互功能，并将提取的特征直接传递到云端，在云端进行高精准识别。这也就使厂商能够在不损失语音体验的前提下，进一步压缩智能硬件成本。另外，鸿鹄芯片还采用了双核HiFi4架构自定义指令集，以提升算力。2020年推出的小度智能音箱2红外版上，就已经搭载了这款芯片，并因此实现了三大性能提升：语音识别错误率平均降低30%；高噪声下首次唤醒率提升10%以上，达到家居场合使用的超低误报要求；平均工作功耗仅在100mw左右，待机功率下降90%。百度造芯，关键在生态总结这样的路径，其实百度造芯的差异化竞争力，也进一步水落石出。关键还是场景和生态。自主研发AI芯片，需要的不仅是硬件技术，还需要对AI算法有深入的理解，要有软硬结合的能力。而作为国内最早开始布局AI的科技企业，如今，百度拥有从硬件到软件，从技术底层到应用场景全覆盖的AI基础设施“百度大脑”。无论是地图、搜索，还是无人车、智能音箱，百度技术多年积累和业务实践都集成其中。2020年发布的百度大脑6.0，已经开放270多项领先AI能力，汇聚230万开发者，企业发布的模型数量超过31万，是业内最全面、服务规模最大的AI开放平台。背靠这样的生态，百度造芯也就有了更多落地实践的机会。就以鸿鹄芯片举例，可以想见，小度音箱仅仅只是第一站，在未来，这样的语音芯片还可能在Apollo无人车，乃至家庭、工业场景中进一步得到验证。而站在当前这样的时间节点，把握AI时代芯片行业的新机遇，进一步从软件的自主可控向硬件的自主可控探索，不仅对于百度自身，对于整个中国的科技行业，都意义深远。所以，在小度估值200亿，Apollo也被调研机构与估值300亿美元的Cruise对标之后……百度最新孵化的AI芯片公司，又需要怎样被评价？从百度最近股价的一路上扬之中，或许可以嗅到一丝风向。且拭目以待。","news_type":1,"symbols_score_info":{"BIDU":0.9}},"isVote":1,"tweetType":1,"viewCount":831,"commentLimit":10,"likeStatus":false,"favoriteStatus":false,"reportStatus":false,"symbols":[],"verified":2,"subType":0,"readableState":1,"langContent":"EN","currentLanguage":"EN","warmUpFlag":false,"orderFlag":false,"shareable":true,"causeOfNotShareable":"","featuresForAnalytics":[],"commentAndTweetFlag":false,"andRepostAutoSelectedFlag":false,"upFlag":false,"length":3,"xxTargetLangEnum":"ORIG"},"commentList":[],"hasMoreComment":false,"orderType":2}