Categories
科技報導

NVIDIA and ARM’s major mergers and acquisitions will change the chip industry



If Nvidia succeeds in acquiring Arm for $40 billion, it is expected to have a significant impact on the chip world, but it will take many years to fully understand the impact of this transaction. Due to a variety of factors, it is expected that more such transactions will occur in the next few years. Increasing interest in the acquisition of start-up companies with innovative technologies and easier raising money from the stock market are important reasons.In addition, there are many emerging markets that are gradually heating up, such as the continuous development of 5G, edge computing, AI/ML, and autonomous vehicles.

NVIDIA and ARM's major mergers and acquisitions will change the chip industry 1

In the short term, mergers and acquisitions in most industries will be affected by the new crown epidemic, but this will not continue.Acquisition is the fastest way to achieve scale growth and improve the company’s products, and it is also a way to quickly attract talents.

Nvidia’s acquisition of Arm is a combination of the world’s largest GPU supplier and the largest mobile processor IP supplier. It will help Nvidia expand its business from the data center to the edge and terminal. The deal also helps the two companies position themselves in the uncertain but emerging world of edge computing, in which highly specialized equipment and servers are used for preprocessing or processing large amounts of data. In addition, Arm has an extensive ecosystem.

“As a combined company, we can do many things. The increased investment will enable us to actively promote the development of data centers and promote the AI ​​that has been widely used in data centers to all corners and to the edge.” Arm CEO Officer Simon Segars said: “Nvidia has a large portfolio of IP products that can be used to build chips, products and systems. We have granted IP licenses to the global semiconductor industry and built an ecosystem around it. Therefore, we will have more IP License to customers.”

NVIDIA and ARM's major mergers and acquisitions will change the chip industry 2

Picture from bioworld

The success or failure of mergers and acquisitions, time will give the answer

At present, Nvidia’s acquisition of Arm has to pass the review and approval of national regulatory agencies. According to several industry insiders familiar with M&A transactions, the acquisition of Burr-Brown by Texas Instruments for $7.6 billion in 2000 seems to be a sensational transaction, but it turns out that the result of this acquisition is more than The initial expectations were much more limited.

The merger of the two companies has enhanced TI’s capabilities in the field of analog technology and strengthened the company’s business focus, but the impact on the entire industry is not so obvious. The same is true for TI’s acquisition of National Semiconductor for US$6.5 billion in 2011.

In contrast, ADI acquired Linear Technology for $14.8 billion in 2017 and Maxim Integrated for $21 billion, making it a strong competitor of TI, which has been dominating analog for decades field. ADI’s acquisition has brought this lucrative market to more price and performance competition, and it has also brought opportunities for start-up companies.

“Intel’s acquisition of Mobileye may be of great significance.” Segars said: “The acquisition of Broadcom by Avago is also an important transaction. People have been talking about semiconductor integration for a long time, and this is indeed happening. If you draw a chart of the market value of these companies, you will find the long tail, which is really interesting. ADI’s acquisition of Maxim is another matter. Due to COVID-19, the progress of the acquisition is slower. For companies that have never had an intersection. It is difficult to integrate together. Maxim and ADI have a long-term cooperative relationship. Arm, Nvidia and Softbank are also familiar with each other.”

EDA and IP mergers and acquisitions exceed other areas of the chip industry

The number of EDA acquisitions exceeds any other area of ​​the chip industry, but most acquisitions in the EDA industry are relatively small.However, the three major EDA companies will never become the three giants if they don’t have acquisitions to promote their growth, and the chip industry is unlikely to develop as before.Synopsys alone has completed more than 100 acquisitions. Mentor has completed more than 70 mergers and acquisitions, and Cadence has completed more than 50. (This number is not precise, because some are splits and acquisitions of assets, not entire companies.)

“Many products are incremental products, which have been improved technically and economically. One of our very large acquisitions was Avant. At that time, Synopsys established the so-called design front-end, synthesis, simulation, timing, power consumption Etc. At that time, the back-end was physical design, it was place and route and some verifications were done by different companies.” said Aart de Geus, chairman and co-CEO of Synopsys.

“In the late 1990s, I was increasingly worried that from a technical point of view, the interdependence between design and back-end design would become stronger. This happened around 2000, when people were worried about the economy. The impact of the downturn. Facts have proved that during the Great Depression in 2001, many consumers intended to reduce spending. There is only one company-Mentor between the two, so we think it is necessary to improve our ability to participate in this competition.”

It was Cadence that initiated the merger and acquisition in the EDA industry. “The two notable acquisitions in EDA are Cadence-Gateway and Cadence-Tangent.” said Wally Rhines, Emeritus CEO of Siemens Mentor. “As the world is shifting from schematics to RTL, Gateway brings Cadence into gateway development. Verilog business, which promotes Verilog to become a standard. Although the industry association is VHDL, Verilog is in a leading position.”

