Taiwan-based startup Velodash has developed an app to allow cyclists to plan cycling routes and arrange group rides as well as to provide management system, including route planning and real-time tracking of cyclists,for cycling event organizers. It has recently raised NT$21 million (US$698,300) in an angel round of funding.Company co-founder and CEO Molly Huang talked to Digitimes in an interview about Velodash's outlook and plans.Q: The Singaporean government adopted Velodash's app for recording participants' performance in a cycling event at the Singapore National Games in late July-early August 2018. What's Velodash's progress since then? What is Velodash's next step?A: At the event, Velodash noticed the need to precisely meet clients' demand, as event organizers have to be responsible for participants' safety. But the app at the time could only display the locations of all participants but did not provide management functions for assigned tasks. Velodash, inspired by the observations at the event, has developed Staff, an app functioning as a smart control center for tracking all participants and displaying locations of rescue resources such as ambulances, and the statuses of incidents that are being handled. Staff will be launched later in 2020.By virtue of cycling events, the number of Velodash app users has increased from over 3,000 in August 2018 to more than 30,000 at present. Velodash has expanded the user base from B2C (individual users) initially to also include B2B (enterprises and event organizers).The fund raised in the angel round will be used to expand workforce from eight staff members currently to 15. Velodash will tap overseas markets in Asia, with Japan to be the first one. In addition, Velodash plans to further raise funds in mid-2021 or 2022 for taping additional overseas markets such as the US.Q: Are B2B clients mainly professional cycling events such as Tour de France?A: Professional cycling events usually adopt customized apps due to confidentiality. Velodash focuses B2B clients on organizers of amateur events or bicycle vendors who organize their own cycling teams for sales promotion.In fact, amateur cycling teams and people fond of cycling make up the largest demand. Members of amateur cycling teams, mostly aged above 35, usually go on cycling in group during weekends and they need mutual care.Q: Will Japan-based Nippon Platform Group, one of the angel investors, help Velodash tap markets in Japan and other Asian countries? Is it in strategic cooperation with Velodash?A: Nippon Platform Group has not asked for strategic alliance despite its angel investment, but will maintain cooperation with Velodash.Q: What are the business opportunities in the Japan market? Are there competing products?A: Japan is the largest market in Asia. In view of concentration of international tourists in main cities, the Japanese government has been making efforts to divert tourists from big cities to rural areas. However, public bus services in rural areas are very infrequent and thus bicycles become an ideal transportation tool. Local governments in Japan are also inclined to sponsor cycling activities or events in rural areas. Therefore, Velodash's sales promotion in Japan will begin in rural areas rather than in big cities. Besides, as there are various folk festivities, such as carrying of sedan chairs, on many traditional festivals in Japan, Velodash plans to provide a common platform for activity organizers to create content on demand.Q: In addition to GPS, AI-based algorithm and IoT, what other technologies will Velodash adopt?A: Besides collection of data on location, Velodash plans to use GPS to establish its own map system, tentatively dubbed VeloMap. Because Google Map contains much unnecessary information for cyclists, our map system will focus on information really needed by cyclists such as locations of coffee shops for taking rests, repair stores and restrooms, and allow users to edit and update information. Velodash will encourage store operators to join the map system.Q: How can Velodash maintain users' stickiness?A: Velodash encourages users to create unique content. For example, Velodash-developed route planner enables cyclists to recommend their favorite cycling routes for sharing.Velodash co-founder and CEO Molly HuangPhoto: Shihmin Fu, Digitimes, February 2020

UMC, once a major rival to TSMC in pursuing advanced manufacturing nodes, decided about two years ago to shift its focus away from joining the race to 10nm and more advanced process technologies. Now the pure-play foundry expects its renewed focus to start bearing fruit in 2020. Following two years of corporate adjustments, UMC has set new goals to be accomplished over the next five years, according to UMC co-president Jason Wang.Wang and another UMC co-president SC Chien have been the architects of the company's transformation. But their roles are clearly defined: Chien is responsible for UMC's core manufacturing operations and technology including R&D, while Wang focuses on corporate strategy and planning, sales and marketing, and customer engineering. Digitimes had an opportunity recently to sit down with the two UMC co-presidents to talk about the foundry's investment strategy, and where it is heading over the next five years, including plans for China.Q: UMC disclosed previously plans to enhance its 14nm and 12nm process offerings but to suspend sub-12nm process R&D. Are you still confident with the decision you made two years ago?Wang: UMC has shifted its focus away from keeping up in the process-technology race and re-positioned itself to a specialty logic foundry. UMC remains confident about its strategy shift.Two years ago, UMC decided to be less aggressive in the process-technology race and to magnify its value in specific foundry segments. We have well-established technology and customer portfolios and manufacturing capacity, and intend to utilize what we have to bring more benefits to the company's shareholders and also for sustainable business development.Q: I believe that was a bold decision you made. Would you elaborate under what circumstances UMC decided to shift its focus?Wang: I have been with UMC for 10 years. UMC has been keen on investing in R&D. For the foundry sector, there are two types of investments - one for technology development and the other for manufacturing capacity. The latter requires much more investment.My finance background has encouraged me to help UMC pursue investment efficiency. For UMC, which generates about US$5 billion in annual revenues, investment in technology R&D is affordable. However, investment in manufacturing capacity calls for more conservative planning.Therefore, it was quite a logical decision for us to make. Besides, UMC has been cautious about capital spending since 2017. Our annual capex has been maintained at US$700 million.As for R&D expenses, UMC spends about 9% of revenue on R&D compared with 6-8% TSMC allocates. But the actual sums differ much because of the sizes of the revenues.Chien: UMC encountered some bottlenecks in the 0.13-micron process race. It was a vicious circle in which we lost market share in the advanced node market segment and saw impacts on our revenues, coupled with a decline in our available R&D capital. Such experience pushed UMC to rethink its strategy to avoid being trapped again in a rat race.In 2017, UMC recognized its role could make a difference. Rather than fighting to be a technology leader in the advanced-node process segment, UMC can be more capable of being a leader in the more mature process segments.Besides, in the 14nm and older process segment, there's still room for growth. In particular, demand for specialty process manufacturing has been rising. We have a great R&D team engaged in the development of more new specialty node offerings for this growing market.UMC used to have only 30% of revenues generated from specialty process technology. The proportion has now surpassed 50%, demonstrating our strategy shift in the last two years.The specialty process market segment will continue to expand, driven by growing demand for chips such as display driver ICs and microcontrollers. UMC has also seen robust demand for 12-inch fab capacity using 65nm and 90nm process technology.Q: If this decision - your increased focus on specialty node offerings - had been made earlier like five years ago, would UMC have transformed better?Chien: This wasn't a decision we could have made lightly. We make decisions based on what we are able to see ahead.Wang: For UMC, stepping into 12-inch wafer manufacturing was a must despite difficulties it faced. UMC's 0.13-micron technology bottleneck already delayed its entry into the 12-inch foundry space for three years. Nevertheless, UMC at that time was capable of pouring resources into 12-inch wafer manufacturing. We were once ahead of our rivals in the 90nm process segment.At that time, it was rather impossible for a pure-play foundry to quit the