The spearhead is directed at Nvidia, the strongest chip war in history, is coming

Visual China.

Text |China Business Strategy.

In 1993, when the Koreans tried their best to defeat the hegemony of Japan's semiconductor industry, a young couple in Missouri, USA, bought an Apple computer for their 8-year-old son.

Thirty years later, the little boy shocked the world by launching a $7 trillion chip program.

7 trillion dollarsWhat is the concept?

It is an order of magnitude more than all the investments in the entire history of the chip industry combined, roughly twice the UK's GDP.

In fact, throwing the entire U.S. federal budget into it is not enough to fill the hole.

It was the 8-year-old boy who proposed this amazing plan, and now he has become a benchmark figure in the global technology industry since Steve JobsSam Altman

In just two years, this young man has used ChatGPT and SORA to rewrite the history of artificial intelligence twice, making the world exclaim:

The rate of iteration of AI horror.

The latest online SORA, from text to **, the demand for computing power is even more terrifying.

And that's just OpenAI. In recent years, artificial intelligence has become popular, and the demand for computing power has soared, but there are only a handful of manufacturers in the world that can meet this demand.

Nvidia alone accounts for more than 80%.

As a result, in the past year, Nvidia chips have become the most scarce commodity on the planet, so much so that Musk exclaimed: It's harder to get than drugs!

If there is a war going on in the field of AI, Nvidia is the most important arms dealer.

In the face of such a pattern, the vast supernatural powers such as Sam Ultraman are also aggrieved. He has been complaining privately and publicly:

The shortage of AI chips has hindered the progress of ChatGPT.

may think that if he does not break this pattern, he will not be able to achieve the ultimate goal in his heart, and even the entire AI industry-strong artificial intelligence (AGI), so he decided to fight and directly pointed the finger at NVIDIA.

In the past few months, Sam Altman has met with the big financiers of the United Arab Emirates, Japan's SoftBank Masayoshi Son, and the head of TSMC.

"We believe that the world needs more chip capacity than people are currently planning to build. ”

Sam Altman's vision is to raise money from the world's richest Middle Eastern billionaires, and then hand it over to TSMC, the most experienced company, to build and operate the factory.

In the end, $7 trillion was invested to build dozens of chip manufacturing plants.

The audacity of the plan dwarfed the scientific community's idea of building a particle accelerator orbiting the solar system, so much so that many thought that Sam Altman was crazy.

Including Huang Jenxun, the most daring founder of Nvidia in the AI industry, who was targeted, also exclaimed:

With this number, you can buy all the GPU chips in the world!

One wants to hold on to the advantage, the other is eager to break the monopoly pattern of Nvidia's monopoly, and this struggle that happened in the United States and Mars collided with the earth is just a microcosm.

Europe plans to invest $43 billion to expand chip production, and Japan is trying its best to win over TSMC and enter the advanced process, hoping to regain the glory of the former hegemon.

Masayoshi Son, the founder of SoftBank, is seeking $100 billion to fund a chip joint venture to compete with Nvidia for AI chip supremacy.

Intel also plans to start production of 1nm chips and is committed to creating a completely independent AI chip fab in the future.

As economic historian Chris Miller, author of The Chip Wars, puts it:

In the current great power wrestling, more and more countries are realizing that advanced chips (represented by AI) have become a strategic commodity.

Marked by American infighting,The next chip war has begun.

In the past half century, the global semiconductor industry has been iteratively upgrading in round after round of war.

In the 70s of the last century, with the efforts of several generations such as Shockley, the father of transistors, and Gordon Moore, the father of Moore's Law, the semiconductor industry in the United States developed by leaps and bounds, promoting the rapid growth of the American economy. With the strong support of the first, the Japanese gradually caught up with the United States by virtue of their cost and scale advantages.

By the mid-80s, represented by Toshiba, NEC, and Hitachi, Japanese manufacturers had more than half of the global market share, forcing Intel to withdraw from the memory chip market.

As the pioneers of the semiconductor industry, the Americans are naturally unwilling to fail, and on the one hand, they gather global forces to force Japan to sign a humiliating semiconductor agreement.

On the other hand,In South Korea and Taiwan, China will support new forces and transfer production capacity.

A man namedZhang ZhongmouAnd the Chinese, who led the semiconductor war of Texas Instruments, the global semiconductor giant at that time, smelled a business opportunity from it. In 1987, he went to Taiwan, China, founded TSMC, and opened a new chip foundry model, that is, only chip manufacturing, not chip design, which gave birth to two sub-industries: chip design and manufacturing.

But before that, most chip manufacturers designed and produced their own chips. There is even a saying circulating in the industry:A real man must have a wafer factory.

Affected by this, Zhang Zhongmou's TSMC was snubbed at first.

At the critical moment, the Americans vigorously supported it, and Intel and other large manufacturers threw out large orders as soon as possible, saving TSMC.

The ambitious Koreans are not willing to be left behind.

In 1989, taking advantage of the opportunity to go on a business trip to Taiwan, Lee Kun-hee, chairman of Samsung, invited Zhang Zhongmou to have breakfast and invited him to join Samsung.

After being rejected, Lee Kun-hee turned around and launched a tragic hand-to-hand battle against Japan, the world's chip hegemon, and made up his mind:

Take a gamble, develop advanced process technology, and squeeze into the world's table!

After two suicidal counter-cyclical investments in 1997 and 2008, the South Koreans finally had the last laugh. Since then, the global semiconductor industry has gradually evolved into a double star competition between Samsung and TSMC, and Intel, the former semiconductor overlord, has become increasingly lonely.

