In 1.2018, the semiconductor market may exceed 500 billion US dollars, and the price reduction of NAND flash memory dragged down the annual growth rate;
Original title: The global semiconductor market is expected to exceed 500 billion US dollars for the first time in 2018, NAND flash memory price reduction or dragging the annual growth rate
Abstract: IC insights predicts that the global electronic systems market will grow by 5% to $1.622 trillion in 2018. The global semiconductor market is expected to grow 14% this year to $509.1 billion, which is the first time to exceed $500 billion. If the forecast for 2018 Realization, the average semiconductor content in the electronic system will reach 31.4%, breaking the record of 28.8% set in 2017.
In the upcoming 2018 McClean report mid-year update, IC Insights predicts that the global electronic systems market will grow by 5% to $1.622 trillion in 2018, and the global semiconductor market is expected to grow this year. 14%, to 509.1 billion US dollars, is expected to exceed 500 billion US dollars for the first time. If the forecast for 2018 is realized, the semiconductor cost in electronic systems will reach 31.4%, breaking the record of 28.8% set in 2017.
Historically, compared to the electronic systems market, the semiconductor industry’s average annual growth rate is higher than the value or usage of semiconductors used in electronic systems. As global handsets, automotive and personal computer shipments are expected to There was a weakness in 2018, and the steady development of the electronic system market led to high growth in the semiconductor market due to the increase in semiconductor usage in electronic systems.
In the past 30 years, the use of semiconductors in electronic systems has been increasing. The increase in semiconductor usage in electronic systems in 2018 was mainly due to the surge in sales of DRAM and NAND flash memory and the increase in the price of electronic systems this year. IC Insights expects that from 2018 From 2022 to 2022, the proportion of semiconductors in electronic systems will not be lower than 30%, and may fall to 30.2% in 2020, but will reach 31.5% in 2022, once again hitting a new high.
At the same time, it is worth noting that in the first five months of this year, the growth rate of the semiconductor industry is more than 20%, but IC Insight expects the annual growth rate to drop to 14%. This is closely related to the price reduction of NAND flash memory.
The trend of increasing semiconductor value in electronic systems is limited. Considering other materials and software costs, semiconductor costs will not rise to 100%. In the future, when semiconductor cost ratio is close to the limit, the semiconductor industry growth rate will also In the system equipment market, the growth rate is very close, for example, 4% to 5% per year. (Proofreading/Lechuan)
2. Self-driving drive 2020, the amount of each car chip reached 1,500 US dollars, FD-SOI became a big winner?
Original title: Auto-driving drive 2020 dollars per car chip consumption, FD-SOI become a big winner?
Abstract: FD-SOI technology can be used in chips or sensors such as emerging ADAS, car networking, infotainment system, power system, etc. FD-SOI is widely used in automobiles. In order to be self-reliant, it is obviously necessary in China. Develop FD-SOI technology.
Episode report (Wen Ai Meng) At the China IC Ecology Summit Forum held in Qingcheng Mountain today, autopilot has become the hottest topic. Although the whole industry chain has been working together, the biggest winner is FD-SOI Ecology. Circle 'partner', why is this?
Auto driving into a big push
When the semiconductor process was developed to 22 nanometers, in order to meet the performance, cost and power requirements, the FinFET and FD-SOI technologies were extended. Since the semiconductor leader Intel led the promotion of FinFET technology, and obtained the foundry of the wafer foundry TSMC The support of FinFET technology has become more and more popular. However, in recent years, FD-SOI technology has attracted more and more attention from the industry. The investment in FD-SOI technology by grid manufacturers, Samsung, Sony, ST, and Core Microelectronics is coming. The bigger the technology, the more its technical advantages and application prospects are optimistic.
Carlos Mazure, president of the Global FD-SOI Industry Alliance, shared relevant data. He said that the automotive market is growing steadily, with a very fast growth rate, with an average of more than eight points globally, while China's growth rate is even faster, reaching 12.4%. In 2020, the average chip on each car will reach $1,500.
