As the three giants of computing power, NVIDIA, AMD and Intel (there is a reason for this arrangement), during the Spring Festival of the Year of the Tiger, the topic of acquisition seems to be the keynote that will affect the competition pattern in the coming year, and it is the cause of considerable changes in computing power competition in the future. Because just last week, the three companies happened to have various news events about the acquisition, so we took this opportunity to have a little gossip.
NVIDIA: Missing an arm is not necessarily a bad thing.
The most shocking news last week must be that NVIDIA's acquisition of arm failed due to obstruction from all sides, which can almost be said to be a deal that may affect one third of the semiconductor industry. According to various sources, quite a number of regional market management agencies failed to approve the transaction, which was the main reason for the failure. NVIDIA exchanged US$ 654.38+02 billion for the so-called 20-year license. Many people think that NVIDIA has lost money. Financially, NVIDIA's acquisition will not only lose 65.438+0.2 billion yuan. After all, the various third-party evaluation fees acquired before will inevitably hit Shui Piao, but it may not be a bad thing for NVIDIA as a whole. After all, the longer the graphics card price is firm, the better the financial report will be.
First, the value of arm is not worth $40 billion. After Softbank acquired arm, it did some subtraction for arm, some services did not participate in the transaction, and China District was partially cut off in the form of a joint venture. If such an arm can be sold at a premium (Softbank only bought $37 billion), considering the poor economic situation of Softbank and the urgency of continuing to cash out, NVIDIA is suspected of taking over. Excluding any inappropriate behind-the-scenes trading speculation, we can only think that since NVIDIA's share price soared, it seems that it really wants to take the opportunity to unify its computing power. After all, look at your two rivals, both of which are fully configured with CPU+GPU+FPGA. Obviously, you can't rely on GPU to maintain the high market value of the first semiconductor share, but winning arm will undoubtedly make you a brand-new ecological "king". It is said that Qualcomm is a patent hooligan. If NVIDIA really wins arm, I'm afraid half of the patents related to computing can't escape this company.
A question worth thinking about is, what happened in the past year and a half from NVIDIA's agreement to buy arm to its final rejection? Arm's revenue has not increased much, but RISCV is becoming more and more popular. Thanks to the uncertainty of many manufacturers after NVIDIA's acquisition of arm (who knows whether the acquired arm will be another MIPS), many computing-related enterprises began to take RISCV as their future plan B, which indirectly caused the explosion of computing equipment based on RISCV kernel. If the transaction continues, this trend will be more obvious. Now that arm continues to be independent, if no one continues to apply for the acquisition of arm in the short term and chooses arm to go public, users' confidence in arm's future independence may be rebuilt, so RISCV may return to its original plan B role. As for Intel or other companies that want to acquire arm, it is estimated that it will be difficult to get regulators in various regions in the short term. Softbank can only take money to promote the listing of arm, so what arm needs to do in the short term is to strongly suppress the RISCV camp, otherwise the valuation will be greatly reduced.
As for NVIDIA, at least the reaction of the stock market shows that investors are cautiously optimistic about the future result, indicating that they are not sure that the acquisition of arm will be a good thing for Invista, but NVIDIA can only move forward on a single core in the short term, and it is still difficult to squeeze into the server market. Of course, it is a pity that Lao Huang imagined a bright future. ...
AMD: Successfully acquired Xilinx, surpassing old rivals.
AMD, NVIDIA's biggest competitor, had a good life in the past two years. Although not as dazzling as NVIDIA, the market value has also increased, and the market share of CPU has increased. The GPU market has benefited from the step-by-step increase in the profit rate of graphics cards out of stock. In addition, the acquisition of Xilinx progressed smoothly, and the transaction was finally completed this Spring Festival. In addition to tying the old rival Intel in processor composition, at the moment of the transaction, the total market value of the two companies surpassed Intel, which was a surprise. You know, at the worst time four or five years ago, AMD's market value was probably less than one twentieth of Intel's. Now if you want to ask AMD or Intel which company is bigger, I'm afraid people have to turn on the financial channel to see the market value before they can answer you.
