1.IHS: Ranking of the Top 10 Semiconductor Suppliers in the World, Nvidia Ranked for the First Time;
According to market research firm IHS Markit, Nvidia ranked among the world's top ten semiconductor suppliers for chip sales in 2017 for the first time; and only the company and Qualcomm (Qualcomm) are strictly non-existent in the top ten list. Fabs (fabless) chip design company.
IHS Markit pointed out that Nvidia’s total sales in 2017 was US$8.57 billion, which is enough to make the company ranked the world’s tenth largest chip supplier; but Len Jelinek, director of IHS technology, media and telecommunications department and chief analyst, said that the rise of Nvidia allows Taiwan's fabless chip designer MediaTek fell out of the top ten list.
Qualcomm, Nvidia, and MediaTek are the only fabless chip suppliers in the world's top ten list of chip suppliers that have ever entered the world's top ten; MediaTek had entered the world in 2014 and 2016. The top ten chip supplier list, Qualcomm entered the top ten for the first time in 2007 and remains on the list.
Overall, global semiconductor sales grew by 21.% in 2017 to reach a record high of US$429.2 billion; this is the highest annual growth rate for the industry in the last 14 years.
Most of the chip industry’s sales growth in 2017 came from the sales performance of memory products; the memory market’s sales in 2017 increased by 60.8% over 2016. In addition to memory, the average sales of other types of components of the semiconductor industry last year grew by 9.9%. The main reason is what IHS calls "solid unit-sales growth" and the strong demand across all application areas, regional markets and technology categories.
Craig Stice, senior director of IHS Markit's memory and storage market, said in a press statement that NAND prices that have risen in 2017 are expected to decline this year.
"In 2018, the 3D NAND conversion rate will account for almost three quarters of the total NAND output, and it is expected to relieve the strong demand from the SSD and mobile phone markets." Stice said: "Prices are expected to begin to drop substantially, but NAND in 2018 The market may still set a record high." Eettaiwan
2. Botong wants to repurchase 12 billion U.S. dollars of shares that have been approved by the board of directors;
Beijing morning news on April 13th, Broadcom just gave up the plan to acquire competitor Qualcomm last month, the company said today that the company's board of directors approved their plan to buy back $ 12 billion shares.
Broadcom is headquartered in San Jose, California, and their main business is the production of smart phone chips. The company said in a statement issued on Thursday that they will use a variety of means for stock repurchases, including through the open market and private transactions. Repurchase. Broadcom also revealed that they may suspend or close the repurchase program at any time.
Last month Broadcom just moved its headquarters from Singapore to the United States. They originally planned to acquire Qualcomm, the biggest competitor, but President Donald Trump called for a halt to 'possibly threatening national security'. The deal.
After the stock repurchase program was announced, Broadcom's stock price rose by about 4%, and the current stock price was 239.43 US dollars per share. Based on current prices, Broadcom's market value is 98.3 billion U.S. dollars.
3. Can machine learning solve the challenges of complex 5G baseband technology?
Zeev Kaplan, Communication Algorithms Team Leader, CEVA
The World Mobile Congress (MWC) held in Barcelona this year is full of exciting product announcements and announcements. Although a few products are trying to rekindle the classic, like the Nokia 8110 4G mobile phone is in accordance with the original "Matriarchal" movie retro transformation from But most of the products are looking to the future. Two of these technologies represent especially the future of the mobile world: artificial intelligence (AI) and 5G communications. But what is the link between these two domains? Continue reading the following article to find answer.
Artificial intelligence is the most popular pop word, but how much authenticity is there?
Artificial intelligence is a huge business opportunity. According to KPMG statistics, artificial intelligence and machine learning achieved a $12 billion VC in 2017, which has doubled from the previous year. But like every hot trend, the use of this term is far reaching. It goes far beyond its true meaning. Don't get me wrong. From robots and robots to smart speakers and headphones, artificial intelligence is changing the world. However, many products use artificial intelligence as a selling point, but they are actually based on Good algorithm, so it is very important to distinguish between popular vocabulary and reality.
