author: Chen Yan, such as edit: to seal, from the title figure: Vision China

Recently, Intel is very upset.

First, Apple announced at the 20th WWDC that it had broken up with a good friend of 15 years and switched to a self-developed chip based on the ARM architecture.

Next, since July, Nvidia’s stock price has risen all the way, replacing Intel, becoming the US chip maker with the highest market value for the first time. Recently, it was reported that Nvidia would acquire ARM.

Nvidia is in the limelight, and CEO Huang Renxun is overwhelmed with joy. As early as 2008, in an interview, he said “Intel is the most popular punching bag.” Whether it is ridicule or self-confidence, it is inevitable to say this at the time. It makes people feel that morality is not matched, time has changed, and now Huang has the confidence to be proud.

In contrast, Intel appears dull. On July 25, the news that Intel’s 7nm chip process progress was postponed directly pushed its stock price to dive. The market opened that day, and it plunged over 15% in an instant, a record low in nearly four months.

Obviously, there is not much time left for Intel. As the wave of the PC era fades, its one-man show may have come to an end. New players such as Qualcomm, MediaTek, and TSMC have all made their debut, squeezing out old actors to the edge. A group of old IT representatives such as Intel, IBM, and Oracle have collectively lost their voices in the mobile era, leaving regrets.

Gates once bluntly said that leaving a vacuum in the mobile phone market is his “biggest mistake made at Microsoft, and it is a technical mistake that can be completely avoided.”

And Intel, its long-time ally, also had the same distress, which caused it to hesitate in the mobile field, killing several back and forth, but always gaining shame. From mobile phones to tablet computers to baseband construction, Intel has done everything that can and should be done, but Intel has been unable to maintain its market position.

It’s like a drowning person. The more you struggle, the faster you sink.


1. Yesterday evening

The lighting of the new world occurred at dusk in June 1971.

Chip designer Fan Jin lights up the small round crystal on the table, just as electricity in the entire space gathers at the tip of a lightning rod, starting the prelude to the information age.

This chip numbered 4004 was the first microprocessor made by Intel Corporation. Its birth marked the start of the computer revolution, and a huge PC empire set sail from here.

From the initial 4004 and 8008 to the Pentium series, to the Core Duo processor, Intel’s every technological innovation has injected new vitality into the IT world, and these great innovations come from Intel’s commitment to Moore’s Law based on the principle of constant change Resolutely abide by.

One of the founders, Gordon Moore, once put forward such a slogan: “Change is our lifelong love. For Intel, this sentence may be as important as Moore’s Law itself: integrated circuits can accommodate The number of transistors will double about every two years. Moore’s Law is based on technological innovation, which determines that change is the way Intel survives and competes.

In recent years, Intel has been dubbed a “toothpaste factory” because of slower update speeds and too little performance improvement, which is increasingly inconsistent with the image of the guardian of Moore’s Law.

The period from Core2 to the second-generation Core i that technology fans generally missed is gone forever. At this stage, the performance of Intel CUP has been greatly improved, and netizens kindly call it “the era of the meat.” At the time when the fierce battle between Intel and AMD was in full swing, Intel processors relied on substantial performance improvements to change the battle and once again achieved absolute dominance in the PC chip field.

I thought this would be the beginning of a triumphant advance, but I did not expect it to be the last golden age. After the second-generation Core smart processor, Intel’s performance growth has slowed down.

From Sandy Bridge to Ivy Bridge, and then to Haswell, Intel’s performance improvement can only be described by a poor single-digit percentage. By the fifth generation of Broadwell, there has even been a retrogression, and the “cut meat” has become “toothpaste”.

Insufficient performance improvement, chip update speed is getting slower and slower. The 10nm process launched in 2015 was not realized as expected. It has been delayed until 2018. The original plan to achieve mass production in 2019 is also yellow. It can be said that it cannot be called out. At the same time, TSMC’s 5nm process chip has already begun. Large-scale shipments, this year is preparing to start the 4nm process.

Although due to the different standards of different companies, the process number is more of a kind of market promotion, but the latter’s strong update speed and promotion strength undoubtedly poke Intel’s pain points, making it difficult to argue.

At this year’s CES conference, Intel’s current CEO Ke Zaiqi announced that it will mass produce 10nm chips this year. He said that Intel’s “technology is advancing at an unprecedented speed, and Moore’s Law is the core of acceleration.” When Intel’s core manufacturing process has not been upgraded for 5 years, the second half of the sentence seems less convincing.

In addition, the direction of the market has also changed. Since 2011, global PC shipments have declined for seven consecutive years, directly threatening Intel’s core business. Intel CPU sales have been greatly affected. Compared with 2011, CPU sales in 2018 decreased by 30%, but its revenue did not decline. This is because Intel increased the price of a single chip.

The sales volume is not enough, and the price comes together, which reflects the embarrassing situation of the traditional business being stagnant and the new business unfounded.

A series of dilemmas faced by Intel are directly related to its failure in the mobile era.

In July 2019, when Intel announced that it would sell its 5G baseband business to Apple, more than ten years of mobilization attempts, the money and manpower it paid were all in vain.

The old IT that used to call the wind and rain was completely caught in an awkward position beside the right to speak. When the mobile chips of Qualcomm, Samsung, and Apple shined brightly, Intel, who was delayed in getting into the car, left a lonely back.


Second, the loss of mobile chips

In fact, Intel’s road to mobile chips started very early.

Return the time to 1997, when Intel acquired the Strong ARM architecture from DEC. The architecture was developed by DEC based on the ARM v4 instruction set and has low power consumption.

Since then, Intel has deliberately made up for its shortcomings in low-power processors. Just like its name, at the time, Strong ARM was more powerful than the public version of the ARM architecture.

In 2000, Intel introduced the Stong ARM series processor based on this architecture, which was the first mobile embedded chip developed by Intel. However, Intel’s chip designers are obviously not satisfied with using off-the-shelf architecture products, so in 2002, the self-developed XScale architecture CPU replaced StongARM.

Compared with ARM processors, XScale has lower power consumption, and has hardware acceleration capabilities for video decoding and 3D rendering as early as around 2004, while the public version of the ARM Cortex-A8 architecture will not be until six or seven years later. , It has a similar floating-point acceleration unit, and the performance of XScale is evident at the time.

Unfortunately, Intel failed to persist on this path to the end. In 2006, Intel sold its communications and application processor businesses, including XScale, to Marvell for a price of $600 million.

One year before that, Jobs had found Intel’s then CEO Otellini, hoping that Intel could provide chips for the first-generation iPhone, but he was declined. Looking back now, this may be the most costly bad decision in Intel’s history. However, from the perspective of the environment at the time, Intel’s choice of giving up mobile chips is hard to be criticized.

2006 can be regarded as Intel’s most embarrassing year since 1985. AMD, the millennium younger brother, jumped up with its super cost-effective performance. The CPU market share was once close to 50%, almost tied with Intel. , And three years ago, Intel still occupied more than 80% of the market.

AMD founder Sanders used to be the head of the sales department of Fairchild. After Noyce and Moore left Fairchild and founded Intel a year later, Sanders also left his old club, rolled up his sleeves and started to become independent Start a business. For more than forty years, Intel and AMD have been in love and killed each other. Perhaps it is a domineering Intel antitrust investigation, or perhaps it was born from the same root and has a constant fate. In short, Intel has never been The little brother rushed to kill him.

Whenever Intel launches a new product, AMD can always quickly make an excellent competitor with comparable performance and lower price. It is said that after Intel employees have finished their work, they will always ask AMD employees: “When will your home produce new products, we will have work after you come out.” AMD just