What is ASIC?
So-called Application Special Integrated Circuit Semiconductor Design (ASIC), that is, design service companies specialize in providing multiple varieties and small batches of chips for specific applications such as home appliances, communications, and servers. From design, development to commercialization, engineers work closely with customers Cooperate to create corresponding customized products.
Why its outlook is promising?
Three waves of tailwind
The demand for ASICs has already emerged in early 2018, but at that time it was mainly for mining machine manufacturers developing Bitcoin (such as Bitmain) to develop ASIC chips to provide needs.
Generally speaking, at the beginning of 2018, most ASICs were needed for edge computing in terminal devices, and now this wave of ASIC customization is spreading all the way to the cloud computing field.
The demand for ASIC customization has slowly taken off in the cloud server industry. This is also due to the fact that manufacturers have become more clear about the development of artificial intelligence algorithms and the services that artificial intelligence can provide. The latest pusher is artificial intelligence. The most convincing example is that artificial intelligence chip manufacturer Nvidia is optimistic about ASIC and is preparing to open a new ASIC department in the company.
ASIC industry prospects
The development of the ASIC industry will benefit from three major trends in the future, including: the rise of automotive high-efficiency computing (HPC), Arm-based CPUs becoming mainstream in the server market, and semiconductor IP silicon intellectual property driving the design of the next generation of AI chips.
In the field of AI automotive chips, benefiting from the take-off demand for automotive HPC ASICs, in addition to Tesla’s fully autonomous driving (FSD) design, mainland China’s electric vehicle manufacturers also need to have their own designs to cope with geopolitical risks.
In terms of Arm architecture CPU ASIC design, in addition to Amazon AWS launching Graviton and Microsoft Cobalt servers, Meta is also considering launching server CPUs based on Arm architecture; Taiwanese creative companies are competing with American counterparts for projects.
How big is the market?
The IDC report estimates that ASIC will account for more than 12% of the overall AI processor market by 2027, and the future growth potential cannot be underestimated.
Alan Weckel of research firm 650 Group estimates that the data center customized chip market will grow to $10 billion this year and double by 2025, while Needham analyst Charles Shi said that by 2023, broader customization ASIC chip market is worth approximately US$30 billion, accounting for approximately 5% of global annual chip sales.
In the narrow sense of the tens of billions of US dollars market, in the broad sense of the 30 billion US dollars market, even Nvidia, the world’s most valuable listed semiconductor company, is interested in ASIC. It is not difficult to imagine the fierce competition in this market in the future.
The leading vendors
In the customized chip market, there have long been two giants: Broadcom (ticker: AVGO) and Marvell (ticker: MRVL). In the high-end ASIC market, Broadcom is firmly in the lead with a market share of 35%, followed by Marvell with a market share of 12%. Both Broadcom and Marvel believe that as data center processors become more diverse, the customized chip model will be revitalized.
The founder of research organization SemiAnalysis said that Broadcom’s customized chip business has a scale of US$10 billion and that of Marvell has reached US$2 billion.
Major ASIC listed companies in U.S.
Broadcom’s ASIC business
Overview
Broadcom’s ASIC chip design-related business under the famous LSI (ticker: LSI), which it previously acquired for application-specific integrated circuits (ASIC), can be said to be the start for Broadcom’s chip design business to step into ASIC.
Major ccustomers
Broadcom’s ASIC business’s full-year revenue in 2020 was US$750 million, up from US$50 million in 2016. In addition to chip design, Broadcom also provided key intellectual property rights to Google, a subsidiary of Alphabet (tickers: GOOGL and GOOG) , and was responsible for the steps of manufacturing, testing and packaging new chips to supply Google’s new data center.
One of the most well-known, successful, and important cases: Broadcom has actually customized TPUs for Google for nearly 10 years, and recently reached a cooperation with Meta to produce customized chips for it. Other customers including Microsoft, and AT&T.
Large customers disclosed by the outside world as of 2024 include Alphabet, META, Apple, ByteDance (the parent company of TikTok) and OpenAI; and there is another large customer that is about to finalize the contract.