The use of artificial intelligence may not be to create a humanoid machine that is as powerful as a human being in the movie Humans, or a robot in Westworld. It can be applied more generally. Professional application. But to be recognized as artificial intelligence, it will include a learning process and generate a result that may not be achieved through simple brute calculations. This type of artificial intelligence is becoming very common, especially around deep learning. And neural networks, many of the technologies they drive are used by millions of people every day.
5G Challenge: Extremely high speed with high complexity
Another huge topic of MWC is 5G. My colleagues gave a good overview of the 5G situation in mid-2017. It is impressive to see how far 5G technology has developed. 5G is expected to become a revolutionary technology. , Driving the development of applications such as self-driving vehicles, smart homes and smart cities, mobile augmentation and virtual reality as well as 4K video streaming. Such expectations have been going on for a while, and now, the potential of 5G is beginning to crystallize.
One example is Nokia's recent announcement of ReefShark Soc preparing for the 5G challenge, including the CEVA-XC architecture for wireless communications. According to Reuters, even Donald Trump’s security team hopes to accelerate the start of US 5G network construction. As a response to threats from overseas networks and economic security, it is now safe to say that everyone knows that 5G is the future of communications.
There are many reasons why 5G is satisfying. Its bandwidth will be much larger than 4G, and it can support streaming media content that was not available before. The reliability is greatly enhanced, enabling critical missions and extremely low latency scenarios such as automatic driving to be guaranteed. It will also dramatically increase the number of devices that can be connected, and the connected devices will grow exponentially to form the Internet of Things (IoT). Therefore, from the smallest sensor in the home to the most sophisticated automated vehicle, all will benefit from 5G technology.
Extremely complex calculations also bring problems. For example, millimeter-wave band technology up to 60 GHz, massive MIMO (multi-input and multi-output support for 64-256 antennas) and low-band sub-6 GHz frequencies, for the 5G physical layer (PHY) The introduction of new challenges and a large number of complex and non-linear calculations. In addition to all these complexities, the 5G definition has not yet been formally standardized. 5G New Airtime (NR) release-15 is the latest release of 3GPP, but all current chips must be maintained Support release-16 flexibility, as well as other versions of non-3GPP standards. To meet these requirements requires a completely new approach to the past.
Take 5G challenges with machine learning and powerful DSP
At our MWC booth, we introduce solutions that address these daunting challenges. CEVA PentaG is an advanced 5G NR IP platform for smartphones and broadband connectivity devices. This unique solution combines us as a leader in communications DSPs. Our extensive experience, and our expertise in artificial intelligence, especially machine learning. The platform integrates an enhanced version of the CEVA-XC4500 DSP plus an advanced artificial intelligence processor, a powerful vector MAC unit (VMU) co-processing , a set of CEVA-X2 DSPs and optimized hardware accelerators.
This combination has the unique ability to handle 5G heavy calculations with very low power consumption and the flexibility of future software updates. Our demo at MWC shows that PentaG can handle complex channel state information (CSI) and supreme support This challenging application scenario for MIMO4x4 and 256 QAM. CEVA's innovative approach is to use artificial intelligence to handle 5G CSI calculations, enabling 5G advanced receivers (eg quasi-maximum likelihood MIMO decoders) to achieve the highest throughput. Intelligence helps narrow the gap between performance mismatches and overcome the complexity barriers of existing methods. It's great to see that everyone shows a great interest in our technology and participates in it. If you miss the demo, be sure to follow us and seize it. Opportunity for the next show.
4. MLCC out of stock Supply chain preparation;
The supply of passive components is tight and the price has multiplied several times. It is also reported that the supply chain may be in crisis due to shortage of materials. In the face of a series of component price increases, the exchange rate rises, and there are still problems with the shortage of components. Quanta (2382 ), Compal, Inventec, Wistron and Heshuo and other assembly plants, in addition to stocking in the off-season, also reduced the amount of spare parts by redesigning products; but Netcom factory is quite nervous, and the general manager ran to the supplier Get goods.
In terms of shipping, Quanta stated that MLCC's capacity expansion is not easy due to lack of MLCCs, and that MLCC's expansion is easy. In the first half of the year, it is off-season. Demand is not very good and will not be affected. Compal, Inventec, Wistron Heshuo said that the off-season will be prepared to supply tight parts first and ensure normal shipments in the second half of the year. This is also a normal procedure, mainly because the price of parts will increase, and the two years will be pre-stocked.