process-technology race. However, the arrival of the smartphone era has reshaped the industry with growing dominance by a few suppliers. We need a different strategy for our next stage of growth.There were a number of factors we had considered behind the decision we made in 2017. As UMC steps into the FinFET segment, we intend to put our capex focus on technology R&D rather than manufacturing capacity. In fact, UMC's investment in 28nm manufacturing capacity has still been a burden for the company. As we enter the era of FinFET, we want to slow down the pace of expansion and pursue our growth in mature and specialty market segments for profitability.Q: Did you take other success cases as reference in formulating the transformation policy?Chien: Texas Instruments (TI) is a very good reference case for us. In earlier times, the US firm used to develop DRAM, DSP (digital signal processor) and smartphone chips, but it later terminated development for them when the time was right. Now TI has become a firm leader in analog technology after continuously accumulating experiences and fully implementing its expertise. Infineon serves as another case with the same development pattern: it is now also leading in some strategic domains. UMC operated as an IDM in the 1980s, but later decided to become a pure-play foundry house, unloading all internal product lines. This was a major decision.Q: Under what conditions would UMC make massive investments again?Wang: We will remain serious about investment effectiveness. Judging from its existing operating scale, UMC have resources and opportunities, as its cash flow has increased by over 3-fold since 2017. Making massive investments is not a problem for UMC, but we care more about investment benefits.We have just fully acquired Japan's Mie Fujitsu Semiconductor (MIFS), renamed United Semiconductor Japan (USJC) on October 1, 2019, for US$600 million, bringing us a monthly foundry capacity of over 30,000 12-inch wafers. It's good investment, given that the capacity, if built on our own, would involve a construction cost more than five times the acquisition cost.USJC mainly engages in 90, 65, and 45nm manufacturing nodes and the acquisition will boost UMC's global market share to 22% in specialty processes, contributing to our goal of raising market shares. The deal may not create major investment benefits in the short term, but the effectiveness will gradually emerge over the long term, also in line with our focus on developing specialty process nodes and generating added-value for our capacity.Chien: We did evaluate the possible synergy of the acquisition in advance. With its larger operating scale, UMC can easily leverage its smart production experiences to help USJC improve production flow, shorten output cycle and automate quality inspection, with investment effectiveness to emerge in three quarters. Meanwhile, USJC's solid clientele in Japan can help UMC expand its Japanese market as customers there will hope to better utilize USJC's fabs, now probably boasting the most advanced process nodes available in the country, to fabricate chip solutions for automotive and military applications.Q: Would you talk about UMC's global deployment plans, as well as the cancellation of public listing plan for Hejian Technology Suzhou (HJTC) and UMC's suspension of cooperation with Fujian Jinhua Integrated Circuit (FJIC)?Chien: China is no longer the factory of the world and supply chain relocation is emerging as a growing trend, with the US and many other countries hoping to have advanced fabs set up locally. For example, many clients have asked to have their products made in Singapore to diversify production risks, and yet Chinese clients also hope to have their products fabricated by UMC's subsidiaries in China - HJTC and United Semiconductor (Xiamen). Accordingly, global deployments must cater to the needs of clients while also meeting different countries' specific needs for industry support.Our future deployments will continue to be based on evaluations of investment effectiveness, and we will also evaluate the benefits of M&A. For the moment, UMC has invested over NT$300 billion in southern Taiwan, with all its R&D teams rooted in Taiwan. Overseas investments have all met government regulations and gained government approvals.Wang: As HJTC has registered profitable operations in China for years, and capacity expansion will also be needed at United Semi (Xiamen), we initially hoped to raise funds through the China capital market to support capacity expansions at both firms. But finally we decided to withdraw the application for listing HJTC shares on the Technology Innovation Board of Shanghai Stock Exchange after securities brokerages and securities regulatory authorities there failed to reach agreements over the listing terms. For UMC, it would be better to get capital support in China but the company can also invest with its own funds. Whether or not HJTC can list its shares will not affect UMC's long-term operation plans in China.HJTC's 8-inch fab has a monthly capacity of 77,000 wafers, and United Semi's monthly capacity will increase to 25,000 pieces in the first-stage expansion from the present 17,000.As for FJIC, UMC holds no stake in it; the Chinese memory maker was only a client for whom we helped develop technology on a project basis. But the development has been suspended following a lawsuit filed by Micron Technology against FJIC and UMC. Now that the case has entered the judicial process, we will do our best to safeguard our rights and clarify unjustifiable allegations against us.Chien: The Micron lawsuit involves many false accusations and we can only vindicate ourselves in court. The lawsuit has had no significant impact on UMC, as our clients, though a little astonished initially, have maintained normal business exchanges with us, trusting that we offer perfect protection of their trade secrets.Q: Both of you have served as co-presidents at UMC for more than two years. How do you coordinate with each other in management concepts and practices?Chien: The co-president system enables a very good division of responsibilities, allowing us to support and supplement each other. I am in charge of technology R&D and operations, and Wang responsible for market planning and business promotion. We consult with each other about how to get our company on the right track for further development and growth. There have been no quarrels between us as we always base our discussions on facts for better communication.Wang: The dual-head system has been meant for complementarity instead of competition. Our decision-making is 100% transparent, based on figures, rather than subjective ideas. I think this is a very pleasant cooperation model.We both have a "stay hungry, stay foolish" mentality. Respecting each other's professionalism, we usually collect data and analyze them rationally before reaching consensuses and making final decisions. Actually, the entire process is virtually the same as big data analysis done with digital tools.Q: UMC's co-president system will enter its third year of operation in 2020. Would you both assess your own achievements and performances so far ?Chien: Our capacity utilization rates have hit over 90% for two consecutive years, but heavy equipment deprecation burden resulting from past mega investments is expected to persist for one more year before starting to ease. Nevertheless, we have performed well in boosting technology competitiveness. We have completed development of 14nm process and remain competitive in 22-28nm nodes. We also see clear growth trends for our product lines in 2020-2021.Additionally, we have outrivaled peers in specialty segments, capturing the highest market shares in LCD driver IC and TDDI IC fabrication in the past two years. We have also started volume production of OLED driver ICs using 40 and 28nm nodes in 2019 and hope to gain a presence in the AMOLED market in South Korea or China in 2020. Our market share in the RF-SOI (radio frequency-silicon on insulator) segment has also reached 15%, and our shipments of automotive MCUs, already validated by clients, are set to grow steadily. But it is a bit pity that these achievements cannot be immediately shown in terms of financial numbers.Wang: I think UMC's financial performance is acceptable, but we are not winning the applause we deserve. I'm trying to win that applause for us, at least securing a reasonable place in the industry. In terms of financial performances, commitments from employees, and recognition by clients and shareholders, we've achieved our goals set two years ago. The management team has reached a consensus over future development direction and will continue to present impressive results in diverse domains.UMC's transformation efforts will start to bear fruit in the next two years. For UMC with 20,000 employees and over 400 clients, it is not an easy job to enforce major operational changes. But UMC has proved that it has the resources and capability to handle the job well, which, however, still needs a little more time to finish.UMC co-presidents Jason Wang (left) and SC ChienPhoto: Shihmin Fu, Digitimes, November 2019