Charlie Kawwas, president of Broadcom Semiconductor Solutions Group, said that Broadcom has captured 15% of the market share in the artificial intelligence chip market in 2023 and is expected to reach 25% of the market share in 2024, with total revenue exceeding US$10 billion.
Advantages
PCle Switch is very important for artificial intelligence chips and data center chips, and this field is dominated by Broadcom. The three major PCle Switch suppliers in the world, Broadcom, Microchip (tickers: MCHP), and Xiangshuo Technology (Taiwan stock code: 5269), together account for approximately 58% of the global share. As the leader in PCI Express switching, Broadcom has shipped more than 1 billion units.
Broadcom is the undisputed monopoly giant in SerDes (serializer/deserializer) communication technology. The SerDes interface converts low-speed parallel data into high-speed serial data before transmission, and then converts back to parallel data at the receiving end. The purpose is to allow data to move from one TPU to another at high speed, improving signal transmission efficiency. In the global 50GB/S SerDes market, Broadcom occupies 76% of the market.
Outlook
Charlie Kawwas, president of Broadcom Semiconductor Solutions Group, said that Broadcom captured 15% of the market share in the artificial intelligence chip market in 2023 and is expected to reach 25% of the market share in 2024, with total revenue exceeding US$10 billion. He emphasized that Broadcom’s The custom XPU outperforms the competition by a full 50% on HBM, and Broadcom can do it better, faster, and with more power than anyone else.
Regarding Broadcom, please see my special article for details:
Marvel’s ASIC business
Overview
Marvell Technology mainly acquired Avera Semiconductor, an ASIC subsidiary of GlobalFoundries (ticker: GFS); Avera was formerly the customized ASIC department of GlobalFoundries (ticker: GFS)). Even earlier, Avera was actually the custom ASIC division of IBM (ticker: IBM). Avera’s transaction helped it become the world’s second largest ASIC manufacturer, second only to Broadcom (ticker: AVGO) in market share.
Marvell also offers Customer Specific Standard Products (CSSPs), where customer accelerators and interfaces can be integrated directly into Marvell’s Octeon processors.
For Marvell in detail, please see my ppost of “How low-key Marvell makes money?“
Advantages
Marvell provides customized ASICs tailored to customers’ specific design goals. It also provides ASIC development services to the aerospace and defense industries through its independent subsidiary Marvell Government Solutions (MGS). In the joint venture with TSMC (ticker: TSM), Marvell will launch 3nm products.
Major customers
The most famous, important, and successful case of Marvell is: Amazon. The Trainium chip of Amazon’s AWS is produced by Marvell.
Microsoft’s Thunder X series is also built by Marvell. One of the two Arm server CPUs developed for Google, code-named Maple, is based on Marvell’s technology.
Moreover, Maiwell also revealed that the company has found another large customer and will soon contribute revenue in the future.
Outlook
Although Marvell CEO Matt Murphy said that Marvell is developing AI chips for a third customer and will enter production in 2026, he also told investors that the gross profit margin of customized chips is lower than that of the company’s off-the-shelf products. The product, the so-called “commercial” chip, needs to rely on higher revenue to offset the lower gross profit margin business in order to maintain current profit levels.
For details about Marvell Technology, please see my special article:
“How low-key Marvell makes money?“
Nvidia
Overview
In February 2024, it was reported that Nvidia had established a new business unit to focus on designing customized chips (ASICs) for cloud computing companies and other companies.
It is not surprising that Nvidia will enter ASIC for two reasons: first, to occupy new markets. Second, protect yourself from substitutes.
Major customers
It mainly targets large cloud computing companies as customers, and Nvidia currently has major North American customers, including Alphabet’s Google Cloud, AWS, Microsoft Azure and Meta.
Because these companies have already begun to develop their own chips, such as Google’s TPU, Microsoft’s artificial intelligence accelerator Maia and Cobalt designed specifically for Azure.
Advantages
Nvidia entered this market mainly with its own high-speed transmission NV Link, software CUDA and Omniverse software suites, which have brought great advantages to Nvidia.