From the relevant data, on the one hand, one of the advantages of FD-SOI is the use of forward body bias technology (FBB), which can be controlled by software to achieve power consumption, dynamic balance of performance, and easy to integrate RF such as transceivers and non-easy Lossless storage unit has unique advantages in the field of automotive electronics. On the other hand, automotive intelligence, network integration makes edge AI and distributed AI necessary, requires higher performance and lower power consumption; 5G is increasingly mature, requiring RF High integration; automotive ADAS requires radar integration to reduce costs, etc., FD-SOI technology advantages and application requirements successfully 'docking', making FD-SOI usher in the best time.
OEM involvement to optimize the underlying architecture
After years of development, the FD-SOI ecosystem has covered tool vendors, IP vendors, design service vendors, chip vendors, manufacturers, etc. to provide easy-to-access plug-and-play solutions to minimize customer costs.
FD-SOI has a longer-term goal. Carlos Mazure said that in order to enable different elements to accumulate, the entire ecosystem can make decisions together and realize the interests of all parties, and recently had the opportunity to invite Audi to explain their needs. Carlos Mazure As mentioned, Audi believes that power consumption is the main obstacle for automotive electronics to break through. Therefore, a three-layer architecture is proposed, the most basic is the sensing brake layer, then the scalable calculation, and finally the seamless Car2X data integration.
Carlos Mazure analyzes that Audi, Google, Apple, etc. are moving from the top to the bottom of the industry chain, allowing them to get in touch with the infrastructure, which allows them to better participate in the entire chain and optimize the underlying architecture system together. Autonomous driving and intelligent network connection. Future AI, 5G and intelligent driving are tightly coupled, which will greatly promote the development of FD-SOI.
Mark Granger, vice president of e-network and network business development of Gexin Automotive, also mentioned that Gexin will invest more in Chengdu to become Chengdu's key industrial base for FD-SOI. Currently, in terms of ecological construction, Gexin and Chengdu government have already cooperated. With 75 partners, it can provide 150 IP projects, from design IP to technical support, and continuously expand.
Technology continues to advance
Autopilot not only carries the high hopes of FD-SOI, but also puts forward higher requirements, such as more stringent standards, integration of more emerging technologies, and so on.
Mark Granger mentioned that Gexin's Autopro technology solution focused on the car has been developed for ten years. For the new standards such as IS26262 and AEC-Q100, relevant certifications have been carried out to achieve better safety and reliability. High voltage BCD, millimeter wave radar and other technologies, as well as reducing cost and power consumption, the core 22FDX technology has been able to cope.
'FD-SOI has cumulatively developed $2 billion in products. 'Mark Granger mentioned. Of course, Mark Granger also admitted that in order to achieve zero defects, zero drift, and better feasibility and safety, this is still a very large Challenge, need to gather the wisdom and strength of partners.
For the potential of FD-SOI, China should also step up preparations. Domestic semiconductor industry expert Mo Dakang once wrote that the research on FD-SOI has formed process research, wafer fabrication, IP, foundry, IC design service company, IC design company. The complete industrial chain. Mo Dakang believes that China's semiconductor industry is in a special environment. In order to be self-reliant, it is clear that FD-SOI technology needs to be developed. This is undoubted. Especially in the process of catching up with foreign counterparts, FD-SOI technology is probably one of the most reliable technologies. Some manufacturers in China's semiconductor industry have been involved in FD-SOI, and I believe that there will be even more amazing performance in the follow-up.
3. South Korea plans to increase support for chip projects to meet the challenges of Chinese opponents;
Sina Technology News Beijing time on July 20 evening news, South Korea's trade, industry and energy minister Bai Yi (Paik Un-gyu) recently said that the Korean government will increase support for large-scale research and development projects to develop cutting-edge memory chips, China The rise of competitors.
Currently, the chip is Korea's largest export product, and China is the world's largest chip market. However, China plans to invest huge amounts of money to promote the development of its domestic chip industry, in order to reduce dependence on foreign products, leading to the industry's long-term commitment to the Korean chip industry. The outlook is worried.
To this end, Bai Yunyi said during the Korean National Assembly group meeting: 'The government is considering supporting large-scale projects to design and produce next-generation memory chips, and plans to conduct preliminary feasibility studies in the second half of this year'
Although South Korean technology giants such as Samsung Electronics and SK Hynix have developed world-class memory chip products through technological innovation, there are still experts calling on the Korean government to pay attention to the industry in response to the challenges of the fast-rising Chinese competitors.