In particular, with the advent of the mobile Internet era, mobile phone chip manufacturers such as Qualcomm and MediaTek have higher and higher requirements for semiconductor processes.

From 65nm, to 28nm, to 10nm, 5nm, 2nm ......

Capital expenditures have also soared, with the cost of building a fab soaring from less than $100 million in the 80s to more than $20 billion today.

Coupled with the tragic cyclical fluctuations in the semiconductor industry, a large number of manufacturers could not keep up with the rhythm and were eliminated. At the beginning of the new century, the 90nm process had nearly 20 players around the world.

In the 5nm era, only Samsung and TSMC remain.

In this tragic chip war, the Americans seem to have suffered a big loss, and the transfer of production capacity in the 80s led to its later falling behind in advanced manufacturing processes.

But it is exactlyThis strategic retreat allows U.S. companies to focus on chip design.

In 2006, when TSMC and Samsung were fighting fiercely on the 65nm process, Huang Jenxun, the boss of Nvidia who was born as a graphics card, had a whimsical idea

The GPU needs to be programmable so that the user can call its computing power directly.

That year, NVIDIA introduced the CUBA programmable architecture, which paved the way for GPUs to be used in scientific research and AI.

Compared with CPUs, GPUs have a simple structure and are good at parallel computing, which is very suitable for the needs of AI, so it has become a hot commodity in many research institutions.

In 2012, Alexnet neural networks made a big splash in image recognition, triggering a new round of AI revolution.

Nvidia GPUs provided it with computing power, which caught Huang's attention. After more than a year of deliberation, late one Friday night, he sent an email to the whole company:

As of next Monday, we're no longer a graphics company, we're an AI company.

Three years later, NVIDIA, which is all in AI, invested heavily to build the world's first AI supercomputing DGX-1 and donated it to the fledgling OpenAI company at that time.

Sam Altman was grateful, but what he didn't expect was that OpenAI also suffered from Nvidia dependence.

It is not only OpenAI that has been put on the Nvidia chariot, 20 years have passed, the AI era is coming, and Intel, which was once the global semiconductor benchmark, has further fallen, but it has seen Intel fall lonelyThe United States, however, is still holding the scepter of chips.

In the past half century, the global semiconductor industry has fought a total of three major wars.

The first time wasPC era,Between the United States, Japan, Japan and South Korea, in the field of memory chips, a production capacity war has broken out.

The second time it wasIn the era of mobile Internet,TSMC and Samsung are fighting to the death around advanced manufacturing processes.

The third time wasAI era,The war between Sam Altman and others and Nvidia is brewing.

As the pioneer of the entire industry, the United States has not always had smooth sailing in these three chip wars, and has even been defeated in the memory and advanced process wars with Japan and South Korea.

But it was,Time and time again, it has led the way of the tide.

In 1947, the world's first transistor was invented by American scientist Shockley and others, which opened the prelude to the semiconductor industry.

Later, Kilby and Noyce invented the integrated circuit, which pushed the semiconductor industry to the era of Moore's Law.

In the era of mobile Internet, companies such as Motorola and Qualcomm took the lead in launching mobile processors, allowing the smartphone industry to prosper.

In the era of AI, Nvidia was born and once again took the helm of the wave.

Round by round, how did the Americans do it? Some people say it's capital, some people say it's the first-mover advantage that has a leading technology base, and some people say it's the support of the best people, these may be reasons, but they are not important reasons.

The most important reason is that generations of geniuses, successive efforts, and freedom to fly.

These people all have one characteristic:

With an almost crazy dream of changing the world, and crazy investment regardless of utilitarian gains and losses.

Shockley's invention of the transistor and Kilby's invention of the integrated circuit were initially driven not to make money, but to look up to the stars and change the world.

In August 2016, when Huang donated the world's first AI supercomputer to OpenAI, he took out a marker pen and excitedly wrote a sentence on the chassis:

"For the sake of computing and the future of humanity, I donate the world's first DGX-1. ”

Sam Altman, the founder of OpenAI, has always compared himself to Oppenheimer during his world tour.

"Every day I work at OpenAI reminds me that the scale and speed of socio-economic change is faster than most people think," he said. ”

Thinking about the future, Sam Altman once worried that Google's artificial intelligence would become the supreme leader, which is why he and Musk joined forces to found OpenAI, their goal is:

Build a friendly AI that will forever change the way humans interact with computers.

However, the dream of changing the world is not only fanciful, but also requires bold bets, regardless of the cost, and sometimes faces huge criticism.

In 2013, when Nvidia was all in AI, artificial intelligence was not yet a hot technology.

Brian Catanzaro, Nvidia's vice president of deep learning, later recalled that he had repeatedly persuaded Huang not to fall into the traps that the AI industry had encountered in the past.

But, ten years later, he was right. ”For the vision of the owner, Catanzaro is full of praise.

Similarly, in 2015, when Sam Altman founded OpenAI, not many people were optimistic. But he is the kind of person who is less optimistic about the outside world, the more ferocious he is.

A well-known investor once commented on him:

"You can parachute him into an island full of cannibals and come back in five years and he will be king. ”

In the end, Sam Altman made the right bet, and he became another iconic figure in the global technology industry after Steve Jobs.

In this regard, Microsoft CEO Nadella sighed:

"He's an incredible entrepreneur, he's made big bets, he's got it right, and OpenAI's success is hard to argue. ”

Today, witness the AI storm sweeping the world across the ocean, from ChatGPT and SORA to the $7 trillion plan ......We should have the confidence to follow our own path, but we should also think:

How to let more talents and dreams grow more horizontally.