Silicon Valley-based Kiwi Campus is a startup dedicated to operating a semi-autonomous robotic platform for delivering foods, beverages and groceries to consumers on college campuses and in neighboring communities. The company, attracted by Taiwan's sound business climate, robust hardware manufacturing, and R&D prowess, is planning to set up operations in the country, according to company founder Felipe Chavez Cortes.He was recently in Taipei to assess the feasibility of establishing an R&D center, a robot assembly and test team and a business operation unit in Taiwan. In an interview by Digitimes, he talked about his firm's service platform and investment roadmap in Taiwan.Q: What drove you to set up Kiwi Campus?A: While still a student at the University of Los Andes in Columbia I got the idea of developing campus economy, believing that time was ripe for operating an automatic robot food delivery platform. We founded Kiwi Campus after spending one month studying the sharing economy, supply chain and AI-enabled autonomous driving technology.We later joined the Skydeck Acceleration Program hosted by the University of California, Berkeley, which became our first institutional angel investor and technology consultant. This also gave us access to resources from California State University through the Deep Drive research project, enabling us to get funds and AI technology needed to build robots.About two and a half years ago, we moved and officially registered our company in Silicon Valley.Q: What is your business model?A: We want to create a balanced market economy, allowing users to happily pay US$1 or lower to get delivery service for foods and merchandises they need and available within two square miles via semi-autonomous robots.Our revenues mainly come from restaurants or suppliers of products for sales in automatic vending machines. We charge 15% of product price as service fee.For the moment, we offer services at two universities and their neighboring communities. One is UC, Berkeley, with its nearby stores and consumers contributing 70% of our revenues there. Another is the University of Denver in Colorado.Meanwhile, our colleagues are making preparations for launching services at another 12 university campuses. We hope that our company can start operations in Taiwan in 2020.Q: What does Taiwan interest you most?A: As our robots are now manufactured in Shenzhen, China, we want to find a university campus nearer the hardware manufacturing base. Also, we need to find a place in Asia to set up a team with strong R&D momentum, and this is why I am in Taipei, where I hope to meet with potential talent and see if there is a chance of establishing an R&D team. I believe that with its high population density and strong consuming power, Taiwan will be a market highly fit for the services we offer.As the TensorFlow platform we use is developed by Google and China bans the use of Google map, we have no other choice but to do the robot assembly and test in the US, which is quite inconvenient for us. This has prompted us to seek a proper place near Shenzhen, where our robotic components are manufactured, to assemble them into robots and do final test. I think robots can be well assembled and tested in Taiwan and then directly offer services to local consumers, given the country's strong ICT manufacturing capability and competitiveness.Q: Is National Taiwan University (NTU) the campus you just mentioned? How will you work with it?A: Besides NTU, we will explore the possibility of cooperating with all other universities in Taiwan. We have amassed abundant experiences in successfully cooperating with university students in the US to deliver foods, beverages and groceries via robots to consumers on campuses and in neighboring communities, and such a cooperation model will also be applied here.Q: How has business been since your company started official run two years ago?A: We completed over 70,000 transactions in 2018 and now register a weekly average of 3,000 deals. If you visit UC, Berkeley, you will find our robots shuttling around. Population density and cooperation with the supply chain are crucial factors for our business. Based on our calculation, one-trip delivery cost for one robot can be covered by fulfilling two service orders in Taipei, and one automated robot can handle 4-5 orders per trip.Q: How is AI applied to your services?A: Our robots adopt Level 2 autonomous driving technology and our artificial neural network enables the robots to automatically move in the middle of sidewalks, shun obstacles and recognize images, street corners and traffic lights. Each robot is equipped with six camera lenses and a Jetson TX2 computer with GPU computing capability, but at low power consumption. We use machine learning and deep learning to train our robots, enabling them to adjust moving speeds and directions in accordance with road conditions, yet without using high-end sensing or LiDAR devices.Q: Will you use the robots to collect big data for future commercial applications?A: Our neural networks will not record road images but can collect data concerning traffic jams and time slots, as well as networking quality. But it will require many more robots if we want to collect sufficient data for useful applications. In this aspect, we will move to discuss with city governments on possible cooperation.Q: What about your business development plan for the next 5-10 years?A: We are mulling setting up robot-operated smart city systems, allowing people and enterprises to access our robot platforms to live in smart cities, which will be our ultimate goal. Actually, we have started building vertical high-frequency trading marts or platforms by combining robots and cities.Q: How about your company's current market valuation ? Will you have a new round of fundraising activity?A: The current market valuation of our company is around US$4 million, and we are proceeding with a second-round financing plan seeking to raise US$15 million in the first or second quarter of 2020. Meanwhile, we hope to kick off our services in eight cities in 2020, with one in Asia, hopefully Taipei. And our business goal for next year is to complete one million transactions with 1,000 robots. Kiwi Campus founder Felipe Chavez CortesPhoto: Michael Lee, Digitimes, November 2019

Taiwan cloud software startup SoftChef has been devoted to developing comprehensive solutions from edge devices to infrastructures since its inception in 2017, providing customized IoT solutions to better serve clients, according company founder and CEO Josh Chai.In a recent interview conducted by Digitimes, Chai highlighted the importance for enterprises to build a "terminal" concept and incorporate intelligence into end devices. He also stressed that future business model must focus more on hardware connections to better learn and serve real needs of consumers.Q: What drove you to set up the cloud software venture?A: After serving networking specialist D-Link from 1999-2013, I left to set up a B2B e-commerce platform GCR in cooperation with the firm's ex-CEO Tony Tsao as I thought the pure hardware production industry could hardly advance farther. But one year later, I returned the GCR management right to Tsao and joined InfinitesSoft, a developer of hybrid/multi-cloud management software, to learn more about cloud technology.At InfinitesSoft, I found great business potentials in the Asian cloud market, especially in Southeast Asia and China, prompting me to establish SoftChef to develop cloud-based IoT platforms for IT hardware makers.Q: After venturing into the IoT management platform domain, did you see any difference between the actual market demand and what you envisioned initially?A: Initially I adopted the serverless application framework of Amazon Web Services (AWS) seeking to create a general-purpose IoT management platform allowing enterprise clients to manage IoT devices, collect and store data anytime.But later, we found IoT is a highly customized domain. Actually, an ideal IoT must be like an "oven" allowing enterprises to put in their own materials and process them into distinct "business breads" they need, as so-called standardization and automation can hardly be actualized in IoT vertical applications. In other words, clients require a versatile and agile oven rather than a vending machine selling standardized services.Q: How do you assess Internet of Everything that many think may materialize after 5G commercialization?A: In the IoT era, it doesn't mean that all devices should be connected to the Internet, as not all the data collected can generate values and meaningless connections are just a waste of resources. In this regard, a "terminal" concept should be established, meaning that data can undergo edge processing at the local end allowing only valuable data to be stored at the cloud end. In short, any IoT end device should be given timely intelligence.Q: What kind of value do you think IoT connection can generate for enterprises?A: If enterprises know little about the purpose of such connections and what data they want, IoT connection can hardly generate useful value for them and adopting even the cheapest NB-IoT transmission system will be not cost-effective for them.IoT is not a technology but a process of transforming digital economy into infonomics and data economy, with production-oriented business model replaced by service-oriented one highlighting sales of services not merchandizes.In the future, enterprises have to reply on hardware devices connection to better learn what consumers need so as to offer services that can really cater to their needs and further make them willing to pay for the services at whatever costs.SoftChef founder and CEO Josh ChaiPhoto: Vicky Liu, Digitimes, October 2019