In fact, Korean local chip makers are also worried that the Chinese government's move may narrow the technology gap with South Korea faster than expected. By then, if the production is not carefully controlled, it will lead to a global chip price crash.
Baiyun said: 'China is stepping up its own semiconductor industry to narrow the gap with South Korea, which may lead to global oversupply. Recently, the growth rate of memory chip prices has slowed down, causing some people to think that 'super cycle' is positive Close to the summit. ' (Li Ming)
4. Samsung: In the 2020 development of 3 nanometer process, the chip design fee will be as high as 1.5 billion;
Samsung Electronics previously released the 3nm Foundry process by 2020. According to the analysis, the 3nm Foundry process chip design cost will be as high as 1.5 billion US dollars. Although the chip design cost is extremely high, but according to expert analysis, its current efficiency and performance The increase is not directly proportional to the cost, and given the high cost, there are only a handful of companies that can design 3 nanometer projects.
On July 17th, the International Business Strategy (IBS) of the semiconductor market said that the chip design cost of the 3 nanometer chip project will be as high as 400 million to 1.5 billion US dollars. IBS shows that the design cost of the chip is relatively high in the design complexity of the GPU. The company's data shows that the average design cost for a 28-nanometer chip is $5,130, while the 7-nanometer chip design with FinFET technology costs $20.97 million, a nearly six-fold increase. The design cost of a semiconductor chip includes IP, Architecture, Inspection, physical verification, software, trial production, etc.
This is also the reason why the semiconductor industry's Fabless factory has always favored the 16nm FinFET and Samsung's 14nm FinFET process. For Foundry manufacturers, the cost is also quite a headache, not only that the 3nm process is quite difficult.
Samsung Electronics' 3nm process will use GAAE (Gate-All-AroundEarly), GAAP (Gate-All-Around Plus) technology for the first time, and will be called MBCFET (MultiBridge Channel FET). The core of the technology is to ensure the reliability of each Gate channel. Current exists. If the FinFET structure is 3-sided current, GAA is guaranteed to have current on all sides of Gate. As the current channel becomes larger, the performance will also increase.
Samsung Electronics' MBCFET technology is jointly developed by IBM and GF in the United States. If the FinFET is vertically arranged in the fish scale, the GAA is the horizontal superposition arrangement of the Gate. In order to produce such a Gate structure, a series of pattern development, evaporation, etching, etc. are required. Engineering innovations, and in order to reduce parasitic capacitance, it is also necessary to introduce new materials such as cobalt, niobium, etc. instead of copper.
According to industry sources: 3 nanometer engineering development and chip design are feasible if supported by huge financial resources, but the key is whether it is worth investing. And the companies that can use such projects are only Qualcomm, Apple, NVIDIA, Apple, etc. A few companies, this will also be one of the stumbling blocks of 3nm engineering. ETNews
5.7nm 56Gbps SerDes blessing, this ASIC or new ideas for AI chip packaging;
Abstract: Recently, eSilicon launched the NeuASIC ASIC platform manufactured by TSMC's 7nm process. The 56Gbps SerDes core under this platform was designed by the former Marvell Italian team that joined in 2017.
Episode News (Text / Xiaobei) Recently, eSilicon introduced the NeuASIC ASIC design platform manufactured by TSMC's 7nm process, including hardware and software macro commands for network applications and new architectures and IP libraries for building AI accelerators.
The NeuASIC platform provides designers with a variety of power-optimized memory compilers, SerDes and 2.5D IC packages. The 7nm library includes 56Gbps SerDes, HBM2 PHY, Tri-State Content Addressable Memory (TCAM) compiler, network-optimized I/O and Other components.
In 2017, Marvell closed most of its European operations, and eSilicon thus “acquired” Marvell's team of Italian engineers who developed a 56 Gbps SerDes for 28-nm process for Marvell. The team developed with the same architecture based on ADC/DSP. The 7nm 56Gbps SerDes is out, and the core appears on the NeuASIC platform. At the same time, the core can be licensed separately. For the chip, power consumption and performance seem to be two indicators that cannot be considered at the same time. The SerDes core PAM4 and NRZ encoding can be implemented, and its programmability allows designers to perform long/short channel performance and power consumption adjustments.
