news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

The giant's Achilles' heel.

Text丨Huashang Taolue

On August 2, US time, Intel’s stock price plummeted by more than 26%, and its market value fell below the historical mark of US$100 billion.

In 2000, Intel stood proudly at the top of the technology wave with a market value of $300 billion. More than 20 years later, Intel's former younger brothers, TSMC with a market value of $50 billion, has a market value of over $1 trillion, Nvidia with a market value of $5 billion, has a market value of over $3 trillion, and AMD, which Intel once couldn't even touch, has a market value of over $200 billion...

A generation of overlord, most of his body is buried under the loess.

【The Worst Overlord】

On August 2, Beijing time, Intel released its second quarter financial report, with revenue of US$12.8 billion and a net loss of US$1.6 billion. It has been in the red for two consecutive quarters this year.

Although Intel CEO Pat Gelsinger said that the company has achieved key milestones in product and process technology, "We are taking decisive actions to improve operational and capital efficiency by leveraging new operating models... Together with the launch of Intel 18A next year, we will regain our leadership in process technology, strengthen our market position, and improve profitability." The capital market still chose to slap it in the face. On the day the results were announced, the company's stock price fell 5.5% to $29.05 per share, and the next day it fell more than 25% at one point, and the company's market value officially fell below $100 billion.

On March 5 this year, Intel announced that it had received the world's first new-generation high-NA EUV lithography machine from ASML.

As an essential "weapon" for mass production of advanced process chips below 2nm, its price exceeds $300 million. Intel, which has spent a huge amount of money, has revealed to the world its ambition to catch up with TSMC, the world's "foundry king".

But at that time, public opinion no longer favored its efforts, believing thatFor Intel, which has fallen to the tenth-last place in the global wafer fab rankings, it is an "impossible task" to make a comeback.

In 1971, Intel, which had just been established for three years, developed the first commercial processor, the Intel 4004, which marked the beginning of the world's computer and Internet revolution.

Since then, from memory to personal computers and server CPUs, from chip design and manufacturing, Intel has long been the absolute overlord of semiconductors in the United States and even the world.

At the end of the last century, the "Wintel" alliance formed with Microsoft Systems created a wave of the global PC era. In 2000, Intel's market value exceeded $300 billion, and its glory was greater than that of Nvidia today.

But after reaching its peak, it was beaten almost everywhere and fell into an unstoppable decline in the past decade.

In 2017, an explosive news that "Samsung's performance surpassed Intel and became the new semiconductor overlord" swept the industry.

As the king of memory chips that focus on lower profit margins and greater price fluctuations, Samsung has long been suppressed by Intel, which has a high share and high profits in key markets such as personal computers and servers.

Samsung prudently declined to comment on the sudden accolade. Intel, on the other hand, said:“We are very pleased with the strategy and performance.”

The fact is that it was Intel's last act of stubbornness: the original 10nm chip plan was postponed, which was extremely fatal in the semiconductor industry where technology is king. However, its two biggest rivals in the CPU and foundry markets, AMD and TSMC, have made breakthroughs.

Due to the difficulty in developing technology and the rise of competitors, Intel has become particularly passive.

From 2018 to 2019, Intel's revenue and profits hit new highs, but because its growth rate lagged far behind that of its competitors and its market share was constantly eroded, it was gradually overtaken by its competitors.

In July 2020, Intel, which had already been left far behind by its competitors, fell further into the abyss.

The disaster came from the helpless statement made by then CEO Bob Swan in a conference call: the company's future chip manufacturing plants (wafer fabs) may never catch up with the progress, and it may even have to consider using contractors to manufacture 7nm chips.

Although he tried to remain calm, every analyst present could sense that he was stuttering when he spoke. After all, this might be the most radical "bad news" in Intel's 52-year history - publicly admitting that it had fallen from the pinnacle of high-end chip manufacturing.

Some American media even exclaimed that Intel's "amazing failure" foreshadowed the end of the American chip era. Overnight, Intel's market value evaporated by nearly 300 billion yuan.

In February 2021, Intel fired Swan, the CEO who had just taken office for two years, in despair. The person who was appointed in the crisis was Intel's eighth CEO Pat Gelsinger.

In 1985, Andy S. Grove, one of the founders of Intel, approached Kissinger, who was pursuing a doctorate, and said, “You can fly a simulator here, or stay at Intel and fly a real plane.”

More than 30 years later, Kissinger became the "captain" of Intel, but his mission was to save this overlord that once rewrote and led history from being completely abandoned by the times.

He ambitiously declared:

Intel is back!

As a veteran who has worked at Intel for nearly three decades, Kissinger knows that "mistakes made over more than ten years cannot be solved overnight."