Both 5G and Wi-Fi 6 networks provide advancements in performance to enhance mobility, capacity and data rates. Leveraging artificial intelligence (AI) and Internet of Things (IoT) technologies, computing is becoming more powerful, enhanced by wider bandwidth, high speed, low latency and massive connectivity. Servers are almost everywhere, but things not just happen in the data center, but also remotely.The booming business opportunities of high performance computing and edge computing improve user experience in every specific use cases. With strong growth and increased competition in the data center sectors of North America and Europe, the trend is expanding to Asia region to deliver quick and massive growth.The latest strong growth of Lenovo's data center group is a good example. In an interview, Han Chon, general manager of Lenovo Data Center Group Central Asia Pacific region, talked about the strategy for achieving growth in the data center market. He is responsible for leading sales and marketing across 12 nations and regions including Taiwan and Hong Kong in this area. Having achieved significant growth and industry leadership in 2018, he expressed gratitude to his entire team for making great efforts.Embracing AI technologies reducing human resource constraintsThe customers of the Central Asia Pacific region are showing strong intention to welcome and embrace new technologies, Chon said. The high performance computing and AI technology have a lot to benefit from each other. As the more data generated by various mobile devices, AI is helping us make sense of the data. The Big Data Analytics is becoming the driving force to create new or enhanced business models more closely attuned to the needs of end customers. AI makes enterprises to increase efficiency and spend less on administrative tasks.Lenovo Data Center Group (DCG) looks at this trend and very openly to cooperative with different AI technology partners to maximize the opportunities presented by the digital economy. Meanwhile, DCG keeps investing more resource to work closely with educational organizations, research institutes, global enterprises and big electronics and semiconductor companies. The collaboration helps to introduce new solutions to tackle major issues that enterprises face.More fast-growing companies that leverage AI technology for their core business create an immense business benefit through data center applications. However, for performance reasons, the remote data-generating devices must be close to computing and storage resources. A slimmer data center, or called a mini version of a data center server, could work as edge computing node takes hold in various industries. The edge computing trend is gaining ground with good reasons.Crafted edge server series accelerating business opportunitiesOver the past few years, Chon watched there has been a steady increase in the number of mobile devices provided high resolution and quality video streams easily. AI technologies make edge computing to deliver reliably, speedily while taking more care of data security. The AI-enabled edge servers can better manage their resources and clearly understand the merits. This is why Lenovo DCG introduced the first edge computing server series for developing the business in the beginning of 2019. The Lenovo ThinkSystem SE350 Edge server equips with small footprint and power efficiency allowing for reliable server-class performance at many remote locations.The SE350 can handle temperatures from 0 degree to 55 degree C, as well as tolerate locations with high-dust and vibration, such as construction site trailers and manufacturing floors. It can be deployed equally well in a traditional office or branch location due to its office-friendly acoustics. The essential cyber-security offerings include key-encrypted storage and physical security features, such as a locking bezel, as well as intrusion and tamper-detection mechanisms. And the networking connectivity options equip wired 10/100Mb/1GbE, 1GbE SFP, and 10GbE SFP+, as well as secure wireless Wi-Fi, cellular LTE and 5G connections in the future.For decades, computing was done solely on big servers in the central controlled room of companies. Over the last two decades, however, businesses gradually began shifting their workloads to the cloud and helped companies reduce capital expenditure and increase return on investment. But the edge servers are terms that are well understood by companies today, even we could see the edge computing approach making data center everywhere, Chon describes to support the ideas.Building a Thriving Ecosystems in Taiwan for further developing marketOwing to the growing presence of hyper-scale cloud service providers, extensive penetration of Internet connectivity as well as keen focus on data security, the demand for data storage and cloud services is expected to grow exponentially across the ASEAN and Taiwan region. Lenovo DCG works closely with customers to deliver value to Taiwan industries and continue to expand the products portfolios to make customers to grow faster.Chon highlights Taiwanese electronics manufacturers are on the frontline of cutting-edge technology to deliver premium products to global brands. Taiwan manufacturers are also investing AI-enabled data center solutions to help boost growth and innovation. Upcoming data center markets are poised to experience exponential growth due to abundant availability of resources as well as government policies supporting. That shows a natural growth of Taiwan data center business.For fast responding the market demands, Lenovo DCG is creating the collaborative business ecosystems to accelerate innovation. "Because no one can do it all and do it alone, the partnership of ecosystem partners or channel partners will be the key strategy to grow regional business," Chon says. Now that more industries are becoming digitally savvy, the Lenovo DCG partners will play an important role to navigate digital disruption and become more agile and responsive to customer's needs. "Without their help we could not go to the level we today's position", Chon says.Although Lenovo DCG is a very young organization for only 2.5 years, the business scope and team members are both growing fast. "All Customers are chasing the solutions to break the limits of computation and push the boundary every day," Chon adds. They want to do better and more. Things like to have better management of their fleets, or simply such as growing new business and improving efficiency.For Taiwan ecosystem partners, "Lenovo DCG teams are here to stay and to play," Chon continues very positively, "Today we extend the high-performance architecture from data center to the edge with the complete product families, becoming valuable partners and bring tomorrow today for Taiwan customers." With these high-performance products and ecosystem partners, Lenovo DCG is ushering in a new age for data center solutions.Han Chon, regional general manager, Lenovo DCG Central Asia Pacific