Disadvantages
Nvidia’s rush to acquire ASICs has aroused great concern in the industry, but Nvidia’s move may conflict with the interests of customers. Mainly for cloud service manufacturers, Nvidia already provides artificial intelligence GPUs. Artificial intelligence GPUs are quite expensive. In order to avoid being dominated by Nvidia, major manufacturers will definitely use their own ASICs. At the same time, cloud service manufacturers have Their own development plans and direction do not want to be limited by the shortage of artificial intelligence chips. Moreover, ASICs are customized for a single customer, so Nvidia must also consider whether it is worth entering.
Regarding Nvidia, please see my previous article for details:
- “How does nVidia make money, Nvidia is changing the gaming rules“
- “The reasons for Nvidia’s monopoly and the challenges it faces“
- “Why nVidia failed to acquire ARM?“
- “Revisiting Nvidia: The Absolute Leader in Artificial Intelligence, Data Center, and Graphics“
ARM
Overview
In April 2023, relevant reports from the British “Financial Times” pointed out that ARM (ticker: ARM) will carry out the manufacturing of the most advanced chips at its headquarters in Cambridge, England. It is reported that this chip has been manufactured six months ago Research and development has begun, and ARM has established a new “solution engineering” team to develop these prototype chips for mobile devices, laptops, and other electronic products.
Partners
In mid-April 2023, Intel (ticker: INTC) and Arm announced a cooperation. Intel’s chip foundry department will produce OEM chips for Arm based on the Intel 18A process, which means that Arm has Production has also gained a foothold.
Advantages
The concept of Arm Total Design is to bring together a group of industry leaders to work together to accelerate and simplify development based on the Arm Neoverse Computing Subsystem (CSS). In just a few months, ARM Total Design has proven to be a rapidly growing professional ecosystem that can innovate on ARM’s Neoverse platform to meet the massive growth in data center needs.
In addition to being the most upstream of the semiconductor industry, Arm has customers covering mobile phones, electric vehicles, cloud, AI and many other fields. It can be said that it has spared no effort in entering ASIC. In October 2023, ARM launched the Neoverse computing subsystem (CSS). Neoverse N2 is the first-generation CSS product and can provide a complete solution for ASIC development. It has been adopted by Microsoft Cobalt self-developed chips. Less than half a year later, the third-generation technology roadmap was launched in mid-February 2024, providing two new platforms, Neovers CSS V3 and N3, with higher performance and energy efficiency, once again exciting the market.
Disadvantages
Like Nvidia, ARM is rushing into ASIC, which may conflict with the interests of customers.
Regarding ARM, please see my previous article for details:
Current status of Taiwan’s ASIC industry
Three major vendors
According to the Nikkei Shimbun, the total sales of Taiwan’s top three related companies that specialize in providing semiconductor design support to customers have more than tripled in four years, driving the rapid development of the U.S. technology industry and creating opportunities for local foundries to produce chips. Taking advantage of the opportunity, without realizing it, Taiwan’s chip design has started a trend of custom-made chips.
Revenues of top 3 soared
The top three companies in Taiwan’s chip design services are Alchip, Global Unichip and Faraday. As of December 2023, the total annual sales of these three companies are 68.6 billion New Taiwan Dollar (approximately USD 15.504 billion ), more than three times that at the end of December 2019.
Advantages of Taiwanese ASIC vendors
Although giants such as Broadcom and Marvell can provide diversified design services, the advantage of local Taiwanese companies is that semiconductor supply chain manufacturers are currently concentrated in Taiwan. The so-called first-come-first-served basis means that both chip manufacturing and chip packaging can be done on the island of Taiwan. Local companies also have an inextricable relationship with OEM packaging and testing companies, which is difficult to compare with Silicon Valley in the United States. Compared with being close to the market, being close to the source of the factory gives them their own unique advantages in customizing chips.
In fact, TSMC, the leading foundry, is the major shareholder of Creative Electronics, while UMC, also a major foundry, is the major shareholder of Faraday Technology. Although Shichi does not have a foundry as a major shareholder, it has a cooperative relationship with TSMC. Together, the two have created an ecosystem that connects design services and production outsourcing.