The acquisition of Xilinx has almost no disadvantages for AMD. Financially, there is not much financial cost in choosing equity swap; Technically, after the introduction of FPPGA, AMD has a more comprehensive pool of computing technologies, which greatly enhances the AI technology reserve capacity (if we take NVIDIA as the AI strength of 100, Intel can probably reach 85 points, while AMD before the acquisition was less than 50 points, and now it can reach the level of 70+); From the market point of view, after Xilinx's FPGA joined, AMD instantly expanded its competitiveness in the server market, and it can also penetrate into many market segments such as industry, medical care, national defense and aerospace. Previously, both companies had to face the joint competition of Intel's CPU and FPGA in terms of servers and core computing equipment. Now, if we evaluate the server-side strength of both parties, it is difficult to distinguish the obvious strength instantly. The only uncertainty is whether AMD can better integrate FPGA technology with the server ecosystem that was slightly lacking before, and how to face the ecosystem construction that Intel is best at.
Intel: IDM 2.0 is very interesting
Intel quietly announced the completion of the acquisition of the semiconductor foundry company Tower. The acquisition of 7 billion yuan is not too expensive, even considering the sharp increase in wafer factory profits caused by shortage, which is already a high premium for Tower with low annual income. However, we can also see that Intel's strategic direction is firm. From IDM2.0 to later OEM service strategy, to later open X86 architecture, and now to the acquisition of Tower Jazz, Intel has continuously enriched its technical strength in semiconductor manufacturing. If Intel seems to lack advantages over its two major computing rivals in terms of products, then manufacturing is likely to be the key link for Intel to suppress its competitors now.
For Intel, catching up with Samsung or even TSMC is an important means to regain the performance advantage of CPU. However, to catch up with its competitors, it needs enough R&D investment, which means not only profiting from its own products, but also sharing the R&D cost as much as possible from OEM, otherwise the sky-high process R&D cost will only drag itself down. Therefore, the acquisition of Tower is of great strategic significance for Intel to continue to implement its IDM2.0 and OEM business strategies.
First of all, although Tower is small, it is a standard foundry with a perfect OEM service system. As a perennial IDM, what Intel lacks most is OEM service system. Buying is much easier than building your own team. What's more, with Tower, Intel will naturally enter the OEM circle and can provide more OEM services, not just on the production line before Tower. We analyzed the current OEM market, and found that Tower is the largest OEM that Intel can buy at present, except for AMD's GF. GF is now in the hands of the Middle East, while Tower is controlled by Israeli capital. For Intel, Israel is basically the second largest R&D and production base outside the United States. At this point, Jewish genes ... you know.
Secondly, in the past two years, Intel has been striving for financial support from various fabs in the United States, and has also obtained semiconductor construction funds from Israel, and hopes to get more support from TSMC from the US government. One feature of government support is that it is hoped that the fabs can meet the national defense standards of the United States and serve them, which Intel will naturally not let go easily. Tower has two fabs in the United States, and Intel can apply for national defense certification for these two fabs after acquisition. More importantly, although Tower is small, it is also the largest simulated wafer foundry, while Intel lacks simulation experience and has no simulated wafer foundry. Most defense applications are mainly analog and mixed signals. Winning Tower will enable Intel to quickly have a relatively complete analog chip production capacity, which is very meaningful for obtaining more financial support from the US government in the future.
It is not difficult to see from these two points that it is of great strategic significance for Intel to quietly acquire Tower. On the one hand, Intel has directly entered the role of foundry, and can quickly transform the original foundry with a mature foundry service system. On the other hand, Intel has become the most powerful analog foundry in the United States at present, and may even build a 12 inch analog foundry in the United States in the future, which will be the most ideal bargaining chip for obtaining government financial support. Of course, it is difficult to predict when Intel will be able to re-establish its leading edge in computing technology through OEM business.