Chip giant Intel Corp seems to be able to breathe a sigh of relief.

IC Insights, a semiconductor market research company, released a ranking of the world's top 10 semiconductors (IC and OSD optoelectronics, sensors and discrete devices) for the first half of 2020. Among them, Intel Corp's sales of semiconductor products in the first half of the year were 38.951 billion US dollars, an increase of 22 percent over the same period last year.

Sales of the world's top 10 semiconductor manufacturers totaled $147.093 billion, up $21.113 billion from $125.98 billion in the same period last year, up 17 per cent from a year earlier and more than triple the 5 per cent growth rate of the global semiconductor industry in the first half of the year.

In terms of sales, although Intel Corp has been in turmoil recently, he has still stabilized the position of "the world's largest semiconductor manufacturer". In terms of sales and growth, he has shrugged off Samsung, which ranks second. But behind the seemingly prosperity, Intel Corp is under a lot of pressure.

Apple Inc announced at the Global developer Conference in 2020 that he would break up with his partner of 15 years, Intel Corp, to install a self-developed chip with a set of ARM instructions on his Mac series of products, so as to completely open up the underlying computing architecture of mobile and PC.

Then, since July, NVIDIA Corp's share price has soared, replacing Intel Corp to become the largest chipmaker in the United States by market capitalization for the first time, and recently there has been news that NVIDIA Corp is going to buy ARM. On July 25th, the news of the delay in Intel Corp's 7nm chip process directly pushed its share price into a dive. At the start of the day, it immediately plunged more than 15%, reaching a four-month low, while the share price of AMD, its biggest competitor on PC, rose more than 70%, with a market capitalization of more than $100 billion.

The possible market of Intel Corp 7nm chip extension has been expected. As early as 2016, the 10nm process that Intel Corp had planned to launch was finally delayed until 2019, and the 7nm process technology that should be launched in 2019 was also delayed again and again due to the problem of good product rate. But what makes Intel Corp more nervous is the shift of Apple Inc's Mac series products to ARM. Bernstein, the Wall Street investment bank, estimates that Apple Inc's laptop business accounts for 2-4 per cent of Intel Corp's sales and 6-8 per cent of its revenue. In terms of the proportion of sales alone, although the impact seems to be limited, the real blow may come from the demonstration effect brought by Apple Inc.

In fact, before Apple Inc tried to transfer to ARM, Microsoft Corp began to adapt the ARM architecture on the Surface Pro X series of products, but the demonstration was not strong, and there was not enough motivation to go deep into ARM, not to mention the seemingly inseparable Wintel alliance. But in any case, ARM has become an important choice for all brands of PC chips except X86 architecture.

Also within the X86 architecture, AMD, the former younger brother of "overlord" Intel Corp, took the lead in mass production and commercial use of 7nm chips, with the stock price rising 6.8 times in five years; under different instruction sets outside X86, ARM has a solid position on Internet mobile intelligent terminals and is eyeing Intel Corp's PC market covetously. Why did the former chip "boss" Intel Corp be lonely in the mobile Internet era? And can ARM lead the next trend of the times?

Intel Corp's defeat of the mobile chip

Bill Gates once said bluntly that not paying attention to the mobile phone market was the biggest mistake he made in Microsoft Corp, and it was a technical mistake that could be avoided. "

Microsoft Corp is not alone, and it is not easy for giants to maintain their huge size across the deep ravines of the times. Obviously, Intel Corp did not take this step. As the tide of the PC era faded, Intel Corp appeared to be at a loss in the smart mobile end and lost step by step in the competition for the mobile Internet chip market.

Intel Corp is not doing nothing in the mobile Internet chip market.

In 1997, processor giant DEC sold its semiconductor manufacturing business to Intel Corp, giving Intel Corp the right to develop StrongARM. To create a faster ARM processor, Intel Corp replaced their previously developed i860 and i960 with this chip. StrongARM uses ARM v4 instruction set architecture. At the frequency of 233MHz, this CPU consumes only one watt of electricity, which was an excellent product with high performance and low energy consumption at that time.

On the basis of StrongARM, in 2000, Intel Corp developed the CPU--XScale processor, which began with the v5TE instruction set of ARM architecture, which not only extends the architecture but also retains the backward compatibility with previous products. At the same time, compared with the standard ARM ARM9 and ARM11 processors, XScale has lower power consumption, better scalability, and the core frequency has been improved to more than 400Mhz, so XScale has a good reputation at that time.

Although the performance of XScale is strong, it is a pity that it was born at the wrong time.

At the beginning of the 21st century, the mobile phone market is still dominated by functional phones, and the market for smart phones is very small, but the demand for mobile chips with high performance and high energy consumption is not high. You know, feature phones at that time needed to be recharged every three to five days, which was completely different from people's daily charging habits. Phones with more energy-efficient chips and reduced standby time were unbearable for users at that time.