But he still tried his best to make up for it and announced the IDM2.0 strategy of running the design and manufacturing departments separately, that is, to use a more flexible wafer foundry model, more aggressive spending expansion and technological innovation to seize the lost market share and even return to the top.

To this end, Intel took the initiative to divest its secondary businesses and focus on design and manufacturing. In March 2021, it invested $20 billion to build two new wafer factories in Arizona, USA; a year later, it invested more than $20 billion to build two new factories in Ohio.

In July 2021, Intel announced its “Four Years, Five Process Nodes” plan, which is to advance five process nodes, namely Intel 7, Intel 4, Intel 3, Intel 20A and Intel 18A (1.8nm), in the next four years and regain process leadership in 2025.

In a conference call on Wall Street, Kissinger said bluntly:“Great companies are able to recover from difficult and challenging times and emerge stronger, better and more capable than ever before.”

A group of industry analysts, however, poured cold water on the new roadmap: "The new roadmap is very radical, but we don't know how it will be achieved." Intel's continued miserable performance is a strong basis for them to remain pessimistic.

In July 2022, Intel's Q2 financial report showed total revenue of US$15.321 billion, a 22% drop. It was the first net loss of US$454 million in 10 years, which can be described as "the worst performance in history."

In the first quarter of 2023, Intel's net profit loss was US$2.76 billion, the largest loss in history.

On January 26, 2024, Intel released its financial report showing that its revenue in 2023 was US$54.2 billion, a year-on-year decrease of 14%; its net profit was US$1.7 billion, a 79% year-on-year cut. Upon hearing the news, its stock price fell another 10%.

Now, another half year has passed, and two consecutive quarters of losses have occurred. Even Kissinger, who is extremely eager to turn things around, can only face reality helplessly:If Intel wants to return to the forefront of the industry, it still has a long way to go.

[Defeat]

In the 1960s, Gordon Moore, one of the founders of Intel, proposed the famous Moore's Law: when the price remains unchanged, the number of transistors that can be accommodated on an integrated circuit will double every 18 to 24 months, and the performance will also double.

From then on, Intel became the most loyal defender of the law:I firmly believe that as long as I follow this "mantra" perfectly, I will be invincible.

Even during the industry recession, Intel has always kept up with Moore's Law by investing more than 10% in research and development. As a result, it has continuously surpassed its products and markets, and became the world's largest semiconductor company in 1992.

From 2005 to 2007, Intel launched a unique Tick-Tock production model, which uses two-year rotation units. The "tick year" focuses on chip manufacturing, updating chip processes and improving technology; the "tock year" focuses on chip design and architectural innovation.

Under this framework, as long as chip performance and manufacturing processes continue to improve, Intel will not waste the performance of the previous generation of processors due to its too fast R&D pace; the large investment in wafer fabs can also be effectively shared by the huge production capacity demands.

As a result, Intel can gain more market and financial advantages over its competitors at a lower cost, further promoting the continued growth of higher-performance chip research and development and production capacity expansion.

As a result, Intel won a golden decade without wasting any opportunities or costs, and its R&D from 45nm, 32nm to 22nm and 14nm was a natural progression, leading the way. Samsung, TSMC, AMD and other competitors could only follow closely behind.

This huge success makes Intel one of the few giants qualified to stick to the IDM model:The core links of semiconductor production, including design, manufacturing, packaging, testing and sales, will be taken care of by one person.

At first, global semiconductor giants such as Intel, IBM, and Fujitsu all adopted the IDM model of "designing and manufacturing by themselves".

However, the cost of developing and producing advanced process technologies is too high, and the large companies that are losing market share have gradually stopped investing in fabs. The number of IDM global players has dropped sharply from 30 at 130nm to less than 5.

However, the IDM model has obvious advantages. Each link of the chip can be reasonably configured, and smooth internal communication can reduce transaction costs, significantly shorten product launch time, and thus obtain higher profit margins.

Intel was able to dominate the market with this model and stand out from the crowd with huge profits of over 60% gross profit margin.

However, in the era when Intel was in a winner-takes-all situation, a force that could subvert it was secretly growing.

In 1987, Andy Grove, who had a close personal relationship with Morris Chang, had just taken office as CEO of Intel. He changed the fate of TSMC, which was founded in the same year, with just one sentence: Perhaps Intel can use you.

After that, TSMC, which took over some of Intel's low-end chip production capacity, tasted the sweetness of foundry. But Morris Chang, who was not satisfied with the "leftovers" of IDM manufacturers, made a decision that had a far-reaching impact on the world's semiconductor industry chain: to become a professional wafer foundry.