Top ASIC vendors in Taiwan
Alchip
Introduction
Shichi (Taiwan stock code: 3661) was established in 2013. In addition to ASIC, Shichi also provides NRE (commissioned design) piece-rate services.
Benefiting from the booming Bitcoin market, Alchip’s revenue in 2014 grew from NT$2.58 billion to NT$4.69 billion, and its net profit after tax doubled from NT$113 million to NT$211 million. When it was listed in October 2014, The next month, the stock price rose from NT$95.9 to NT$118.5, which can be said to be both fame and fortune.
After 2015, the Bitcoin market continued to decline, and Alchip suffered a heavy setback. Its stock price, revenue, and profit plummeted, which caused it to fade out of mining machine chips for a time. It only returned after the market picked up. However, at this time, Alchip had turned to high efficiency in time. Capable of computing and artificial intelligence, this also paved the way for it to become the leading ASIC manufacturer in Taiwan.
Customers
Alchip’s current two main customers are Amazon and Intel.
Amazon’s Inferentia series is a famous product designed by Alchip.
Alchip also participated in the design of part of Tesla’s (ticker: TSLA) supercomputer Dojo 1.
Li Auto (ticker: LI) also uses Sechip’s design services to design its own automotive chips.
Operating performance
Alchip has grown rapidly, with sales increasing approximately 7 times. It has set new records for six consecutive years. Revenue, profit, net profit and earnings per share have all reached record highs. Although the company’s stock price has been in a correction phase recently, it is still the highest in Taiwan. One of the stocks with the highest price, already 10 times the share price four years ago.
Alchip, who is good at seizing opportunities, not only made the biggest pot of gold in the Bitcoin wave, but also made the biggest profit when the artificial intelligence trend began. It is no wonder that it has been listed on the Taiwan stock market twice. “King of Stocks”.
Global Unichip
Introduction
Founded in 1998, Global Unichip’s performance has reached record highs for four consecutive years, but its profits declined slightly in 2023, with net profit per share of NT$26.18.
Customers
Global Unichip designed Athena, a dedicated artificial intelligence chip for Microsoft starting in 2024, but its operational help does not seem to be that obvious.
Faraday
Founded in 1993, Faraday is a member of the UMC (ticker: UMC) family of major semiconductor foundries.
Introduction
So far, it has not won enough new orders from the giant, and its performance is even more dismal. Its revenue and profits will decline in 2023, with net profit per share of NT$6.39.
Customers
Faraday IP and ASIC design service customers include Broadcom and Socion Technology.
Faraday also recently announced that it is cooperating with Intel and ARM to develop a 64-core system-on-a-chip using Intel’s 18A process.
Andes Technology
Introduction
Due to the U.S. embargo on Chinese semiconductors, China has been forced to develop a large number of its own chips in recent years. RISC-V, which is an open system, does not require licensing fees, and is not subject to any government control, is the main system platform for the chips developed by China.
Andes Technology (Taiwan stock code: 6533), which has been working hard on RISC-V for a long time, has a large amount of IP stored in RISC-V CPU technology and may also become a key partner for China’s CPU autonomy in the future.
In terms of next-generation artificial intelligence chip design driven by semiconductor IP silicon intellectual property, Andes has launched its AX65 RISC-V core.
Customers
Jingxin said it has secured two to three large customers from mainland China and the European market.
Lin Zhiming, chairman of Anders Corporation, confirmed for the first time in an interview that Meta (ticker: META) is Anders Corporation’s largest customer, and this has been going on for at least three years. “Meta contributed about 15% to the company’s financial report revenue. The self-made artificial intelligence chip “Training and Inference Accelerator” (MTIA), which is expected to be used in 2025, is the latest result of the cooperation.
MediaTek
Introduction
MediaTek (Taiwan stock code: 2454) began to provide ASIC design services as early as 2011. In early 2018, MediaTek officially announced that it will vigorously expand its ASIC design service business.