At the same time, because Intel Corp has accumulated insufficient baseband patented technology, mobile phone manufacturers need to configure additional baseband chips when carrying XScale, which increases the production cost of mobile phones and the difficulty of mobile phone design. The quality of baseband technology will even affect the success or failure of the whole processor. Qualcomm Inc (QCOM.US) dominates the global communications market in the whole 3G and 4G era because of his powerful baseband technology. Now, after winning the antitrust lawsuit of the Federal Trade Commission (FTC), Qualcomm Inc continues to adopt the business model of patent fees and chip sales to win the 5G era.

While the prospect of the mobile chip market is bleak, in 2006, Intel Corp's PC business was fiercely attacked by its old rival AMD, and its market share was leveled. Just a year ago, something that is now sighing for the industry, Intel Corp's fifth CEO Otellini rejected Jobs' offer that Intel Corp provide chips for the original iPhone.

Looking back now, Intel Corp seems to have missed the best time to lay out mobile chips, but in the eyes of Intel Corp at that time, it was a reasonable choice: in 2005, the global mobile application processor market totaled only $839 million. Intel Corp's revenue reached $38.8 billion in the same year, obviously, the mobile chip market at that time was difficult to meet Intel Corp's appetite, and the PC market was Intel Corp's revenue base.

In order to keep the basic market of its PC end market, Intel Corp directly cut off the mobile communications business that did not seem to make money at that time (this was also the largest layoff in Intel Corp's history. XScale sold the communications and application processor business with Intel Corp to Marvell in June 2006), shrinking the front and devoting all its efforts to the PC side. Naturally, he would not consider dividing the R & D team and production line to produce mobile phone chips for Apple Inc at that time.

Under the background at that time, it seemed not a bad choice to cut off the mobile business and stick to the position of PC, and Intel Corp did retain the number one position in the PC CPU market and beat AMD again. But after missing the historical window, it became extremely difficult for Intel Corp to return to the mobile chip market.

After that, in 2008, Intel Corp tried to restart the mobile chip market by developing Atom through Intel Bonnell microprocessor architecture, but due to unclear market positioning, unimproved high energy consumption, failed baseband technology and other reasons, Atom products wandered between mobile phones, tablets and low-cost PC. In 2019, with the reconciliation between Qualcomm Inc and Apple Inc, Intel Corp sold the 5G business to Apple Inc and completely withdrew from the mobile baseband market.

In the era of mobile Internet, Intel Corp, who is faltering, faces the cruel reality that the market share of PC is shrinking, while the territory of mobile Internet chips is constantly expanding. The market base supporting Intel Corp's rapid growth has changed, so Intel Corp is in an awkward position.

At a time when Intel Corp is losing step by step in the mobile chip market, ARM is taking advantage of the growth in the era of mobile Internet.

ARM to be a giant

If Intel Corp is the "overlord" of the PC era, then ARM (Amou) has boarded the "throne" of the mobile terminal era.

The original abbreviation of ARM is Acorn RISC Machine. Founded in 1978, Acorn Computer is headquartered in Cambridge and was founded by Andy Hopper (Cambridge University), Chris Curry (Sinclair Research) and Herman Hauser (Cambridge University).

In the era when Intel Corp Intel x86 architecture dominated PC, Acorn applied to Intel for a 80286 processor sample, but it was rejected, so Acorn chose to develop its own processor.

In October 1983, Acorn launched a project codenamed Acorn RISC, which was produced by chip manufacturer VLSI Technology. On April 26th, 1985, VLSI produced the first Acorn RISC processor ARM1. In order to avoid competing with the giants in the market at that time, ARM chose the design concepts of low-cost, low-power and high-performance, which just stepped on the demand for smartphones.

Although in the following years of development, because the explosive growth of mobile smart terminals was as late as 2008, ARM had a very bumpy experience. But it is in the shadow of giants such as Intel Corp, ARM, which avoids the edge of giants, grows up quietly with the rise of mobile intelligent terminals.

According to the latest data, ARM has more than 1000 processor licensing partners worldwide, 320 processor optimization packages and physical IP package licensing partners, 15 architecture and instruction set licensing companies, and more than 95% of the world's smartphones and tablets use ARM processors. ARM's annual shipments have increased from 10 billion units in 2013 to 20 billion units per year now.

If it is difficult to feel the popularity of ARM in the market just from the numbers, then look at the support of ARM competitors Intel Corp, Qualcomm Inc, Western Digital, SK Hynix and so on to ARM competitors-SiFive Inc--, which is committed to commercializing open source semiconductor technology, we can feel the importance of ARM to the industry.

To be able to break through the tight encirclement among the giants has a lot to do with the ARM model.

One of the important reasons for the rise of ARM is to give up the traditional IDM mode (Integrated Device Manufacturing, integrated equipment production mode), so that many manufacturers can continue to design and innovate in their instruction sets. According to the different needs and capabilities of chip design companies, ARM provides three different modes of external authorization:

One is the IP authorization mode of the processor. ARM grants the IP patent to the company that bought the design, and the company can produce it directly according to its design.

The second is the authorization mode of processor and physical IP. Under this mode, enterprises need to strictly follow the technological process stipulated by ARM to produce chips.