SerDes is short for Serializer/Deserializer. As the name implies, it is a serializer and a deserializer. However, SerDes is only described as a serializer and a deserializer. This explanation is not complete. In addition to the serializer and deserializer, the SerDes system also Including the driver side of the transmitter and the analog front end of the receiver. For the low-speed SerDes system, the design of the analog front end is less difficult and the power consumption is low. The use of the ADC will increase the design difficulty of the system. For the high-speed SerDes system, the implementation is high. The high speed ADC of accuracy is inherently more expensive than implementing an analog front end.
In April, MediaTek launched the industry's first 7nm 56G PAM4 SerDes IP. The solution is based on DSP technology and uses high-speed transmission signal PAM4, which is expected to be available in the second half of 2018.
Both eSilicon and MediaTek's SerDes solutions can achieve 56Gbps and adopt a 7nm process, which may form a competitive relationship in the future.
SerDes is part of the NeuASIC platform, and 'communication' is one of its important tasks. The implementation of NeuASIC AI performance is mainly in its AI accelerator, etc. The integration of AI accelerators is relatively 'new', and this has a large package with NeuASIC. relationship.
In order to maximize the bandwidth of the memory, the eSilicon NeuASIC network communication chip stacks the ASIC and DRAM through a silicon interposer and encapsulates it with 2.5D package technology. For AI accelerators, NeuASIC allows designers to implement Deep Learning Accelerator (DLA). Converged into the ASIC, as shown below. The industry believes that this is a new way. (Proofreading / Jimmy)
6. TSMC's Q2 revenue net profit both fell, 7nm will become 2H main force, no listing plan in the mainland
Original title: TSMC: Q2 revenues both fell, 7nm will become the main force in the second half of the year, no listing plan in the mainland
Abstract: At the second quarter of the law conference held yesterday, TSMC CEO Liu Deyin said that according to the new wave of tariffs issued by the United States, no TSMC customers were listed in the taxable list, so there was no impact on TSMC. Whether the United States has a third wave of tariff lists, TSMC closely observes.
Set micro-network news (text / Mao Mao), TSMC yesterday (19th) held a law conference, and announced the second quarter financial report, this is the first financial report after the official retirement of the chairman Zhang Zhongmou. TSMC's Q2 revenue was US$7.6 billion, down 5.97% from the previous quarter; net profit was US$2.359 billion, down 19.49% from the previous month.
In response to the decline in Q2 revenue and net profit, TSMC Finance Chief and spokesperson He Limei said that it is mainly affected by the seasonal factors of mobile terminal products, but the cryptocurrency mining operation continues to have strong demand and a favorable exchange rate environment. It slows down the impact of weak revenues on mobile terminal products.
Re-adjust this year's revenue forecast and capital expenditure
Looking forward to Q3, TSMC CEO Wei Zhejia said that thanks to the release of a new generation of iPhones in the second half of the year and the benefits of 7nm process chips, TSMC's third quarter revenue is expected to pick up. TSMC estimates Q3 revenues at 8.45 billion to 85.5 The profit margin of Q3 is expected to be between 48% and 50%.
In addition, future smartphones, automobiles, high-performance computing and the Internet of Things will further drive the semiconductor industry to continue to grow, and AI and 5G will bring significant changes to human life.
Although TSMC expects Q3 revenue performance to meet market expectations, TSMC has lowered its full-year outlook. TSMC estimates that if annual revenue is calculated in US dollars, the annual growth rate is high single digit, about 7-9%, low. In the last season, the law said that the estimated 10%, this is also the second time this year TSMC lowered its annual forecast.
It is reported that at the law conference held by TSMC at the beginning of this year, Zhang Zhongmou said that it is estimated that this year's revenue is expected to grow by 10-15%. In the first quarter of the law conference, TSMC will reduce its annual revenue growth this year. For 10%, TSMC said that the reason for the downgrade is because of the poor demand for smartphones.
