NVIDIA is considering buying arm? What an abacus!

Home > News content

NVIDIA is considering buying arm? What an abacus!

via:驱动之家 time:2020/7/24 12:50:17 readed:1231

In the early morning of July 23, Beijing time, Bloomberg announced that NVIDIA was acquiring arm, which attracted the attention of the semiconductor industry.

In 2016, Softbank, a Japanese consortium, acquired arm for $32 billion. At that time, Softbank was in a strong fashion. Sun Zhengyi made a big bet on artificial intelligence and IOT. The wholly-owned acquisition of arm was a huge investment of Softbank in the field of IOT.

However, in the past two years, as several companies invested heavily by Softbank are not operating as expected, Softbank is choosing strategic contraction. One of the important measures is to merge the IOT Department of arm into Softbank, and the remaining arm will be re listed or sold to prospective buyers.

It was at this time point that news of the acquisition negotiations between NVIDIA and arm came out.

Why is NVIDIA interested in arm?

Unlike Softbank, which has been declining since 2016, NVIDIA's share price has risen more than eight times since 2016. As a leader in the artificial intelligence field of semiconductor industry, NVIDIA is not short of money at present, so it is reasonable to consider high-level acquisition of some important targets.

Nvidia考虑收购Arm？打的什么算盘！

So why would NVIDIA consider buying arm? We believe that there are at least two motivations in terms of technology and market. First of all, although NVIDIA is far ahead in the field of cloud based artificial intelligence, its attempts in the mobile market over the years have not been successful enough.

Ten years ago, with the rise of smart phones, NVIDIA tried to enter the mobile device market with Tegra series. Unfortunately, Tegra was not recognized by many mobile device manufacturers, and eventually failed to become the mainstream. After the rise of artificial intelligence, NVIDIA tried to tap the terminal AI market by virtue of its advantages on the server side, so it launched Jetson series of mobile AI platforms.

Jetson series is still based on a lightweight GPU architecture, and on the mobile side is very demanding for efficiency and cost, so Jetson design is still too much

Another attempt of NVIDIA to enter mobile AI is open source nvdla, which is the IP of deep learning accelerator released by NVIDIA. It is not based on GPU architecture, but optimized for terminal deep learning.

NVDLA's performance figures are pretty, but they do n' t dominate the market and release time

On the contrary, no one can shake the dominant position of arm in the mobile SOC. At present, the mainstream mobile device SOC can not do without arm's IP authorization. Therefore, if NVIDIA wants to unify the cloud terminal AI market, it must be a reasonable consideration for NVIDIA to realize penetration of the terminal AI Market by acquiring arm.

Arm's products in mobile terminal include CPU, GPU and deep learning accelerator. Its IP and NVIDIA's CUDA Programming ecology in artificial intelligence also have very good complementarity. In addition to the mobile terminal, NVIDIA's other motivation to acquire arm may not be so obvious, but it is also very important, that is, the processor of cloud server.

At present, the mainstream processor of cloud server is still based on x86 architecture, but arm has a good momentum in the field of cloud server. Amazon earlier announced the AWS cloud server based on self-developed ARM CPU, while Huawei also released Taishan series servers based on ARM architecture and self-developed Kunpeng CPU. Its processor performance is comparable to Intel Xeon processor.

For NVIDIA, the current application in the cloud server is mainly based on the GPU accelerated by artificial intelligence, and it is indeed the world's leader in artificial intelligence ecology. However, if you want to make further breakthroughs in cloud servers, you need to go beyond the application of artificial intelligence in the cloud, or apply it to more fields in the field of cloud artificial intelligence.

Both require CPU

If NVIDIA wants to strengthen its market share in the cloud server field or consolidate its dominant position in the cloud artificial intelligence field, it will be a very good solution to acquire a CPU company and integrate it with its GPU ecology.

Nowadays, CUDA is already recognized as the standard ecology of artificial intelligence in the field of GPU, and if Nvidia can rely on its strength in the field of artificial intelligence, packaging GPU and CPU ecology to open the server market is also a good opportunity.

What are the variables in the acquisition road?

The main variable of arm acquisition is government approval. As the absolute overlord in the field of mobile IP, merging with other companies in a semiconductor industry will cause strict anti-monopoly review by governments.

Before that, it has also been reported that Samsung and apple are interested in acquiring arm. Apple and Samsung are mainly interested in mobile SOC. However, as apple and Samsung are also important players in the field of mobile SOC, it is difficult to obtain the approval of the government for the acquisition of arm. For example, once Apple acquires arm, it is very likely that arm will continue to provide IP to other competing mobile phone SOC manufacturers, which will cause antitrust problems.

At that time, Softbank successfully acquired arm mainly because it was not a semiconductor company, and it was in a neutral position in the semiconductor and mobile device market, so there was no antitrust problem; however, apple or Samsung were not.

Relatively speaking, NVIDIA is also a semiconductor company, but it does not have direct competition with arm's current customers. Therefore, the chances of passing the antitrust review will be higher than those of apple and Samsung. However, because both NVIDIA and arm are in the semiconductor industry, the antitrust review will still be a great challenge for NVIDIA to acquire arm.

Another possibility is that NVIDIA chooses to acquire only a part of arm's business, such as the server business. Since arm's server IP is not mainstream, there is little antitrust problem, and arm's customers in the server field (such as Amazon, Huawei, etc.) are also NVIDIA's customers. Therefore, such acquisition is of great significance to NVIDIA.

However, the main problem with such an acquisition is whether Softbank / arm agrees to spin off the server business, or whether they want the server business to remain in arm and sell it to other interested companies at a higher price, or go public again.

On the whole, NVIDIA's acquisition of arm is still in a stage with great room for change. However, we will wait and see.

Nvidia考虑收购Arm？打的什么算盘！