The third is architecture and instruction set authorization. This is the most liberal licensing model, and powerful mobile phone manufacturers such as Samsung, Apple Inc and Huawei can design and optimize the chips themselves under the ARM architecture. Among them, RISC-V instruction set is more radical, which is directly based on RISC (reduced instruction set) principle and open source instruction set architecture, which allows anyone to design, manufacture and sell RISC-V chips and software without paying royalties to any company. The RISC-V architecture is used in the Xuantie 910 (XuanTIE910) processor developed by BABA's semiconductor company, Pingtou, in 2019.

The IDM mode represented by Intel Corp undertakes all aspects of the chip from design to manufacture to packaging. In the early stage of the development of the integrated circuit industry, enterprises generally use this all-inclusive IDM model. In the field of mobile terminals, low-cost, low-profit processors can not support this heavy investment model.

The authorization model created by ARM separates the architecture and design of the chip (Fabless, such as Qualcomm Inc, MediaTek, Hayes) and production (Foudary, such as Taiwan Semiconductor Manufacturing Co Ltd). It activates the vitality of the whole industry chain, not only improves production efficiency and the speed of new process iteration, but also reduces production costs, thus forming an increasingly prosperous ARM ecology.

In other words, the light ARM can run faster than the huge Intel Corp, and finally became a big worry for Intel Corp.

ARM has embarked on a road that is different from Intel Corp's business model and market positioning. With the saturation of the PC computer market and the arrival of the mobile Internet era, ARM will also extend its tentacles to Intel Corp's PC end market. But are Intel Corp and X86 a thing of the past? The actual situation is not that simple.

Who will control the ups and downs in the future?

It is also controversial to tell which X86 architecture is better than the ARM architecture, and who will unify the market industry, which involves not only the technical differences between the underlying RISC instruction set and CISC instruction set, but also different use scenarios of servers, PC and smart mobile devices, which requires designers to make choices in chip performance, efficiency, energy consumption and application scenarios.

Technology is important in the chip market, but it is no longer the only yardstick. With the further development of technology, both ARM and X86 are gradually making up for their shortcomings. They no longer only adopt a certain instruction set in RISC and CISC, but become RISC and CISC. There is almost no substantial and clear difference between the CPU architectures marked RISC and CISC.

Although Intel Corp is beset with difficulties in recent years, he is not poor in the realization of earnings. According to Intel Corp's second-quarter results on July 24, Intel Corp had revenue of $19.7 billion and net profit of $5.1 billion, up 20% and 22% respectively from a year earlier.

Although Intel Corp's 10nm process chip has "skipped tickets" several times, its technical background is still there. According to Intel Corp's top engineer, the Tiger Lake micro-architecture CPU, which has been difficult to give birth for several years, will be released on September 2. It will adopt the upgraded version of the FinFET structure-SuperF technology, which can provide stronger epitaxial source / drain, improved gate process, and additional gate spacing. The transistor density of the 10nm process chip is twice that of Taiwan Semiconductor Manufacturing Co Ltd and Samsung, which is about the same density as Taiwan Semiconductor Manufacturing Co Ltd's 7nm. In other words, the 10nm process efficiency launched by Intel Corp can be equated with the process efficiency of other manufacturers' 7nm chips.

Internally, Intel Corp began a series of reforms. Brian Kozaiqi, Intel Corp's former CEO, has been trying to transform Intel Corp into a data company, providing the technical foundation for the new data economy. Intel Corp's share of traditional PC business has dropped from 80% to 50% now, and the decline has been replaced by data center business. Intel Corp occupies the top position in the industry, whether it is storage, traditional PC business or new business such as data center, and it also has a layout in new markets such as industrial systems and self-driving.

For ARM, in addition to continuing to upgrade and innovate in technology, we also need to improve the ecological layout:

The linkage of upstream and downstream to build the industrial availability of ARM. ARM server chip should be used as the underlying application sequence to meet the basic applications of downstream manufacturers in server scenarios, including data, storage, invocation and so on, so as to reduce the compatibility cost of customers.

Speed up the industrial layout and establish ARM ecology. Windows/X86 has developed for more than 30 years, and there are countless application software in the ecology. The barrier formed by unified hardware, development and infrastructure is an inevitable barrier for ARM to occupy the PC market. And younger companies are already coming.

On August 12th it was reported that SiFive Inc had raised an additional $60 million from investors including SK Hynix and Saudi Aramco. Its existing investors, including venture capital departments from Intel Corp, Qualcomm Inc and Western Digital, have also joined the round, valuing the company at more than $185 million.

SiFive is heavily supported because the young company gives the boss hope against ARM: SiFive's line of CPU blueprints launched in October 2018 threatens ARM's fee-based products and has 64-bit architecture products that the latter lacks. ARM even launched a website, riscv-basics.com, to criticize RISC-V, despite its notoriety.

Intel Corp's story has proved that the technical leadership and the success of the business model have temporary limitations. In the new era of the Internet of things composed of 5G and trillions of connected smart devices, maintaining innovation and embracing change is still the foundation of Intel Corp and ARM.