“Equally important is Cadence’s acquisition of Tangent. Tangent is a gate array router company (the best gate array router company), and they have also developed capabilities for standard cell migration. Earlier, Mentor IC Station and Cadence moved from Solomon The predecessors of Virtuoso from Design Automation all provide similar products. Today’s IC Station may be Virtuoso, but the internal competition between SCS (Silicon Compiler System) and Mentor’s IC Station brought them to a complete standstill. Provides an opportunity.”

It can be said that the most significant acquisition in the EDA industry involves the acquisition of Mentor by Siemens AG in 2016.Siemens is a large enterprise group and the largest industrial manufacturing company in Europe. This $4.5 billion transaction allows Siemens to provide complete services from design software to complete semiconductor processes.

At the same time, Mentor has become a powerful company with a capital scale much larger than the sum of all EDA companies. From this perspective, in 2019, Siemens’ revenue exceeded 100 billion US dollars. The sum of the market value of all EDA companies is only a small part of it.

IP companies are also actively buying other IP companies, and Arm’s growth is at least partly due to acquisitions over the years. Some EDA companies, especially Cadence and Synopsys, have also acquired many small IP companies.

K. Charles Janac, Chairman and CEO of Arteris IP, said: “So far, the most important acquisition in the IP field is Synopsys’s nearly $1 billion business, mainly focusing on peripheral I/O IP and PHY technologies. .Nvidia’s acquisition of Arm is extremely important for both the IP industry and the semiconductor industry. Nvidia is trying to become the next generation computing platform, directly competing with Intel and AMD. Therefore, with the support of Nvidia, maybe the Arm architecture will become the next core of the SoC. “

The key is synergy. “I used to work for Joe Costello, when he was the CEO of Cadence, he had a few interesting ideas.” Janac said, “This transaction is a good financial transaction, but it must also have some synergy. If Putting the two companies together, their sum should not be equal to 2 or 2.5, it should be 3 or 4. The two companies also need to be culturally compatible and actively cooperate. Another thing Costello said is that the most A good deal is when both parties are dissatisfied with the deal. If someone is really happy, it may not be a big deal.”

Major acquisitions in processors and memory

In terms of processors, the significance of Nvidia receiving Arm is not limited to a certain aspect. Arm’s ecosystem is so extensive that Arm processor cores dominate everything from application processors in smart phones to various portable devices. Of course, this is not the only important significance of the acquisition.

AMD acquired GPU manufacturer ATI for $5.4 billion in 2006, making AMD an important competitor to Intel in the data center market.“AMD acquired ATI’s graphics business, which allowed them to further develop the CPU business on the basis of Intel-compatible CPUs.” said Wally Rhines, emeritus CEO of Siemens Mentor. “ATI is a very unique company. If you look back in history, you will find that as graphics standards change, a new graphics company will appear about every 10 years. ATI is the only company that has survived for several generations. “

In the storage field, a series of acquisitions contributed to the revival of Micron.He said: “Hitachi and NEC merged to become Elpida (Elpida), which was acquired by Micron. This did enable Micron Technology to retain its DRAM business. Today, there are three major DRAM suppliers in the world. Although Micron is the smallest, that This transaction gave them enough technology and competitiveness, as well as the influence of Japanese companies.”

Shane Rau, vice president of IDC research, agreed. He said: “Micron’s acquisition of Elpida enabled them to cross the threshold. There are now three major memory companies, which is necessary for them to have sufficient supply and demand in DRAM. We expect consolidation in the NAND market.”

Government intervention and influence on acquisitions

However, not all announced acquisitions can be realized. The review process may produce surprising results, and the government may think the country is threatened.In the past, the US government was mainly responsible for terminating or postponing major transactions, but other countries such as China began to exert their market power, China rejected Qualcomm’s $44 billion plan to acquire NXP.

There are many similar cases.The most important first transaction in the chip industry was Fujitsu’s attempt to acquire Fairchild Semiconductor in 1987.At the joint request of the U.S. Department of Defense, the Department of Commerce, and the CIA, then President Ronald Reagan intervened. This transaction brought the chip industry, especially the government, to the attention for the first time, and prompted the birth of another industry giant.

“Due to the failure of the Fujitsu acquisition transaction, National Semiconductor acquired Fairchild Semiconductor.” Rhines said, “This transaction is significant because National Semiconductor has minimized its R&D investment for decades, and Don Brooks (at the time) President and CEO) Fairchild Semiconductor has developed outstanding new technologies. It gave National Semiconductor a new life because a company that minimized R&D investment and maximized operational excellence acquired a R&D investment Maximized company.”

The US government also played an important role in two other transactions that indirectly affected the chip industry. The first relates to the consent decree signed by IBM in 1956. At that time, the decree was designed to restrict IBM’s monopoly pricing of bundled services, software, and mainframe computers, known as the “market basket.”

IBM was the only company that provided these three functions at the time, and because the platform at the time was also an industry standard, all software must be compatible with IBM’s equipment. Therefore, IBM uses its market influence to use low-price strategies to compete with any of these three market segments.

