After years of development, cloud computing will be strengthened as more and more ICT infrastructures are in place, and many Internet giants will enter this market, but for new entrants and investors - the giants Their 'king of all' effect, as well as the surprisingly high valuation of mature companies, have turned them away.
Cloud opportunities have become fewer and fewer, and through more and more sophisticated algorithms, more powerful terminal-side computing performance, and the rapid development of network connectivity, workloads can also be executed on the terminal side.
Gartner's data shows that the total number of smartphones shipped in the next five years will reach 8.5 billion units.Has already occurred in the global smartphone market shipped 1.47 billion units in the whole of 2016. According to Qualcomm's forecast, Mobile terminals will become the world's most popular artificial intelligence platform.
The future, the terminal side of artificial intelligence will usher in tremendous opportunities for development.
Giant cloud field has been out, 'after childbirth' chances slim
When we talk about artificial intelligence, nothing more than algorithms, calculations, voice, image recognition and visual keywords.In fact, almost well-known domestic and foreign Internet giants are in full swing in these areas in the layout of the cloud Applications.
For example, Alibaba.com announced the first quarter of fiscal year 2018 (April 2017 to the end of June 2017) results show that the number of paid users in Aliyun in the quarter increased from 874,000 in the previous quarter to 1,021,000, becoming Asia's first to achieve One million user-scale cloud computing company.
And Baidu CEO Robin Li also said that financial services and cloud computing revenue contributed to (Baidu) revenue growth, such as the expansion of personal cloud services, the use of the increase.
Ma Huateng has repeatedly said that the future of the Internet, in the cloud using artificial intelligence to deal with big data.From 2015 annual report for the first time mentioned Tencent cloud, cloud business in Tencent's earnings has been repeatedly stressed.In Tencent's 2017 In the first quarterly earnings report, Tencent's cloud services and payment-related services were included in 'Other Business Revenue', an increase of 224% over the same period of last year.
In addition to competing for users in the domestic market, Ali cloud and Tencent cloud are starting overseas layout.And overseas, the pattern of the cloud computing market by Amazon, Google, Microsoft and IBM and other manufacturers control, the largest player is Amazon .As a global The largest cloud computing manufacturers Amazon, China BAT three Internet giants have the number of data center servers and less than half of them.And players such as Amazon as early as the Chinese market, has long been eyeing.
'Cloud' war intensified.It can be observed that both the 'GAFA' in the United States and the 'BAT' in China are both companies with years of development and accumulation. The battle in this area will still show a strong Matthew effect - - "Cloud" requires a lot of technical and data support. The investment in technology by small companies is hard to contend with the giants and it is difficult to obtain the data that the giants have. On the other hand, the Internet giants naturally have the advantages of content, channels and users Ecosystem complete.
In a nutshell, there are fewer and fewer opportunities for latecomers who want to be in the cloud.
Why manufacturers and capital gathers to the terminal?
Over the years, when people talk about artificial intelligence computing or storage, most focus on how to strengthen the computing power of the cloud.For mobile terminal considerations, the basic remains how to collect data faster and more abundantly.
In most people's perceptions, the processing and calculations associated with deep learning are all in the cloud, and the phone is just an interface.For example, if you want to identify an artifact, it is customary to send its image to the cloud and back again Mobile terminal.
But as the one that has been mentioned countless times, a driverless car needs to react quickly during driving. If there is no net or the network is not good, human life will encounter a tremendous threat. Even if the network is in good condition, if the system still needs to The road and vehicle information collected is uploaded to the cloud and processed before the result is obtained, which of course is more of a treat for the latter than to be done directly on the car.
Reliable - and this is one of the advantages of the terminal.Overall, the advantages of the terminal include: privacy, it is easy to understand, the problem of data privacy protection; low latency, and efficient use of network bandwidth.
In the past, most Internet giants have spent more energy in the development of the cloud. On the one hand, they are based on their original business models. On the other hand, this also has a great relationship with the development of terminal-side technologies.
The challenge with terminal-side AI is that the AI workload itself is very complex and requires a lot of intensive computing while it has an always-on and live-on-demand requirement; on the terminal side like smartphones, drones , Cars, etc., and its computing environment is very limited, so they complete the artificial intelligence workload to overcome the thermal efficiency, battery life, memory and storage space and other issues.
However, these problems are being solved or optimized with the advancement of mobile technology.Currently, mobile processors have reached the 10-nanometer process node in the process technology. The more advanced process technology means the smaller chip size and the higher Processing performance and lower power consumption.