Supermicro, a first-tier server vendor worldwide, has been aggressively developing gaming-related products with its experience from the server business. Supermicro is able to offer customers products with server quality.Digitimes recently talked with Supermicro's senior application system product manager Steve Lee and senior product manager Tony Fan about the company's gaming business.Q: How long has Supermicro been in the gaming motherboard business?Lee: We have been developing gaming motherboards for five years. In fact, Supermicro has been in the motherboard business for around 20 years already, but has mostly focused on consumer, server and embedded solutions.Recently, with gaming applications growing popular, Supermicro also crossed into the development of the related product lines. However, in addition to customers from the PC DIY market, we have seen growing demand from the enterprise sector as customers from the field are building desktops using Intel's top-end CPUs and need reliable motherboards to unleash the CPUs' full potential.All our gaming motherboards are equipped with server-grade components and designs to optimize the stability of the systems, making our gaming motherboards the top pick for our customers.We currently have several business models for selling our gaming motherboards including cooperating with system integrators, specific regional channel retailers and e-commerce platforms.In Taiwan, we work with a local channel retailer CoolPC to sell gaming and server motherboards, while in the US our motherboards can be found on both Amazon and Newegg. However, partnering with system integrators is still the main business direction for Supermicro's gaming business.Q: What direction is Supermicro planning to push its development to?Lee: We have seen many brand vendors in the gaming market shifting their development to features that make products more appealing to customers, such as adding support of RGB lights to hardware and peripherals.This is a growing trend in the gaming market, but Supermicro at the moment will still focus on maintaining products' high quality, performance and reliability, as these are what Supermicro's customers are mainly looking for.Q: Does Supermicro manufacture its gaming desktop in house?Lee: Supermicro is collaborating with peripheral makers including Thermaltake and Lian Li for desktop and chassis product lines.Supermicro has co-developed two liquid-cooling gaming desktops with Thermaltake that are currently available in the US channels. The company's chassis co-designed with Lian Li was showcased at Computex 2019.Q: Does Supermicro's gaming products have any advantages over competitors?Fang: Supermicro is designing its gaming motherboards with a concept 100% identical to designing server motherboards. Our gaming motherboards are all equipped with server-grade components that are provided by a sole supplier in order to keep quality stable.To make every gaming motherboard with the same quality is the goal that Supermicro has been looking to achieve and by choosing only one supplier, the company is looking to avoid the concern on the tolerance discrepancy of the components.For example, Supermicro's gaming motherboards use high-quality PCBs with extra layers, which are the same as those used in server motherboards. Although the such motherboards are pricier than regular ones with fewer layers, they provide better and more reliable electrons transmission via circuit.In addition, several of our gaming motherboards are also equipped with server-grade remote management chip from ASpeed Technology and Supermicro has a team of 300 software engineers to help customers manage their systems via the Intelligent Platform Management Interface (IPMI) feature. Customers can also obtain assistance via their smartphones.Thanks to Supermicro's expertise in server motherboard design, some of Supermicro's gaming motherboards can support up to a total of four graphics cards with the addition of a PCIe switches chip to expand the motherboard's Lane support.Supermicro senior application system product manager Steve Lee (left) and senior product manager Tony Fan (right)Photo: Joseph Tsai, Digitimes, July 2019

AR can create brand-new gaming experience, but for industry applications, AR can help firms tackle their long-existing pain points, according to Roger Lu, founder and CEO of Mind & Idea Fly (Mifly), a Taiwan startup dedicated to multimedia AR/VR content development solutions.In an interview conducted by Digitimes, Lu stressed that real-time transmission will play a key role in determining the success in the AR domain in the 5G era, and that solid hardware support also counts greatly for AR applications in people's daily life.Q: Why did you migrate to the industry AR segment after 10 years in the gaming segment? What is the difference between gaming and industry AR?A: I witnessed the changes of the gaming sector during my 10 years in the sector. Mobile games started to emerge after Apple released its first 3G iPhone in 2008, but not until 2013 did large capital funds start to enter the sector, leading to the birth of large-size gaming enterprises such as Tencent and NetEase in China.The leading firms tended to acquire work teams and IPs of smaller firms that develolped gaming products with great sales potential. As a result, many independent gaming studios were acquired or failed.I found that AR can allow interactions whether it is applied to games or general industries. But for gamers, AR interactions can mainly bring them new gaming experiences, such as Pokémon GO. For industries, AR can help tackle their long-existing pain points. E-commerce operators, for instance, can use AR to develop visual apps for consumers to try shoes on via their smartphones.Q: How do you see B2C and B2B AR applications? What business model is adopted by your own platform?A: For AR glasses, B2B is the dominant business model, with such glasses having four main application scenarios including repair workshops, production plants, on-site inspection and security operation. I think future B2B AR applications will move closer toward large-size businesses and government units.As to the B2C business model, it can make an AR platform grow quickly, but the ratio of subscription payment by consumers will not necessarily rise with the platform usage growth.Our MakAR editing tool, designed to help enterprises create AR content to serve consumers, adopts the B2B2C model, meaning that businesses pay the cost for the services to consumers.Q: How will 5G affect the development of VR and AR application to industries?A: 5G will certainly have immediate influence on VR, as high-quality games will no longer be downloaded and can be directly played in the cloud, and high transmission performance of 5G can concretely match quality VR content.The influence of 5G on AR is not so clear for the moment. But when AR identification technology SLAM (simultaneous localization and mapping) is applied to process bigger data in shorter time before generating application content, 5G real-time transmission will be critical to supporting the application.Q: Besides the support of 5G transmission capability, what other crucial aspects will spur the development of the AR industry?A: In order for AR applications to be materialized in our daily lives, the popularity of hardware AR glasses also counts greatly. Though AR applications are operated via handsets, the most intuitive interface will remain AR glasses, which can also better integrate the virtual and physical realities.Mifly founder and CEO Roger Lu (center)Photo: Viki Liu, Digitimes, July 2019

ARPlanet Digital Technology, founded in 2010 as an AR startup, provides a full range of virtual and physical realities integration solutions. The company is now one of few startups in Taiwan able to export its AR PaaS (Platform as a Service) solutions.In a recent interview conducted by Digitimes, the firm's founder and CEO Jennifer Pai said AR is the friendliest and most intuitive interface for humans to access global information. She expressed optimism that AR will become a pivotal next-generation technology capable of integrating virtual and real worlds.Q: You used to be a Chinese-language magazine editor and were also responsible for marketing in gaming, financial and retail service sectors. How did these job experiences help you set up an AR startup?A: I majored in Chinese literature at university, but I was then quite interested in learning new tech products in addition to taking courses in journalism, psychology and economics. After graduation, my first job was to do copyediting at a Chinese-language computer magazine, and at the time new technologies were fast changing work patterns.Just three months after joining the magazine, computerized typesetting desktop systems quickly emerged and prevailed, prompting art editors to quickly learn and master such systems, with many manual typesetting service houses forced to shut down. From this, I got to know that any new trend would appear extremely fast and wait for nobody.Later I migrated to do marketing for diverse business sectors, witnessing the evolution of the Internet from web 1.0 and 2.0 to the Internet of Things (IoT) as well as the impacts of digital technology on various industries. When I first learned of AR more than 10 years ago, I deeply felt that "vision" would be the next trend and AR would play a crucial role in the trend.The outlook is bright for AR because AR is an all-match, able to be paired with diverse business models and suitable for cross-domain applications such as transportation, tourism, smart city, smart manufacturing, 5G, IoT, medical care, and long-term care. While smartphones are indispensable devices in the daily lives, I think AR will be not only a technology in the future, but also an interface for us to access the world in our daily lives.AR can be used to explore many things that are invisible to the naked eye and present