When the PC era began in the early 1980s, IBM was still earning income mainly through mainframes and microcomputers. The view of its executives was that the PC was just a toy. Therefore, IBM’s idea is that instead of stifling market competitors again and facing more government intervention, it is better to sign deals with Intel and Microsoft instead of trying to own all the technology.

The US government also interrupted the Bell System transaction in 1982, allowing another transaction to be concluded. AT&T gave up control of Bell Operating Company, and Bell Operating Company was split into regional operating companies. Bell Labs has carried out a lot of work in cooperation with the federal government. At that time, Bell Labs was one of the world’s major R&D departments, on par with IBM.Bell Labs invented the first transistor in 1947, which was the Unix operating system that became the basis of Linux and the first optical router.

Bell Labs was sold to Alcatel as part of Lucent in 2006, which was part of the spin-off, and was acquired by Nokia in 2016. At the same time, GlobalFoundries acquired IBM’s microelectronics business in 2015. These two acquisitions ended the two largest semiconductor research businesses.

Although IBM is still doing chip research for AI systems, in the United States, the heyday of corporate and company/government-sponsored semiconductor research, especially communications and computing research, is over.

Governments have also blocked other transactions in the chip industry. The Committee on Foreign Investment in the United States (CFIUS) blocked Tsinghua Unigroup’s $23 billion acquisition of Micron in 2018. The U.S. Department of Justice also blocked the $9.3 billion merger of Applied Materials and TEL (Tokyo Veliki) in 2015.

40 billion U.S. dollar merger scale cap

Nvidia’s acquisition of Arm remains the largest in the semiconductor industry, albeit in a small number. Avago acquired Broadcom for $37 billion in cash and stock. However, it is worth noting that there seems to be an acceptable upper limit.

Rob Lineback, senior market research analyst at IC Insights, said: “A few years ago, we determined that the semiconductor merger agreement (excluding transactions related to the system level and software business) has reached the upper limit of approximately US$40 billion.Qualcomm’s failure to acquire NXP for $44 billion was cancelled in July 2018 because China has been delaying approval of the transaction during the trade war. US Broadcom’s $121 billion hostile takeover offer for Qualcomm (later down to $117 billion) was blocked because of concerns about the country’s loss of its leadership in cellular communications technology. “

Although there is sufficient funding and the company’s willingness to participate, scale is still important. “Due to the high value of large-scale transactions, the rise of trade protectionism among more countries, and the intensification of trade frictions, approximately $40 billion seems to have become a feasible acquisition limit for the semiconductor industry.” Lineback said,

“The geopolitical environment and trade wars may continue to limit the scale of semiconductor mergers and acquisitions. However, Nvidia’s $40 billion agreement violates this hypothetical upper limit. The Nvidia-Arm deal not only affects major players in many areas of the IC industry, but also seems It is also a test of the restrictions on chip mergers and acquisitions under current geopolitics.”

In China, there are also share restrictions on acquisition transactions. In order for a non-local company to establish a joint venture, the Chinese partner company must own 51% or more of the shares in the subsidiary. IDC’s Rau said: “This transaction makes Arm divest 51% of Arm China, which is of great significance. This is one of several transactions to bring intellectual property to China. MIPS is open to China and RISC-V is also open to China. .”

The semiconductor industry is entering a new stage

It remains to be seen how the new acquisitions will change the chip industry, but the slowdown of Moore’s Law, the development of chip designs in a more heterogeneous direction and the driving force for processing and intelligence requirements are changing the dynamics of the chip industry.

“We seem to be entering a new stage in the semiconductor industry. In the 1970s and 1980s, vertical integration was usually carried out with large semiconductor companies that developed their own processor cores, EDA tools, and sometimes even processing equipment.” Codasip Said Roddy Urquhart, Senior Marketing Director of.

“By the 1990s, this situation had been broken by companies such as Texas Instruments and Siemens Semiconductors (spun off in 1998) and abandoned internal EDA tools in favor of commercial tools. In the same time frame, including Arm IP companies such as, MIPS, ARC, and Tensilica emerged to provide internal core alternatives. By 2000, most companies in the world depended on the three major EDA companies and Arm to meet most design tools and IP requirements. In the same time frame, we have seen the emergence of pure wafer foundries and the growth of fabless semiconductor companies. With a stable business environment, global IC design has developed in many regions, especially in China and India.”

Recent geopolitical tensions have disrupted this stability. “Because China cannot use American technologies, such as Android, certain EDA tools, and, they will be forced to establish a local ecosystem. Arm is considered to be a third-party company completely independent of its licensees, but will be acquired by Nvidia. This independence will be broken. In the processor field, RISC-V is an obvious alternative to the company’s attention. For EDA companies headquartered in neutral regions such as Germany or Canada, this is also an opportunity.”

More mergers and acquisitions allow the market to reorganize in different regions, the adoption of new technologies and the need to process more data in more places, there will be more acquisitions, but the speed is unclear.