The scale effect of mobile technology can not be underestimated.According to Gartner data, from 2017 onwards in the next 5 years, the total number of smart phone shipments will reach 8.5 billion, smart phones will undoubtedly become the world's most popular artificial intelligence platform Qualcomm vice president and senior vice president of Qualcomm Ventures Shen Jin had previously said that the current online traffic less and less, more and more expensive, after the entrance are all large and small terminals, including mobile phones, cameras, Sensors, robots, etc. We need to know these portals, and if we do not have those portals, these values will stay in the cloud, which is why it's the AI at the terminal side.
Meng Pu, chairman of Qualcomm China, made a similar point in his speech entitled "5G Achievements in Intelligent Age." He also believes that 5G will play an important role in the intelligent age of all things interconnected. Gigahertz transmission rate, lower latency and higher reliability, more importantly, it can support cellular base station to connect a large number of devices, which is 3G, 4G can not be achieved, 'future machine learning in the cloud Coordinated development with the terminal '.
Lenovo CEO Yang Yuanqing has also said that artificial intelligence includes two major aspects, perception and cognitive ability.He believes that the intelligent terminal is the human senses, the cloud is the brain, the smart terminal and cloud brain perfect combination is artificial intelligence The future direction.
For a time, terminal computing is more and more optimistic, manufacturers and capital began to gather closer to the terminal.
In early September, Huawei unveiled Kirin 970 with dedicated NPU hardware processing unit. Intel acquired several companies such as Altera and Yogitech successively and gained relevant technologies in fields such as FPGAs. Softbank, with a market value of 23.4 billion pounds Price acquisition of ARM, intended for embedded artificial intelligence; Tesla did not bring the case of Nvidia quietly come up with unmanned car AI chip in production test. Just released a new generation of iPhone, all equipped with Apple A11 Bionic processing It also comes equipped with a neural engine that specializes in AI requirements.
Capital, cloud-related company valuation has become more and more impossible to climb, the terminal chip has recently become a hot area of the AI field.
August this year, Cambrian Technology announced the completion of 100 million USdollars A round of financing, becoming the world's first AI chip field unicorn company; the end of October, the domestic AI chip startup deep Kam Technology completed A + round of financing, its investors include Ant Financial Services, Samsung Ventures, China Merchants Ventures and CRE Capital, raising about 40 million U.S. dollars ...
Qualcomm focuses on terminal-side AI
Although artificial intelligence is a concept that has been proposed only in recent years, Qualcomm has been working in artificial intelligence for more than a decade as a leader in global mobile communications.
In 2007, Qualcomm began exploring machine-learning impulsive neural approaches to computer vision and motion control applications and subsequently expanded its research range from bionic approaches to artificial neural networks - primarily to deep learning. In July 2017, Qualcomm began Developers are provided with Snapdragon Neural Processing Engine (SNPE) software development kits. SNPE is the first deep learning software framework designed for Snapdragon mobile platforms to support TensorFlow, Caffe / Caffe2, with autochanger Modify, analyze performance through the SDK, and then integrate. It provides developers with software tools that accelerate deep neural network workloads for mobile terminals equipped with Snapdragon mobile platforms and other types of network edge terminals.
Not long ago, Qualcomm had just announced the acquisition of Scyfer BV, a subsidiary of the University of Amsterdam, specializing in cutting-edge machine learning technology to complement its team of talented people.Scyfer has built artificial intelligence solutions for companies in many different industries worldwide, covering manufacturing, medical Industry and finance.
On the morning of October 20, ShangTang, the largest artificial intelligence startup in China, jointly announced with Qualcomm that both parties will cooperate on artificial intelligence (AI) and machine learning (ML) around mobile terminals and the Internet of Things. Technology algorithms will be targeted specifically for Qualcomm's Snapdragon platform, while Qualcomm's Snapdragon chips will be open to Shang Tang's ability, which will give full play to the algorithm + chip integration advantages.
From an investment perspective, Qualcomm Ventures, Qualcomm's venture capital arm, has invested nearly 40 companies since entering China in 2003, including many in cutting-edge technology companies, Related companies and projects in the field of investment include: The US brain, which takes seven years from researching people's brains to providing software for robots for mobility and navigation; and US Face Recognition Company Clarifai; China smart voice company cloud knows sound and so on.
Shen Jin has said publicly that the scale-up of smartphones can result in rapid technology iterations, rapid cost reductions, reduced cell phone power consumption, and reduced cell phone size, while the economies of scale of smartphones will also be radiated to other industries, including Various industries that require the use of silicon and hardware, such as smart city, automotive, healthcare, networking, wearable devices, industrial internet of things, smart home and more.
For entrepreneurs and investors, there will continue to be plenty of opportunities for areas around end-intelligence. "I believe IoT + AI is a good place to start if you focus on smart hardware or the Internet of Things. .