Now, once again, the annual revenue growth rate is lowered. TSMC said that the main reason is that the demand for virtual currency has begun to cool down. At the same time, this year's smartphone-related chip revenue is expected to decline compared with last year, which also led to lower-than-expected TSMC revenue performance.
In addition to lowering its full-year revenue growth forecast, TSMC said it will cut its capital expenditure plan this year, from $11.5 billion to $12 billion to $10 billion to $10.5 billion. He Limei said that capital expenditures have decreased by about $1.5 billion this year. There are three main reasons:
The first is the deferral of equipment payments. The time is postponed from this year to 2019. This part affects about 700 million US dollars. The second reason is that TSMC's production efficiency is improved, and equipment parts can be shared, thus reducing some equipment purchases.
The third reason is that the US dollar is strong. Originally, some of TSMC's equipment payment currency is in euros and yen, but the strong dollar makes these currencies depreciate, and the amount of payment is reduced.
TSMC is confident in 7 nanometers
In TSMC's second-quarter earnings report, 7-nanometer revenue has not yet been reflected. In this regard, TSMC said that 7-nanometer is progressing as planned, and it is expected to contribute revenue in the second half of the year. Meanwhile, TSMC is confident in the development of 7-nanometer process. It is expected that the proportion of 7nm process in the third quarter will reach 10%. The proportion of the 7nm process in the fourth quarter is expected to reach 20%. It will contribute more than 20% next year, that is, the performance of 7nm process will surge next year. More than double.
TSMC said that the number of customers using 7nm will increase from 20/16/10nm in the second half of this year. At present, TSMC has dominated the 7nm process node and won large orders from many large customers. It is reported that TSMC's latest InFO technology has been approved by Apple, enabling it to obtain orders for the A12 processor for the iPhone released this year. In addition, TSMC will also give Huawei Unicorn, Qualcomm, AMD, NVIDIA, etc. in the second half of the year. Yujia customers produce new chips.
For the enhanced 7-nanometer process, TSMC is expected to mass-produce in the second quarter of next year, and it is confident that it will be the first wafer foundry to use extreme ultraviolet (EUV) light.
In addition, TSMC also said that it will conduct risk trial production of 5 nanometers in the first half of next year, while 3 nanometers have set up factories in Nanke, with an investment of over 20 billion US dollars.
The second batch of tax increase list does not involve TSMC customers, no IPO plan to the mainland
However, TSMC currently accounts for more than 50% of the global foundry market. Bloomberg columnist Gao Canming believes that the European Commission has imposed a huge fine of 4.3 billion euros on Google. TSMC should be the next locked antitrust for the EU. Goal preparation.
TSMC told shareholders last November that the European Commission is conducting a preliminary competition investigation on the company, but the market has not paid much attention to this matter. According to Mlex's information, TSMC was investigated by the EU for allegedly' Involved in the abuse of market forces, illegally excluded competitors in the semiconductor market, including the inclusion of improper conditions in semiconductor supply contracts. The US Fair Trade Commission is also conducting a similar investigation.
This week, the EU issued a $5 billion sky-high price ticket to Google. Gao Canming believes that TSMC executives and investors should not underestimate the power of the EU's competition authority, Weistag.
As for Sino-US trade warfare, Zhang Zhongmou, the founder of TSMC, stressed that when the TSMC corporate headquarters was renamed Zhang Zhongmou, the Sino-US trade war was a "no script" reality show that President Trump was performing. The United States and China will find a solution. It is not expected to be a long trade war.
It is worth mentioning that the proportion of mainland customer revenue in TSMC's total revenue has increased from 19% in the first quarter to 23%, which also reflects the rapid growth of the mainland semiconductor industry, and the dependence on TSMC continues to rise. Sino-US trade expansion and semiconductor projects will naturally impact TSMC operations.
At present, TSMC Nanjing 12-inch factory with an investment of NT$90 billion has been officially shipped in May this year. TSMC said that the Nanjing plant will give priority to filling 20,000 production capacity per month, and will not rule out further expansion. According to TSMC, there are no plans to list in mainland China, and there is no plan to set up factories in the US in the future.