Customers
In 2021, MediaTek successfully broke into the supply chains of Google and Amazon and won Google’s new generation tensile processor
(TPU) Large order.
Since MediaTek’s ASIC business began to enter Cisco (ticker: CSCO) with server Serdes chips in 2018, it has successfully won multiple projects from major customers such as Google and Amazon, and currently there are also new US series Customers discuss new projects with MediaTek.
RDC Semiconductor
In April 2024, it was reported that China’s public sector will begin to phase out personal computer products using American processor manufacturers Intel and Advanced Micro Devices (ticker: AMD), and gradually replace them with domestically produced processor products.
Since RDC Semiconductor can provide x86 architecture processor ASIC services, it is regarded as one of the beneficiary vendors that has the opportunity to win relevant cases under this wave of trends. Jinlike is also optimistic about this and emphasizes that it will continue to seek more development projects. opportunity.
I am the author of the original text, the essence of this story was originally featured on Smart Magazine, Issue of June 2004
Related articles
- “ASIC market is getting bigger, and related listed companies in the US and Taiwan“
- “How low-key Marvell makes money?“
- “The reasons behind Broadcom share price consistantly outperformance“
- “How does Applied Materials, lord of semiconductor equipment, make money?“
- “How does Lam Research, top chip equipment player, make money?“
- “Artificial intelligence benefits industries“
- “TSMC negative corp culture and management style are detrimental to its future and growth“
- “Four chip companies account for one-third of S&P 500 gains so far this year
- “Top five lucrative artificial lucrative intelligence listed companies“
- “What’s TSMC DCF intrinsic value?How to calculate it quickly with a free tool?“
- “How many fabs and houses does TSMC have currently and in the future?“
- “Comparison of TSMC, Samsung, Intel’s Yield and Advanced Process“
- “The TSMC cost, sell price, and R&D cost of chip foundry“
- “ASML, who dominate TSMC’s fate“
- “Comparison of SMIC, Rapidus, TSMC, Intel, and Samsung’s new process roadmaps for future chips“
- “Two long-term threats to TSMC: US and SMIC“
- “Why is TSMC’s profit margin much greater than competitors?“
- “How does TSMC make money?“
- “Zyvex and sub-nanometer semiconductor processes, will Zyvex threat TSMC?“
- “TSMC gets emerging and serious challenges“
- “How does nVidia make money, Nvidia is changing the gaming rules“
- “The reasons for Nvidia’s monopoly and the challenges it faces“
- “Why nVidia failed to acquire ARM?“
- “Revisiting Nvidia: The Absolute Leader in Artificial Intelligence, Data Center, and Graphics“
- “Will Intel go bankrupt?“
- “How does Intel make money? and the benefits to invest in it“
- “Intel’s current difficult dilemma“
- “How does Micron make money? What investors should know?“
- “How does Texas Instruments make money? Amazing long term capital reward and company net profit margin!“
- “How AMD makes money? A rare case of turning defeat into victory“
- “Why is AMD’s performance so jaw-dropping?“
- “Qualcomm diversifies success, no nonger highly dependend on phone“
- “How does the ubiquitous Arm make money?“
- “How does Broadcom make money? Significant changes in Broadcom’s business approach“
- “Data center, a rapidly growing semiconductor field“
- “The lucrative semiconductor supply chain“
- “Global semiconductor chip market in detail, big dominators and markets“
- “6 common wrong semiconductor investment myths“
- “Gen 3 semiconductor“
- “Three EDA oligopoly vendors: Synopsys, Cadence, and Siemens’s Mentor Graphics“
- “How does, the EDA oligopoly, Cadence make money?“
- “How does Synopsis, the EDA oligopoly, make money?“
Disclaimer
- The content of this site is the author’s personal opinions and is for reference only. I am not responsible for the correctness, opinions, and immediacy of the content and information of the article. Readers must make their own judgments.
- I shall not be liable for any damages or other legal liabilities for the direct or indirect losses caused by the readers’ direct or indirect reliance on and reference to the information on this site, or all the responsibilities arising therefrom, as a result of any investment behavior.
