In the wake of the coronavirus outbreak, many firms in the IT world have revised their business outlook for first-quarter 2020. But TSMC still sees significant demand for its foundry services, reportedly seeing its upcoming 5nm capacity fully book by clients. Its major handset chip clients, such as MediaTek, are all gearing up shipments for 5G applications. The handset market is still clearly transitioning to 5G, but overall handset shipments in China plunged in February, thanks to the epidemic.TSMC to kick off 5nm chip production in April: TSMC is set to kick off volume production of chips built using 5nm process technology in April, and has already seen the process capacity fully booked by clients, according to industry sources.MediaTek ramping up 5G chip shipments despite outbreak: MediaTek has started ramping up its 5G chip shipments, driving up its revenues for the first two months of 2020 by 25.11% on year to reach NT$30.43 billion (US$1.01 billion), with the sales performance much better than expected despite the coronavirus outbreak, according to industry sources.China handset shipments fall over 50% in February: Handset shipments in the China market fell 56% on year to 6.384 million units in February 2020, as the coronavirus outbreak halved demand for mobile devices, according to data compiled by China Academy of Information and Communications Technology (CAICT).

Astra, a Taiwan-based startup specialized in commercializing facial recognition technology, began to cooperate with Japanese telecom carrier NTT East to provide cloud computing-based image streaming and subscription-based facial recognition services for small- to medium-size enterprises in Japan three years ago, and the cooperative business has since accumlated over 1,000 monthly paid users, according to company founder and CEO Gary Kao.Facing increasing competition from mobile telecom carriers in the Japan market, NTT East teamed up with Astra to provide value-added services trying to maintain user basis and increase revenues, Kao said.Most of NTT East's over two million enterprise users are small- to medium ones, including retail stores and restaurants, and they are target users of the collaborative efforts, Kao said.As facial recognition services are expensive, Astra and NTT East provide such services on a subscription basis, with monthly rents being 1/200 of corresponding sales prices, Kao noted, adding that monthly payment coincides with telecom carriers' charging schemes and is affordable for target users.Astra has developed a MSP (managed services provider) platform specifically for NTT East's management for service operation, including users' devices and processing of data, Kao indicated. Via the MSP platform, users of the facial recognition services can get information on customers' age, sex, the numbers of their recurring visits, and volumes of visitors, Kao said.While facial recognition can help users learn more about their members and regular customers, most users hope to understand non-members and irregular customers for retargeting, and this is currently a bottleneck in application of facial recognition technology as well as a problem Astra strives to solve, Kao indicated.Facebook and Google provide data on clicking and visiting rates but are unable to let retail stores know who have watched online advertising about their business, or how many customers and sales the online advertising has brought, Kao noted.Astra compares customers detected via facial recognition with fan groups' user behavior and thereby finds highly loyal VIPs that visit stores and click "likes" in fan groups as well, Kao explained. For example, Astra had recognized visitors to a wine and spirits exhibition and compared them with the exhibition's fan group to screen out the overlapping persons, and later found they were frequent buyers with average annual spending of NT$200,000 (US$6,633) on wine & spirits each, Kao noted.Astra, unlike online marketing's influence on retail stores (offline), compares offline information with online data and hopes that the approach can help clients find additional customers, Kao indicated.Astra was runner-up in a startup contest held under CIAT (Cloud Computing & IoT Association Taiwan) Acceleration Program in November 2019.Astra founder and CEO Gary KaoPhoto: Company

Memory vendors have been cautious about increasing output in the wake of the coronavirus outbreak, despite significant demand from certain maket segments, such as datacenters and enterprise-class SSDs. It is foreseeable that memory prices will continue to climb in the second quarter of 2020. Foundry services are another sector that has been left almost unscathed by the outbreak. Lead-times at 8-inch fabs have extended because of tight capacity, as clients concerned about possible supply disruptions have stepped up order placements. Passive components makers have reported sales decreases for February because of output contraction at their China plants in the wake of the outbreak, but they remain optimistic with Yageo mulling raising prices to reflect increased costs. Memory contract prices to see double-digit increases in 2Q20: DRAM and NAND flash memory contract prices are expected to register double-digit increases in the second quarter of 2020, driven by growing demand for data centers, and enterprise-class SSDs and other applications, according to industry sources.Eight-inch foundry delivery lead time extended: Eight-inch wafer fabs' delivery times have extended to more than 12 weeks to reflect their tight capacity, according to sources at Taiwan-based analog IC firms. The lead times were below eight weeks before the Lunar New Year holiday, the sources said.Yageo, Walsin post significant revenue decreases for February: Passive components makers Yageo and Walsin Technology both posted sequential and annual falls for February 2020 revenues due to significant contraction in output of MLCCs and chip resistors at plants in China amid the coronavirus outbreak.