them via computer vision. All the knowledge and data now stored in hard disks and in the cloud have to be presented externally through computer vision in the future, so that virtual and physical realities can be combined to demonstrate the true value of AR.Q: Besides the core AR solutions, your company also provides VR services. How do you define the market positions for AR and VR?A: In terms of software, VR is actually a carrier and device, and AR is a kind of user interface (UI) and user experience (UX), enabling us to access information.Both AR and VR are experience technology, but only through the integration of virtual and physical realities can the true value be generated. VR is purely a virtual concept and can help you enter the virtualized world after wearing a VR carrier, but physical reality cannot be totally replaced by VR.The reason AR is so fascinating is that it does not separate you from the real world, and can instead combine the real life to create new business models by leveraging technology. AR is not a world formed with pieces put together, but a new user-oriented application based on the development of humanity.For instance, if AR is applied to serve as a shopping guide for you, it can create a special shopping atmosphere allowing you to purchase the exact products you want to buy without quitting the real environment. With AR, consumers will no longer have to make online price comparisons or search for product items via written text, and can complete shopping simply by scanning products with their handsets. Accordingly, AR enables consumers to enjoy a slight sense of entertainment in the shopping process.For commodities vendors, AR-based shopping guide will generate high-accuracy one-to-one shopping pattern and help enhance brand loyalty by consumers.Q: How can VR achieve a booming development?A: Content will play a key role in determining the development of the VR industry. Currently, only a few people have their own VR carriers, and most others still have to go to specific dens to experience VR. Accordingly, the VR industry still focuses on vertical market applications such as education and video games, allowing users to watch immersive images or images invisible to the naked eye. VR cannot enter an actual popularity period until after VR carriers become highly affordable pricewise so that most consumers are willing to buy them before mature edge and cloud computing technologies are available to support diverse VR applications.At the moment, the average VR carrier unit price is still over NT$30,000 (US$966.40), which is still too high for consumers to accept. VR industry players still have little idea about the aceptable price levels of VR hardware in the market, but they can take current market prices of Nintendo Switch or Sony PlayStation consoles for reference, which may be the price range acceptable to users.But what counts most for the development of the VR industry is still content or user experience. Sony owns massive gaming content and users, and can therefore develop VR technologies in its gaming ecosystem. Without content and users, it would be a highly difficult job to build VR ecosystems.Q: What changes will AR and VR bring to Taiwan's 5G industry? And what government policy is needed to help spur the development of the AR industry?A: The two major advantages of high efficiency and low latency of 5G cannot be fully appreciated without the stimulus of AR, VR and even 8K, which are therefore badly needed to persuade consumers to upgrade to 5G from 4G, which currently does not have any problem accessing the Internet.AR, VR and 8K are crucial for promoting 5G because 5G is best suitable for vertical applications carried out at places shared by many people, such as a sports stadium, where AR, VR and 8K are mainstream applications. AR and VR enable high real-time interactivity, and current 4G smartphones cannot support 8K image transmission.Accordingly, while enforcing 5G development policies, the government must also spur the development of AR and VR content in addition to hardware equipment and technologies, in order for consumers to enjoy real 5G experience.In the wake of "Pokemon Go" becoming a popular AR application, many hardware makers have come to us for cooperation. Nevertheless, related Taiwan businesses usually lack software and hardware integration capability, and the government can serve as the best third party to bridge and integrate the strong ICT hardware industry and the fledgling software industry in Taiwan to facilitate the country's AR industry development.ARPlanet Digital Technology founder and CEO Jennifer PaiPhoto: Company

IOTA Foundation, a German blockchain NPO, announced on May 17 that it is setting aside US$2 million to set up an IOTA Ecosystem Fund to support developers. The announcement has caught much attention in the blockchain community, but not too many are aware that IOTA Foundation is also building its first Asia outpost in Taiwan.Digitimes recently talked to IOTA Foundation co-founder Dominik Schiener and their business partner in Taiwan, BiiLabs CEO and co-founder Lman Chu about why the German NPO chose Taiwan as its springboard in Asia and their vision for the future machine economy.Q: What role does IOTA want to play in future economy? What is your vision for that future?Schiener (S): The way we look at the future economy, there will be a few distributive ledger technologies, two or three, maximum. But those will be focused on specific use cases or industries. I don't think there will be a ledger only for mobility, another only for energy, because the future is inter-connected.For example, the car needs to pay the grid for electricity, so they need to have a shared protocol and virtual coin. The way to eliminate the friction of the trade is to have one ledger for the machine economy. And that is our vision, that one ledger will be used universally. But there will also be other ledger to be used by banks and central banks. We are not really focused on that area; we fundamentally focus on machine-to-machine payments.IOT and blockchain are all about automation. They will be two very essential technology as the backbone systems around us in the future machine economy. The way we envision the future is that you don't even notice where IOTA is used. We want to automate all the processes around us and make our experiences much more seamless. For example, you get into a vehicle and pay per meter for energy as you drive. The vehicle pays the charging station and toll station autonomously. The automatic machines make the entire network much smarter, simply because they are sharing information with each other all the time: "Hey here is an accident, need to call an ambulance," or "Reroute the traffic." We are building an intelligent network around us, where there is the trust layer behind us that nobody can control it, and where there are machines coordinating each other. Automation will make our lives easier, without having to get off the car to get a stupid parking ticket.Q: Will that require a lot of cloud computing power?S: Much more is on edge. We are focused on how one machine can pay another machine for that computing power. For example, one car needs to pay a computing station to execute a specialized algorithm, so you pay for the computation itself. One device collects data, one device buys data, one device sells data, and another buys computation. It is really a network of different actors with different capabilities, but they all work together, like a symbiosis. The most empowering thing is you give machines a wallet. With a wallet the machine can become an autonomous economic agent. As such, a small Raspberry Pi can have the same computing power of a supercomputer, simply because it can buy this computing capability. This is how the future looks like: a machine gets the resource from another machine.Q: How does that work? When a machine transacts with another machine, do they need to identify each other?S: No, that's the whole point. One machine can do a micropayment to another machine. For example, a car pays the charging station. As soon as the payment stops, the energy flow stops. You don't need to sign a contract for that. This doesn't require identification for trust.Chu (C): In the real world right now, we can only do monthly subscription for many services. That is because we still need a complicated system of multiple agents to make sure that you get the service for the money you pay. But in the future, when micropayment works on blockchain, you can pay exactly how much you use the service, by minutes or seconds. That will create many new business models.Q: To do that, you will need good hardware too. Taiwan produces good IoT sensors and AI chips, and are training more software engineers for AI.C: That's why they are setting up the first office in Asia here in Taiwan. Like I said, blockchain is very suitable for Taiwan, but probably not the mining machines. The opportunity lies in