However, the lack of spare parts has seriously affected the order shipment rate and revenue performance of China Netcom's factories. Each of China Netcom’s factories has been tempted to pick up stocks. Cai Wenhe, the chairman of the company’s chairman, said that the first half year of stocking should not affect shipments. Cai Wenhe stated that MLCC is expected to (Laminated Ceramic Capacitors) Out-of-stock condition may be relieved in the 4th quarter, so the strategy has been improved inventory, stocking half a year.
Including passive components such as MLCCs (Laminated Ceramic Capacitors) are the most basic components of all electronic products. However, passive components began to appear tightly in supply since last year, and prices have continued to rise. More recently, passive component suppliers have asked for bids from customers. Can get the goods.
For this purpose, Chen Ruicong, general manager of Compal said earlier that since many raw materials, including memory, MLCC, PCB, and even paper, have risen together, and the cost pressure is high, EQ customers have understood that for the newly developed products, customers are expected to share some of them. Price increase costs.
Quanta, Wistron, said that it is discussing with brand customers to change the design, hoping to reduce the use of related parts to reduce the impact. Wistron pointed out that through product change design, the number of MLCCs used in some products can be One-third to one-half. Inventec also stated that it is reducing costs by finding alternative materials.
A China Netcom Factory said that in recent quarters, the shortage of parts and components has risen again, and the relationship with suppliers has also been likened to three warmths. At the beginning, the lack of goods was the reason why suppliers called out to the office to get a stinky head, forcing priority to supply. , After the lack of goods, everyone was scared. "Please go to the supplier's office yourself."
Netcom Factory said that a product has many parts and components assembled together. It is indispensable and the materials are all available. There is no single MLCC or memory, and the entire list cannot be shipped. When the parts are out of stock, they are even more upstream. Customers circle directly to component suppliers to ensure orders are shipped in a timely manner. Economic Daily
5.Samsung expands DRAM production capacity, and capital expenditure will increase by 8% this year;
With Samsung Electronics expanding its DRAM production capacity, analysts estimate that capital expenditures of manufacturers in this (2018) year are expected to grow by 8%, and the prospects for semiconductor device groups are bright.
Barron`s.com reported on the 12th that Keybanc analyst Weston Twigg published a research report stating that chip makers’ capital spending on equipment this year is expected to grow by 8%, which is higher than the 5% he previously estimated, and demand has a chance to increase by 10%. In the above, the demand for logic IC devices is relatively stable. Although foundries are somewhat weak in the first half of the year, they should improve in the second half of the year.
Twigg said that Samsung seems to be actively expanding its DRAM production capacity. It is estimated that the demand for memory will be stronger than that estimated by the equipment makers before the end of the previous quarter. DRAM expansion and 3D NAND production capacity will continue to increase, and it is hoped that the capital expenditures of chip manufacturers will increase all the way to the end of this year. Although Samsung has basically stopped expanding its production of OLEDs and significantly reduced the utilization of existing OLED production capacity, Applied Materials has a relatively low exposure to this factor and its OLED related revenue accounts for only 7% of the total.
Applied materials rose 2.69% to $56.43 on the 12th, returned to the 60-day moving average, and hit a close on March 26th. Lam Research Corp., a manufacturer of semiconductor etching machines, rose 2.57%. At US$206.02, it was the highest since March 26. The KLA-Tencor Corporation, a wafer inspection equipment manufacturer, also rose 2.33% to US$109.12.
Taiwan Semiconductor Manufacturing Co. (2330), Samsung’s interest in advanced process competitions will also be very beneficial to semiconductor equipment manufacturers.
The European Department of Foreign Investment Research released on April 9 pointed out that recent rumours suggest that TSMC may postpone the delivery of orders for EUV lithography equipment. However, according to field surveys, the demand outlook for EUV has not diminished. On the contrary, there are indications of an increase in temperature. It is estimated that ASML Holding NV, the leading semiconductor equipment manufacturer in Europe, will know that the EUV production capacity for (2019) should be snatched up as scheduled.