Taiwan-based startup JWI Intelligent Technology has developed an in-house-developed automated machine learning engine as an educational tool to help enterprises introduce machine learning as an initial step to realizing smart manufacturing, according to company president Jerry Huang.Many enterprises' digital transformation bids lack methods for implementation and experienced personnel for AI integration. JWI provides integrated consulting services to help clients build forecast models using their own operating data as the first mile of reaching AI-based digital transformation, Huang said.JWI's machine learning engine helps clients build quality forecast models, for example, for solder paste printing process in SMT (surface mount technology) and thereby recommending optimal printing parameters, Huang noted. If clients relocate factories or set up new ones abroad, they can apply the proven forecast models to the same equipment, Huang indicated.JWI's business model provides 10-day POC (proof of concept) services plus one-year licensed use of the machine learning engine on a subscription basis. The engine can be used to build various forecast models using the clients' equipment and operating data and based on their domain knowledge, Huang said.While the machine learning engine is to educate clients to build preliminary forecast models, JWI focuses on the educational purpose for the time being. After the engine has been widely adopted, JWI will consult with industry experts to build domain knowledge-based forecast models tailored for individual industries, Huang said.As clients may apply forecast models to different factories, management of different forecast models simultaneously poses a problem, Huang noted, adding JWI will seek to set up a related management platform based on experience gained from large clients' cases.JWI was winner at the 2019 startup contest sponsored by CIAT (Cloud Computing & IoT Association Taiwan) Acceleration Program in November 2019.JWI Intelligent Technology president Jerry Huang (right)Photo: Company

Apple's ban on its engineers from traveling to Asia in the wake of the coronavirus has reportedly affected the development of its next-generation iPhone devices that are supposed to be launched in fall this year. The travel restrictions has reportedly delayed the engineering verification test for the 5G iPhones at manufacturing facilities in China. Handset vendors may still be optimistic that 5G smartphones will become mainstream devices sooner or later, but 5G phone shipments will be much lower in first-half 2020 than originally expected, thanks to the outbreak.Travel restrictions reportedly delaying tests for next-generation iPhone: Apple reportedly has extended the deadline by one month to the end of April for the removal of a set of travel restrictions preventing its engineers from making business trips to Asia, a decision that could have a direct impact on the planned launch of 5G iPhone devices in the second half of 2020, according to industry sources.5G handset shipment boosts unlikely till 2H20: Shipments from the 5G handset supply chain for the first half of 2020 are expected to be one third of those for the second half, instead of 50% as estimated earlier, as first-quarter sales of 5G smartphones have been hard hit by the coronavirus outbreak and second-quarter shipment prospects remain unclear amid the worsening epidemic, according to industry sources.

Taiwan-based Hugreen has cooperated with Netherlands-based developers of greenhouses to make greenhouses smart through establishing sensing networks and monitoring systems, according to company founder and CEO Clement Lee.About 80% of Netherlands-based developers' overseas greenhouses have been left idle simply because Dutch experts assigned to these greenhouses are not willing to stay abroad for too long, Lee said. This is one of the problems Hugreen aims to solve, Lee noted.The Netherlands is a global leader in using greenhouses to grow crops, especially floriculture, Lee said. Netherlands-based greenhouse developers have advanced agricultural technology, proprietary crop varieties and global sales channels, but basically rely on manual labor to manage greenhouses, Lee noted, adding Hugreen's capability in setting up LoRa-based sensing networks and related monitoring systems exactly complements their strengths, Lee indicated.Sensors can monitor soil conditions, including moisture, temperature, acidity and fertilizers, with data collected transferred via LoRa for analysis to control optimal soil conditions, Lee explained. In addition to fixed sensors, mobile cameras are used to go around greenhouses to take photos for comparison with sensing data, Lee said. Hugreen has passed Netherlands-based greenhouse developers' testing of sensing accuracy and data transmission for resulting cooperation, Lee noted.Hugreen has 13 patents about monitoring and adjusting soil conditions, Lee said. While there are at least six methods of measuring soil moisture, Hugreen adopts permittivity to avoid influence by fertilizers and other factors, Lee noted, adding it has sent trainees to th eNetherlands to learn how to evaluate maximum effects from photosynthesis, estimate cost structure for greenhouses and other professional knowledge.Currently, Hugreen's clients are of two categories: one is greenhouse developers and irrigation project undertakers in more than 20 countries; and the other is small farmers and academic units in universities, Lee said.A 160-hectare greenhouse in Shanghai, China has adopted more than 1,600 sensors and 1,900 smart switches supplied by Hugreen, which has also helped it analyze optimal methods and conditions for growing 67 varieties of vegetable and fruit, Lee indicated.Hugreen's business model is to first provide hardware and consulting services for clients and later charge certain percentages of their sales revenues after successful operation of greenhouses, Lee said.Hugreen founder and CEO Clement Lee (front right) and colleaguesPhoto: Company

