Where should ARM go in the future?

NVIDIA’s acquisition of ARM is in trouble, and the acquisition has entered a stalemate stage. Most companies in the industry have expressed their opposition to the acquisition, and even the US government opposed the acquisition last week. Why most companies want ARM to remain independent and whether ARM will change its name to N in the future is worth discussing.
Position of ARM
The predecessor of ARM, Arcon, was founded in Cambridge, England, in 1978. In the late 1980s, Apple began to cooperate with Arcon to develop a new version of ARM core. In 1985, Arcon developed a new processor using the reduced instruction set, called ARM (Acorn RISC Machine), also known as ARM 1.
Due to Arcon’s financial situation, on November 27, 1990, Arcon was funded by Apple and VLSI Technology to split ARM (Advanced Risc Machines) into an independent subsidiary. Using the proven RISC technology, ARM is committed to developing the market of low-cost, low-power, high-performance, 32-bit reduced instructions.
At first, ARM also produced chips, but its performance was mediocre. So they changed their business strategy and authorized their technical intellectual property (IP core) to many famous semiconductor factories in the world to make profits. ARM has three authorization methods: processor, POP and architecture authorization.
Processor authorization refers to authorizing the partner manufacturer to use the processor designed by ARM. The other party cannot change the original design, but can adjust the frequency and power consumption of the product according to its own needs.
The second is POP (processor optimization package) ARM provides a series of processor design schemes. The other party can select the appropriate design scheme for production according to its own needs,
The third kind of architecture authorization is that ARM will authorize its partners to use their own architecture, so that they can design processors according to their own needs. This model has the greatest freedom and is suitable for companies with strong technology. Typical customers are Qualcomm, Apple, Huawei, etc.
The licensing fees and royalties have become the main revenue of ARM, and also helped ARM share the risks. More importantly, ARM has established its own ecosystem.
Where will ARM go in the future?
ARM “universe”, source: ARM
ARM’s architecture has absolute dominance in mobile terminal, automatic driving and other fields, and is also struggling to catch up in the server market where the X86 architecture remains the leader.
Why does Softbank sell ARM?
In 2016, Softbank Group of Japan acquired ARM, the most valued technology company among the London listed companies at that time, at a price of $32 billion, which was also the largest acquisition in the semiconductor industry at that time. As a Japanese venture capital company dedicated to IT industry investment, the acquisition of ARM was once regarded by the industry as a step for Softbank to enter the frontier technology field. Softbank said that the acquisition was because it saw the potential of ARM in the field of the Internet of Things, which is the area that Softbank wants to develop.
However, this acquisition does not seem to bring the expected benefits of Softbank.
Where will ARM go in the future?
Source: Softbank financial report, organized by the semiconductor industry
From the financial performance of ARM in the past four years, it can be seen that the growth of ARM sales has entered a plateau. Softbank’s financial report shows that due to the weak economy in 2020, ARM’s technology non-royalty income (technology license income and software and service income) decreased by 1.7% year-on-year. The bigger problem of ARM is that its profitability is not stable. It can be seen that between 2017 and 2020, there are two years of negative revenue. Although the number of startups building their first products around ARM technology will increase in 2020. ARM has signed more than 160 license agreements, far higher than its historical operating rate. However, many high-value transactions contributed little to the revenue of FY2020. Softbank said: “Most of the increase in the revenue from the royalties of ARM in fiscal year 2020 is due to the licenses signed in 2017 or earlier.”
In 2020, the operating loss reached 88 billion US dollars and had to sell its investment to reduce its liabilities. ARM, which was “valuable” but could not make Softbank “make big money”, was sold.
ARM is a key part of NVIDIA’s consolidation of the data center.
Data center business has become the direction of every chip company, and NVIDIA will not miss it. According to Nvidia’s financial report, the revenue of Q3 data center has reached 2.94 billion US dollars, up 55% from the same period last year, making it the fastest growing platform among Nvidia’s four major product platforms. What is more noteworthy is that the revenue of the data center has been close to NVIDIA’s highest game revenue ($3.22 billion).
The revenue source of Nvidia Data Center is mainly its GPU products. Nvidia’s graphics card has been adopted by many large-scale customers, including cloud providers such as AWS, Microsoft Azure and Google Cloud. At the same time, Polaris, the largest GPU-based supercomputer in Argonne National Laboratory in the United States, will also run on Nvidia’s accelerated computing platform.
NVIDIA, which has tasted the sweetness in the data center, announced in April this year that it will launch the processor Grace for large AI models and high-performance computing. NVIDIA’s Grace chip takes advantage of the flexibility of the arm architecture, so for NVIDIA, the arm architecture is a necessary link for NVIDIA to enter the server market.
In 2020, ARM launched products, including Neoverse series processors for data centers and Ethos machine learning processors using AI. In September of the same year, NVIDIA announced its acquisition of ARM with US $40 billion.
Where will ARM go in the future?
Where will ARM go in the future?
Comparison of market share of ARM architecture, source: Softbank financial report
It can be seen from the financial report of Softbank that the ARM architecture has occupied 75% of the market for in-car entertainment and driving assistance systems, and has also started in the data center market. More importantly, it accounts for 90% of the Internet of Things application processor market, which lays the foundation for the expansion of ARM architecture in the data center business.
If the acquisition is successful, NVIDIA’s products can be better compatible with ARM architecture. Through the integration with ARM technology, NVIDIA’s DPU, CPU and GPU have great hope to open the server market monopolized by Intel and AMD.
Why can’t ARM last name N?
Simon Higgs, the CEO of ARM, once believed that listing was bad for ARM. He believes that IPO will affect ARM’s innovation ability. In fact, it is not difficult to understand that if the company is not listed, these shares are only held by a small number of people. ARM is free to conduct technology research and development and is open to all. However, after the listing of the company, the equity was dispersed, and the right to speak could not be grasped by the company’s personnel, and even the product path was disrupted by the usurpers. In this way, ARM’s business behavior will be interfered.
However, more voices worry that the acquisition of ARM will lead to the emergence of a monopoly market. As mentioned earlier, ARM has authorized almost all the most important chip companies in the world. When ARM becomes a subsidiary of NVIDIA, companies in the same industry are worried that this will affect its openness. Even though Huang Renxun promised to maintain the open authorization mode of ARM, he still met the opposition of the industry and many governments.
An earlier article analyzed that under the situation that the US government wants to strengthen its dominance in the semiconductor field, Nvidia still has a high probability of being approved to acquire ARM. However, this view was overturned by the United States Federal Trade Commission (FTC) in December. The reason why the FTC announced to prosecute the merger of NVIDIA and ARM was precisely: “The merger may kill the next generation of technology”.
No matter how Huang Renxun guarantees, once NVIDIA has acquired ARM and both have jointly developed a technology, the licensing of this technology will always be affected by NVIDIA. In this case, the market has formed a monopoly for companies using arm architecture. For ARM itself, it has greatly reduced its public nature, which may lead customers to turn to other structures, and its dominant position may be greatly shaken in the future.
For example, the deadline for the transaction of Nvidia’s acquisition is still ten months away. If it fails, Nvidia will lose $1.25 billion in advance payment, while Softbank will lose $74 billion in input. Softbank will probably choose to IPO ARM.
Whether ARM will remain independent or fall will affect the global semiconductor pattern.