the big ecosystem formed by networks through the IOTA distributed ledger core on IoT devices.Q: Tell us more scenarios of machine economy.S and C: When it comes to IoT, we really want to solve the "trust" problem. Today IoT is owned by a few companies that are centralized. And it is a very insecure system, with a single point of failures.And we want to solve the trust problem for data, so we can trust and verify the data from machines. We also want to bring trust into the network, so that you can execute transactions between two machines by setting the transactions themselves. That would make it much more secure. With these two properties - data security and transaction easiness - you can derive other use cases in different industries. For example, audit trails for supply chain, or chain of custody in the factory, and digital twins, are all very important use cases.Or collecting data in a vehicle, you cannot tamper with it. Or in a financial use cases that I mentioned before, when a car has a wallet inside, it becomes an autonomous economic agent. You can start paying the infrastructure, such as the tolling stations, towing and parking stations, energy, etc. And you can also start earning money by selling excess electricity back to the grid. It is also a brilliant way to balance the grid in the future.Another big use case is data - data security and selling of data. When a car drives on the street, it autonomously collects data: "Hey here's a parking space available," "There's a charging station," and so on, and sell the data for the use of other cars.There is unlimited potential. That is why we say permissionless ledger is very important. We as founders can envision what the future can be like. We want to create a protocol and ecosystem that everyone can use our technology in a way just like what BiiLabs has been doing in Taiwan and Japan. In Taiwan BiiLabs works with the Taipei City Government to create Taipei Digital Citizen Card, while cooperating with National Cheng-kung University on the digital existence credentials for graduate diplomas.Q: How many partners do you have all over the world?S: Too many to count! We have a very strong focus on Europe, and more in the US than Asia. We also have dozens of partnerships where we work with big companies, universities, startups, research institutes, in projects that they build with us.We already have an office here in Taiwan, hiring four engineers. Our intention is to grow the office here and hire more people, grow to 10-15 people by the end of the year. For us, Taiwan is really our entry to Asia, simply because the government has shown willingness to work with us. We are very excited about Taiwan. We are already working on multiple smart city projects in Europe and America, for example, Austin, Texas. And now we will do the same thing in Taiwan, setting up a test bed. That is very interesting, because we can have these test beds with citizens to try up new applications. And our focus is also to replicate the success stories we had in Europe, like the charging stations, smart grid system, etc, and do the same here in Asia, in general.We have test beds here surrounding citizen applications, for example, digital identities and data sharing. While in Austin, we have test beds on smart mobility, also cooperating with the government. What we are doing now is specifying user cases, such as how IOTA can help with congestion problem, how IOTA can help car data sharing with cars and the government, and how IOTA can help improve service provision of the government, etc. And in Europe we have smart energy districts in seven cities. We plan to do the same work here in Taiwan with our mobility and energy test beds.Q: Is BiiLab's applying blockchain on car insurance also included in your work here in Taiwan?S: Probably in the future, but not the primary task for now.C: BiiLabs has a different role from the IOTA Foundation. They are the creators of the IOTA Tangle technology, and are responsible for maintaining and developing the technology roadmap. As for BiiLabs, we are a startup. We explore all kinds of business model that we can develop with that technology to solve the current problems of the society together with IOTA Foundation.The user case of car insurance which our customer uses on-board sensors to record drivers' behavior, is to provide data for evaluating the premium of car insurance policy. The customer decides to use blockchain technology for this data audit trail, because they want to ensure that nobody can tamper with the evidence. Drivers would not know there is such a record - all they have to do is obey traffic rules and drive well, and they can get better ratings and lower premiums for their policies.Q: When blockchain becomes readily available in our daily lives, does that mean in the future everyone will need a smartphone with blockchain function?S: I think a blockchain smartphone makes very little sense, it is more of a niche product that cater to the crypto currency club. But we'd rather see the trend as more and more smartphone manufacturers will allow digital wallets in their app stores. In time, the world will be more open for crypto currencies and digital wallets. The interaction with machine economy will be much more seamless when IOTA is used in the payment.Q: Your vision for IOTA Tangle technology is to achieve a frictionless machine economy; however, as there are so many crypto currencies and tokens out there, there's due to be transactions between this currency and the others. How do you solve the problem?S: We see the future very differently, because from my point of view, tokenization makes little sense. If you tokenize, for example, a token for energy, a token for this company and that company, then it will start to fragment the ecosystem again. That will make people difficult to trade with each other. The way we see the future is that there is going to be only one currency for the machine economy, because through that you reduce the friction. If we don't do that, our machines will become foreign exchange traders, constantly checking: "Hey what's the price for this coin, and what's the price for that coin?" That will introduce new layer of complexity. Like I said before, there will be two or three crypto currencies that will prevail, because most of the other ones don't make any sense, they don't have adoptions.Q: Some big countries are considered to be the winners for AI. Do you think the machine economy will come under control of big countries, too? Or will it be different?S: It will be very different. Crypto currencies and the blockchain were introduced to empower the people, so they can decide what happens to the money and have more opportunities for financial inclusion. Blockchain is all about empowering. The Internet brought us free flow of information, and now crypto currencies and blockchain give us free flow of assets. Financial inclusion, just like this new ecosystem, people in Africa can develop an app for you to use it in transactions. At the end of the day people own their assets, and with the transparency of blockchain, you can verify it. People now talk about the social media and fake news that divide people. Blockchain is a powerful tool that can stop people from spreading lies.Q: Yes, since people will have digital ID in the future, it would be easy to trace that on blockchain. But can you also sell your own data? Or the information you curate?S: That is exactly the future of data. You can have your own sovereign over your data, your identity. Besides you can decide who can see your data. You can also decide who gets access to which data. It is also important to be able to sell your data. You can decide whether you want to participate in the data market or not. This becomes a very interesting selling point for companies in the future. One company can say, "Hey our car allows you to sell data that you collect when you drive around," or the other one might say, "Our product will never collect data from users unless they decide." You can differentiate by creating different business models. So people are more empowered as a result of that.We encourage businesses and organizations to research and explore different services and business models with us. The time to experiment is now. Otherwise you would have to pay expensive consultancies to catch up in the future.Q: One of the major concerns with blockchain is security. How secure is IOTA?S: IOTA is open-source and open network and battle-tested. Like Bitcoin, IOTA is battle-tested too. There is a community of very smart people all around the world testing its vulnerability trying to get rich. On the other side, we have very smart people on our team, pioneers of this entire blockchain space to improve the protocols and firewalls. We also have a big community of institutes and partners participating in our research. Through that we gain more trust.Q: What is the major difference between IOTA and other blockchain technology?S: We have a very different approach, and no longer based on blockchain. We are much more scalable, and we have no transaction fees and no miners. Second to that, we are a German non-profit organization, the only one NPO; the others are all for profit. We are based in Germany, work closely with the industry, so we have very strong partnerships there. Our focus is not on crypto currencies. Right now we have more than 100 employees in 23 different countries. Very decentralized.C: BiiLabs is also like that. We have 15 employees in six cities - five in Taiwan, the other is Osaka, Japan.Q: How do you plan to expand to other parts of Asia through Taiwan?S: First we want to do some bigger projects here to build up success stories. We are also talking to some businesses in Japan and South Korea. But Taiwan is definitely our primary focus right now. In the future we can replicate those experiences to other countries. Asia is so big and so spread out, maybe we will expand to Singapore or Hong Kong in the future too. But we need to focus. Right now our focus is in Taiwan.IOTA Foundation co-founder Dominik Schiener (right) and BiiLabs CEO and co-founder Lman Chu (left)Photo: Michael Lee, Digitimes, June 2019