Supply chains had been keen to move their production lines out of China amid intensifying trade tensions between the world's two superpowers before the coronavirus hit. Now they have turned concervative about relocating their manufacuring as the fast spreading outbreak clouds the global economy. But passive components maker Yageo still plans to expand production capacity outside China to fulfill rush orders. For Apple, the epidemic reportedly is delaying production for an entry-level version of AirPods Pro that the vendor is believed to be launching.ODMs decelerate non-China production expansions as outbreak worsens: Supply chains have turned cautious about capacity expansion outside of China amid growing economic uncertainites in the wake of the widespeading coronavirus outbreak, according to the industry sources.Yageo boosting non-China production to fulfill rush orders: Passive components maker Yageo is gearing up to boost capacity utilization at its plants outside China to mitigate the coronavirus impact on its overall output of MLCCs and resistor chips and to fulfill rush orders from clients, according to industry sources.Production of entry-level AirPod Pro to resume momentum in 2Q20, says sources: Production of an entry-level version of Apple's AirPods device is expected to resume momentum by the end of first quarter or early in the second quarter of 2020, according to upstream component suppliers for TWS devices.Production of entry-level AirPod Pro to resume momentum in 2Q20, says sources: Production of an entry-level version of Apple's AirPods device is expected to resume momentum by the end of first quarter or early in the second quarter of 2020, according to upstream component suppliers for TWS devices.

Handset shipments by Taiwan's brand vendors and ODMs totaled 14.59 million units in the fourth quarter of 2019, up by a single-digit rate from the previous quarter but fell 35% from a year earlier, Digitimes Research has found.For all of 2019, Taiwan's handset shipments reached 56.16 million units, decreasing 32.2% from a year earlier.The Foxconn Group remained the top maker in the fourth quarter thanks to increased feature phone orders from HMD Global.Arima Communications surged to second place, buoyed by orders from clients in North America, while three- to fifth-place makers Asustek Computer, HTC and Compal Electronics all suffered sequential declines in the fourth quarter.Digitimes Research also found that smartphone models shipped by the five makers suffered an annual decline of 60% to account for less than 30% of Taiwan's total handset shipments in 2019.

Taiwan's National Health Insurance (NHI) program under Ministry of Health and Welfare is using big data it collects to train AI in developing smart medical care solutions.NHI Administration director general Lee Po-chang said his administration focuses on three core concepts in using the big data: data management, clinical application of AI technology, and protection of privacy of personal data.Analysis of medical records can result in fast ICD-10 (10th revision of International Statistical Classification of Diseases and Related Health Problems) coding of diseases to reduce hospital staffs' workload of coding diseases, Lee said.For interpretation of medical images, the more NHI images are used, the better AI algorithmic models are trained, Lee noted. For preventive medicine, medical images, such as CT and MRI ones, have to be further classified and combined with diagnostic content and, in addition, related data such as blood lipid, age and sex should be included in analysis, Lee indicated.The administration appreciates the medical value of minimally invasive surgery but, from the viewpoint of NHI insurance benefit, has to evaluate whether the advanced technology can bring higher medical effects than conventional treatments and the cost incurred, Lee explained.The NHI database consists of various medical data uploaded by about 30,000 hospitals and clinics in Taiwan. The data are first incorporated into a database of which underlying architecture is based on Oracle system. They are stored in 32-digit codes for de-identification with a string of keys, which are crucial to protecting privacy, because there is no telling the linkage between the medical data and corresponding personal identification.NHI data have so far been available for academic and research organizations mostly, Lee said. But the administration is communicating with life insurance companies about using NHI data to help them design insurance products that better meet customers' needs, Lee noted.Lee Po-chang, director general of Taiwan's National Health Insurance AdministrationPhoto: Shihmin Fu, Digitimes, March 2020

To build a successful startup, it is essential to start with a domain which the founders are best at and then work out solutions that exactly meet clients' needs, according to Winnie Lee, co-founder and chief operation officer of Appier, one of Taiwan's only two startup unicorns.Founded in 2012 as an AI startup, Appier has managed to raise total funds of US$162 million and has developed four major solutions, CrossX, AiQua, AiDeal and AiXon, to help clients in diverse sectors carry out precision marketing. It has been recognized by Taiwan's National Development Fund as one of the country's first two homegrown unicorns.Lee told Digitimes that at the start it had been difficult for Appier founders, including CEO CH Yu and CTO Joe Su, to convince international venture capital firms of the prospects of AI startups, and that it had been even harder to persuade enterprises to incorporate the firm's AI solutions to address their problems. Lee said after developing eight unsuccessful products, Appier rolled out an AI solution that precisely addressed clients' long-unsolved problems. She continued that the success has justified their insistence on starting in the AI domain where Yu is an expert.The three founders have shared the view that Appier should develop into a global company with AI as infrastructure while working out solutions for business sectors that can generate big data. This is why the company has rolled out precision marketing platforms for e-commerce, advertising and other related online services providers, according to Lee.Appier co-founder and COO Winnie LeePhoto: Michael Lee, Digitimes, March 2020