Shenhao started as a military training anchor

NVIDIA's release of the "most powerful chip on earth" B200 has caused the company's market value to skyrocket by two trillion.

Hawley did not deny the B200’s super computing power, nor did he deny its beneficial effects on large model training.

What he denies is that B200 is not a revolutionary technological product.

The negative is that such high-computing power and high-priced chips cannot promote the development and popularization of the AI ​​industry.

On the contrary, it may put Nvidia in trouble.

In today's press conference, Holly first revealed the useful information of the "Supercomputing Center" on the supply of training AI content, which is already ahead of the United States.

Then throw out the direction of domestic chips that Everest Technology is researching, and question Nvidia’s most powerful chip on the surface.

It can be said that the topic has reached its peak.

Now everyone is looking forward to Holly's explanation. Is there a chip in China that is stronger than B200?
Holly flipped through the PPT and continued, "Comparing the H100 released by NVIDIA, you will find that the B200 has not made progress, nor is it a product of technological innovation.

The algorithm has not changed, the architecture has not changed, it is just that the chip area is larger and more computing units are used. Simply put, it is doing addition.

Since the development of GPU, it has been doing addition to improve computing power.

Including the GB200 released by NVIDIA, two B200 chips are stacked together to realize shared memory through supercomputers.

This kind of stacked chip that only pursues computing power cannot promote the revolutionary evolution of AI, and may even become a constraint on the development of AI.

The reason is simple, it is very expensive and consumes more energy. "

Holly flipped through the PPT: "How expensive is it?

Let's compare the price of H100. A single chip costs close to 50000 US dollars, and G200 will only be higher.

Based on the current demand for chips from these trillion-level models in Silicon Valley, the purchase volume is more than several million chips.

This means that these technology giants in Silicon Valley will spend tens of billions or even hundreds of billions of dollars to purchase Nvidia chips.

Both OpenAI and Meta are currently losing tens of billions of dollars in the AI ​​field every year. Such expensive chips will make their losses even greater. This is not a healthy investment or a healthy industry.

Small companies cannot afford it, and large companies all work for Nvidia. The final result is that things will turn against each other and the industry will collapse.

In an industry chain, if only one company is making money and it is cutting leeks fiercely, the development of this industry chain will be out of balance.

That’s why I say that NVIDIA’s GPU, which claims to have the most powerful computing power on the planet, will not promote the development of AI. "

Reporters at the venue listened and recorded and reported in real time.

This is the fourth disruptive point of view proposed by Hawley in the field of AI.

The unabashed criticism of Nvidia's appearance is definitely big news in the international technology circle.

Holly continued: "Let's talk about energy consumption. The increase in computing power has brought about an increase in absolute energy consumption.

The power consumption of the B200 has reached 1000 watts. We all know that AI is a power-consuming industry. If we blindly pursue ultra-high computing power, our energy will definitely not be enough.

The more GPUs are used, the more energy on the earth may have been exhausted before AI is used and popularized by billions of people around the world.

We train large AI models to empower energy technology and solve energy problems, rather than endlessly consuming energy.

From these two perspectives, the G200, the most powerful chip on earth released by Nvidia, is not the best choice for the development of the AI ​​industry.

What Everest Technology Semiconductor Research Institute and Huawei Semiconductor Research Institute are doing is to develop and produce a chip with low energy consumption and low price.

The name of this chip, we call it Tai Chi, is taken from "The Book of Changes". "

The big screen flashed, and the image of the Tai Chi chip appeared.

The venue was filled with deafening applause and exclamations.

This is the most powerful moment of the entire press conference today.

The domestic chip "Tai Chi" has entered the research and development stage.

And it is trying to break Nvidia’s monopoly in the field of AI chips.

Holly didn't mention a word about the computing power and performance of the "Tai Chi chip". In fact, he was taking the route of overtaking in a corner. The competition between Everest Technology’s “Tai Chi Chip” and NVIDIA’s most powerful computing chip is a bit like the competition between ARM and Intel in the CPU field.

Intel followed the same path as Nvidia back then. In terms of CPU research and development, it continued to build multiple cores and high frequencies, relying on the superposition of computing power to achieve strong performance.

But ARM did the opposite, making low power consumption and light computing power, and used it in various mobile devices and Internet of Things devices. In the end, it became Intel's strongest opponent, and ARM processors even had the momentum to replace Intel. .

Compared with chip technology, computing power, etc., domestic semiconductors will definitely not be able to catch up with Nvidia.

However, if a chip with half or even a quarter of the computing power of NVIDIA can be made, but the price is less than one-tenth of NVIDIA and the energy consumption is much lower than NVIDIA, this chip will quickly rush to the market. Break through Nvidia’s monopoly.

With reduced computing power, it will definitely be more difficult to catch up with international AI giants like OpenAI and meta.

But there is a way to overtake in a corner, which is to separate the inference chip and AI training.

Everest Technology's "Tai Chi Chip" is an inference chip, and the massive learning content of the Everest Supercomputing Center is an AI training.

The two work together.

Finally achieve superiority over the opponent.

It can be said that Hawley is now challenging the entire Silicon Valley.

Of course, it is also inseparable from the support of Huawei and other domestic semiconductor peers, as well as the support of domestic Internet giants.

When everyone at the venue was still discussing the "Tai Chi chip", Holly directly shouted to the giants in Silicon Valley, "If the technology companies in Silicon Valley cannot afford to purchase Nvidia chips, they can consider our Tai Chi chips.

We will not use restrictions and sanctions. "

The taunt value is full.

When he said these words in front of many international media, Holly's main motive was not to advertise the "Tai Chi Chip" and prove how powerful the Tai Chi Chip was, but to create fire.

Let the technology giants in Silicon Valley fight among themselves to get rid of Nvidia's monopoly on GPUs.

He sat on the mountain and watched the tigers fight, developing secretly.

"Today's press conference ends here. The last thing I want to say is that no matter the doubts from the outside world or the pressure from opponents, they will not defeat our generation of technology people.

The road is long and far away, and I will search up and down.

There is a time to regret. "

Holly finally left the venue with a handsome figure behind him.

This press conference became a hot topic of discussion on the entire Internet and even in the global technology circle that day.

"Tai Chi chip, it turns out that Holly is doing big things quietly, I'm really looking forward to it."

"GPU localization will be realized by scientists of Holly's generation."

"After watching the entire press conference, what I want to say is that Holly is awesome."

"It's full of useful information. Is Mr. Huo planning to challenge the entire Silicon Valley?"

"Those black guys came out to talk, are they a slap in the face?"

"I already have a premonition that domestic AI development will accelerate."

"..."

The impact of this press conference also spread to the stock market. Domestic semiconductor companies surged. At the same time, NVIDIA plummeted by 1200 billion U.S. dollars overnight, with losses exceeding one trillion.

When the most powerful chip on earth was first released, the market value skyrocketed by two trillion overnight, and now it has plummeted by one trillion, like a roller coaster.

Huang Renxun never expected that a press conference held by Holly would have such a huge impact.

In fact, the real cause of Nvidia’s stock market crash was Hawley’s provocation to Silicon Valley technology giants.

Let them also feel the crisis caused by Nvidia's monopoly on chips, so the technology giants in Silicon Valley began to exert their efforts.

(End of this chapter)