With one of the strongest processor ecosystems in the industry, Arm's role to accelerate the adoption of machine learning (ML) at the edge can't be overemphasized, and its effort to enhance its CPU performance and software support for ML workloads is now in full swing. According to Jem Davies, Arm fellow, VP and GM, Machine Learning Group, the market outlook, challenges, and Arm's overall strategy to address the huge opportunity are clearly disclosed.ML in edge devices is just beginning"We see ML as one of the most exciting advancements in computers and processors in modern times," Davies said. "As machine learning is exploding across edge devices, we're now to the point where we are seeing huge amounts of activities and some really interesting use cases across all markets Arm technology services."From his point of view, some of the most interesting use cases and active communities for ML are coming out of the IoT sector using traditionally very small processors like the Arm Cortex-M microcontroller family.For example, the use cases come from the embedded and IoT spaces, such as life improving medical devices like smart asthma inhalers, through to industrial sorting and robotics to voice assistants, more intelligent home security and even things like DTVs where there is a lot of activity in super scaling, scene recognition and picture quality enhancement and gesture recognition.Of course, there are some of the more interesting ones, including the well covered autonomous vehicle and driver assistance; in smartphone a huge range of applications are implementing ML improvements like smarter games engines, richer social media applications, and even utility applications built directly into the OS like predictive text and voice assistants."From an Arm perspective, the thirst for ML in edge devices is just beginning and we expect it to continue growing substantially for several years yet," he said. "The use cases are still growing rapidly and we expect an explosion of creativity over the next couple years as the algorithms become more understood and smaller and the developer community really engages with what ML can bring."But, the challenges are...However, opportunities always come with challenges. The challenge for Arm is to ensure people have the improved CPU and other processors along with associated software and tools to support their needs today while also ensuring it is working on the products for tomorrow with even more capability, such as ML dedicated processors.But at the customers' side, Davies saw lots of confusion as ML is too new and too complicated for them to adopt. "A lot of what we're working on now is just trying to help demystify and clarify things in the technology space as there's a lot of confusion and misinformation out there.""Two years ago or even a year ago, it wasn't uncommon for people to think that if you wanted to do any ML on a device, you needed to have an ML dedicated processor - a view fuelled by people with dedicated processors to sell - and so we would get asked which processor was best for ML a lot. It depends on what is important to you. So we've spent a lot of time explaining when a small CPU, large CPU, multi-processor CPU, GPU or ML processor would best meet people's needs."On the other hand, one of the biggest challenges for software developers is the same one they always have: which hardware platform/processor should I target to give my software the widest compatibility with devices?Aside from that, models have introduced a new and critical component to the technology stack when you are doing ML. A lot of work has gone on in the industry the last few years to make these better understood and more friendly and accessible.ML are affecting nearly all Arm's productsAccording to Davies, ML is driving a change in software, and Arm's processors and products are all about running software. As such, his view is that ML is affecting nearly all Arm's products."You can see this in the CPUs we have been releasing the last couple years which have had major performance improvements specifically targeted at ML workloads. Often these are 4x or even 10x on generation improvements. This also extends to our latest GPUs and even our efficient Cortex-M family where we recently released our Helium extensions specifically supporting 15 times improved ML performance in microcontrollers.""So Arm is focused on offering a huge range of 'processors' that can give customers a vast range of price, performance and power trade-offs. We will be extending that further through 2019 with a range of complementary ML processors for all markets given the market options where heavy ML workloads are needed or ML power efficiency is critical."He explained that Arm's strategy has been to add capabilities to its CPU and GPU architectures to support ML for some years now. And the company has launched a range of scalable and extensible NPUs (neural processing units) to provide more efficient ML processing in a range of market segments with performance requirements such as in automotive, right down to the tiniest low-power embedded microcontrollers.As he mentioned, there is no one-size-fits-all approach when choosing processors for ML. "We predict that ML will continue to be run across a range of processors, not just NPUs," he said.In addition to this, Arm has been investing heavily in ensuring the software is there to get the best performance out of current and future Arm hardware as well as ensuring software portability.Arm's advantages in the new ML eraIt is fair to say that Arm provides the architectures at the heart of modern computing and Arm-based devices are everywhere around us. With this advantage, when disruptions like ML come along, developers will use Arm's architectures first to innovate on, Davies indicated."We recently launched a survey to measure where ML processing is being performed, and the most popular was the Arm Cortex CPU architecture, followed by the Arm Mali GPU architecture," he said. "That's a huge advantage for us - it gives us great connections to developers who can then tell us what they need us to develop in terms of software libraries, compilers, development tools etc and what they will want to do themselves.""This major shift in the industry is something we started some years ago and we could see it spreading across various areas of the company. That's when Arm's Project Trillium was brought to life to help ensure we were taking a broad, holistic and comprehensive view of ML in the business."He explained that Project Trillium, which covers all of Arm's ML activities, is Arm's and third-party hardware IP and software running ML workloads and applications. Within it are CPU and GPU improvements started many years ago in research, a broad range of complementary current and future ML processors software to enable ML on Arm, tools, ecosystem and a ramping amount of educational material.As ML is fundamentally a software problem, Arm is also investing hugely to support ML developers on Arm. Arm NN provides a framework to allow ML workloads to be easily run across a variety of processors: CPUs, GPUs, NPUs and other IP blocks."We developed Arm NN, investing over 120 engineering years of effort before donating it to Linaro, and now, with Open Source and Open Governance, our partners are contributing their own efforts to Arm NN, in the confidence that it will become an open standard across multiple industries, not controlled by any one company."With these big investments in tooling and ecosystem support and development, Arm aims to provide the broadest and most comprehensive range of ML solutions for all ranges of edge devices, and hope to bear fruit in the years to come.Jem Davies, Arm fellow, VP and GM, Machine Learning GroupArm is enabling ML on edge devices