At first, TSMC wanted to get back on Intel's good side, but was rejected for financing. Even Moore personally advised Morris Chang: "You have had many good ideas, but this one is not good enough."

But this time, both Moore and Intel were wrong.

After 1991, with the arrival of the semiconductor boom cycle and the rise of the computer industry, the low-threshold chip design industry has seen a surge in new blood. TSMC has become the biggest support for these start-ups that have no funds or manpower to build wafer fabs.

By 1994, TSMC, which had successfully gone public, had 137 customers and a market value of NT$120 billion.

This is a watershed in the history of semiconductors. The IDM model that Intel adhered to was constantly defeated by TSMC's professional wafer foundry model. Many competitors that were ruthlessly strangled by Intel under the original model were able to take a breath and even rise secretly.

In 1995, Huang Renxun, who had been in business for two years, had no money to build a wafer fab, so he had to write to TSMC for help. Soon, Morris Chang called back and agreed to do foundry work, just like Intel changed itself, and changed the fate of Nvidia.

With the help of TSMC's professional and reliable foundry model, the fledgling NVIDIA was able to quickly occupy the market segment. After NVIDIA proposed the concept of GPU, Huang Renxun emphasized:There is no Plan B, all pressure is on TSMC.

Sure enough, with the help of TSMC, Nvidia became the world's leading graphics chip manufacturer. Huang Renxun later said thankfully: "If I had built my own wafer fab, I might only be a CEO with tens of millions of dollars in revenue."

Similar stories also happened to Qualcomm and AMD.

In 1996, mobile phones began to spread around the world. Qualcomm, which suffered from the huge R&D investment in wireless communication technology, naturally found TSMC to outsource the technology, which greatly alleviated the financial pressure and enabled it to focus on CDMA technology and rise to become an industry giant.

AMD founder Jerry Sanders initially stubbornly believed that "only a real man can own a wafer fab." Even though Morris Chang said, "My production cost is half of Intel's, and the quality is twice as good as Intel's," he was not convinced to do OEM.

But in this old model, AMD, which was lagging behind in all aspects, was no match for Intel. After Intel launched the Tick-Tock model, AMD, which lagged behind in both product design and advanced manufacturing processes, was unable to cope and was once pushed to the brink of bankruptcy.

In desperation, AMD divested its wafer fab in 2009, abandoned the IDM model, and moved towards a path of focusing on design. In 2014, AMD welcomed a Chinese-American female CEO, Lisa Su, who further gambled on design, focusing on self-developed new Zen architecture, and relying on TSMC's continuous breakthrough process technology to catch up with Intel at full speed.

As a result, Intel has finally found itself in a two-front battle with the world's top chip design companies and top manufacturing companies, and is being pushed into the abyss step by step by its opponents.

In terms of manufacturing, in 2014, Intel announced that the mass production of the 14nm process would be postponed again. The outside world felt that the Tick-Tock strategy was beginning to fail; TSMC launched the "Nighthawk Project", bringing together nearly 400 researchers, working in three shifts 24 hours a day, to rush to 10nm.

In 2017, Intel's plan to break through 10nm was "difficult to implement", and the 14nm+ process that was barely released was ridiculed as "squeezing toothpaste". The Tick-Tock model that had been running for ten years was declared "bankrupt". TSMC was the first to achieve 7nm mass production in 2018.

In terms of design, AMD has taken the industry by storm with its Ryzen series CPUs, which have faster processing speeds and are less than half the price of Intel's.

In 2018, after TSMC's 7nm mass production, AMD was even more determined to transfer its production capacity to it, and Intel fell further into the vicious circle of the IDM model: when its own factories cannot achieve more advanced processes, it will hinder design innovation, and backward design will lead to market and efficiency lag, and drag down the company's investment, making it even more impossible to achieve advanced processes.

In July 2020, Intel again announced the postponement of the 7nm process, while the 5nm chips produced by TSMC have been launched together with the iPhone 12 and Huawei Mate 40, and 3nm chips are also on the agenda.

At this point, TSMC's manufacturing has become Intel's nightmare of "never being able to catch up". AMD, which took off on TSMC's shoulders, has continued to make a gorgeous comeback and surpassed Intel's market value for the first time in February 2022. Now, with more than 20% of the global PC and server market share, it is rapidly eroding Intel's future.

In this sense, Intel's failure is a failure of model.It was a failure in its attempt to compete with global rivals in the two major fields of design and manufacturing on its own, a failure in wanting to do everything but ultimately doing nothing well.

But in addition to the model, Intel also has bigger reasons for its failure.

Losing the Future

Grove, who led Intel to the top, once pointed out in "Only the Paranoid Survive" that being able to identify the direction of the wind and avoid sinking is crucial to the future of an enterprise.