The report pointed out that TSMC has already placed orders of nearly US$1 billion on ASML and Lasertec, the Japanese semiconductor equipment manufacturer, on April 2. It shows that the number of equipment ordered in the first quarter should be 6-8. Samsung is the most active attitude towards EUV. According to the survey, Samsung may introduce at least part of its 1 y nm DRAM (~1 layer) production capacity into EUV. The time for mass production of DRAM will be even earlier than that of 7nm foundry/logic chips. , because Samsung intends to use a lower risk DRAM trial production first, so that in the future a large number of 7-nanometer process introduction EUV.
6. Luminar key parts cost down to 3 USD
Li Shan Compile and arrange the qubits produced | Public number QbitAI
Lidar is also a huge stumbling block in the development of the unmanned vehicle industry because it is more expensive than a car and is always out of stock.
Luminar, a startup company of Austin Russell, a Stanford drop-out teenager, today released a new product with a 120-degree angle of view. At the same time, it also shows that cheap and efficient production of laser radar has become possible.
A year ago, Luminar, who completed a $36 million seed round of financing, showed off the products he had polished from scratch: Their laser radars used wavelengths of 1550 nanometer lasers, which had a detection range of over 200 meters and a resolution 50 times higher than similar products.
People who have seen the product comparison demo told the media that the Luminar lidar's images are clearer than those of Velodyne and Quanergy.
Russell dismissed competing products as saying that they 'cannot meet the safety requirements of self-driving cars'. However, their own products are also facing a difficult and costly production dilemma.
Now, things are finally getting better. Russell said: 'Last year took a whole day to make a device - it needs to be manually assembled by Dr. Optics. Now we have 136,000 square feet (12,600 square meters) of manufacturing. Center, it only takes 8 minutes to assemble a set. '
Throughout 2017, Luminar produced only 100 lidars. By the end of the year, they plan to increase production capacity to 5,000 units per quarter.
But there is no need to worry that the company will trade off quality by sacrificing quality. Their systems are 30% lighter and more energy-efficient. The effective distance is increased from 200 meters to 250 meters, and even lower reflectivity can be detected. Objects - such as people walking in the dark in black clothes.
Luminar's lidar is basically the same principle as its kind: emits a laser beam, and then calculates the return time. They all need a receiver, a photosensitive surface to identify the returned photons.
But Luminar's receiver is not the same as someone's home.
Most of the photoreceptors in digital cameras and other LiDAR systems are based on silicon. This material is fully researched, the price is very cheap, and the manufacturing process is also very mature.
But Luminar's system was developed from scratch, using a compound called InGaAs. InGaAs-based photoreceptors can be used for different optical frequencies (1550nm, not around 900), and the capture efficiency is drastically improved. Promote.
The general rule is: The more light that is obtained, the better the sensor. The same applies here. Luminar's InGaAs receiver and a single radar transmitter produce images far better than those of similar size or power. There are fewer moving parts.
Like most of the powerful things that are powerful, InGaAs has one drawback: It's too expensive, it also requires specialized design capabilities, and the talents who can use it are scarce...
To solve this problem, Luminar can only minimize the use of indium gallium arsenide, and only use this material when necessary. They have taken some engineering measures instead of using a series of photovoltaics like other LiDAR systems. detector.
Now, Luminar finally announced that they drastically reduced the cost of the receiver, from tens of thousands of dollars to 3 dollars.
How to do it? The answer is to acquire a like-minded, but also very powerful chip company.
Last year, Luminar and a company called Black Forest Engineering co-designed chips and found that their methods were extremely similar. Luminar acquired the company.
At the same time, Luminar also introduced internal designers to customize its own optoelectronic receivers and various chips. 'We developed our own ASIC. We used only a little indium gallium arsenide to put it in the chip. We also customized Chip. 'Russell said.
Today, Luminar has about 350 employees, including 30 from Black Forest and 200 later recruited.
The company announced in September last year that it had reached a cooperation with Toyota. It is said that there are three other car manufacturers working with them. However, the addition of Luminar LiDAR is still not clear, Russell only said that after mass production, each set Lidar's production cost is much lower than $1,000'.
Qubit