But after him, Intel missed the future "wind direction" time and time again.

At the 2006 Mac World Conference, a classic scene in the history of technology appeared:

Intel CEO Paul Otellini, wearing a semiconductor cleanroom suit, handed a large silicon wafer to Steve Jobs and excitedly said, "Intel is ready."

Over the past decade, Intel has dominated the vast majority of the world's personal computer CPU market, and Apple is the last unconquered "virgin land." Otellini, who was recognized by "Silicon Valley idol" Steve Jobs, won a huge victory for Intel.

Since then, the two sides have become close allies, but Otellini won the beginning but lost the end.

When the first generation of iPhone was just being conceived, Jobs personally approached Otellini and hoped that Intel would manufacture chips for Apple smartphones. However, this time Otellini made a historic mistake that will go down in history.

He carefully organized internal experts to analyze and discuss the possibility of manufacturing chips for the iPhone. The experts unanimously agreed that the order price given by Jobs was too low and the quantity was limited, making it an unprofitable deal.

Otellini listened to the advice of these professionals and said no to Jobs.

The experts were right. At the time, Intel had already occupied an absolute share of the PC and server X86 chip market, and its profits were rolling in like a fully operational "money printing press".

However, the emerging mobile chip field is dominated by devices such as smartphones and tablets, which can be met by ARM architecture products with lower performance and power consumption. It is difficult for Intel to create high added value, and if it invests heavily, it may not be worth the cost.

Intel even took the initiative to sell its XScale business that produces ARM architecture products, and developed a low-power x86 chip called Atom to target ARM products from entering the notebook computer market and further defend its advantages in the computer field.

Simply put,Intel chose to focus on defending the computer market and showed little interest in the mobile market.

An opportunity to win the next wave of industry competition was thus ruined by Intel itself.

In the end, Apple's first-generation iPhone sales exceeded initial estimates by more than 100 times, and Apple's revenue from smartphones quickly surpassed Intel's a few years later.

Betting on the wrong future not only made Intel lose the biggest ticket to the mobile Internet era, but also made it die in the icy cave of the continued decline of personal computers. Otellini regretted rejecting Jobs in the past, saying:

"The biggest regret in my career."

But this may not be Intel's most painful lesson.

In 2006, NVIDIA, relying on the trend of GPU, has begun to show its potential to change the future.

That year, AMD acquired ATi, the second largest GPU company on the market, and NVIDIA immediately launched the Tesla GPU based on the CUDA architecture, whose performance was comparable to the "nuclear weapons" in the computing field.

Nvidia's popularity has made Intel wary. In order to curb Nvidia's progress, Intel not only interrupted the integrated graphics cooperation between the two parties and launched a general-purpose central processing unit (GPCPU), but also announced a major decision: launching the Larrabee project.

The project is said to cost billions of dollars and is a rare attempt by Intel to reach the peak of GPU technology. Its leader is Pat Gelsinger, Intel's first CTO and current CEO who led the development of flagship products such as the 80486 processor.

But when the impatient Intel encountered setbacks in product development and could not be implemented as scheduled, it resolutely "cut off" the project because it could not bear the fact that the investment was like a black hole and that even if it succeeded, there would not be much prospect.

Even Kissinger, who insisted on the project, was kicked out of the company.

After canceling the Larrabee project, Intel once had a chance to fight back: it decided to follow AMD's example and buy Nvidia.

But Intel once again missed this opportunity to change its destiny.

then,Nvidia's market value is only about 10 billion US dollars. For Intel, the price is not a problem in acquiring Nvidia, but it is said that the company's management is arguing over "what position should be given to Huang Renxun" and cannot reach a consensus.The acquisition ultimately fell through.

Today, Kissinger, who has returned to the helm of Intel, still feels resentful about the past from time to time. "When I was kicked out of Intel, they killed this project that would change the landscape of artificial intelligence."

Facing Nvidia, whose market value has exceeded 3 trillion US dollars, Kissinger still refuses to accept it and even believes:"Nvidia's success in the artificial intelligence industry was purely accidental," Huang Renxun said.

But Bryan Catanzaro, an Nvidia executive who worked on the Larrabee project, disagrees with Kissinger. In his view, "Nvidia's dominance does not come from luck. It comes fromVision and execution, this is exactly what Intel lacks.”

【References】

[1] “Intel’s Revelation: Lost to TSMC, Lost to Themselves, and Lost to the Times” Jinduan

[2] Chip Wars, Huake Press

[3] Chip Wave, Publishing House of Electronics Industry

Welcome to follow [Huashang Taolue] to get to know the influential people and read the legendary strategies.