Due to the resistance of regulators in the United States, the United Kingdom and the European Union, Nvidia had no choice but to announce last month that it would abandon the $40 billion acquisition of Arm. Shortly thereafter, Intel's current CEO, Henry Kissinger, said intel intends to jointly acquire Arm through a consortium with other vendors. "We're not a big arm customer, but we do use Arm. As our foundry business incorporates this technology, we will use more Arm products in the future. "So, if there's a consortium, we might be willing to get involved in some way." He also hinted that in addition to Intel, including Qualcomm, Samsung Electronics, Apple, Google, Microsoft and even MediaTek and other manufacturers will not rule out participating in the acquisition in the future. As we all know, Intel is one of the most representative manufacturers of the x86 architecture camp, and there is a de facto competitive relationship with the Arm architecture camp. Kissinger said that Intel intends to participate in the acquisition of Arm, what kind of drugs are he selling in the gourd?
1, Arm has chosen the path of IPO listing, but may face challenges in the capital market
SoftBank itself is not a customer of Arm, which is one of the reasons why SoftBank was able to buy Arm in 2016 - SoftBank was able to keep Arm operationally neutral. Later, Nvidia intended to acquire Arm exclusively, which had made Arm's customers have doubts, and now, after Arm announced its IPO plan, the industry's major manufacturers won Arm through a joint consortium, and Arm can still remain neutral in business operations. In fact, arm after losing Nvidia, the big financier, Arm's IPO road may face challenges in the capital market.
On March 15, insiders revealed that Arm's new CEO, Rene Haas, announced in a memo to employees that the company plans to lay off 12% to 15% of employees worldwide. The source said that represents a maximum number of layoffs of nearly 1,000 people, but engineers are largely unaffected. Arm said most of the layoffs will occur in the U.K. and the U.S., but did not provide specific numbers for the cuts in each country. "Like any other company, Arm continuously evaluates its business plans to ensure the company strikes the right balance between opportunity and cost discipline," the company said in the announcement. For Arm's layoffs, the outside world analysis said that Arm's move is preparing for the IPO, Arm needs to reduce expenses, streamline projects, and make the financial report figures look better, so that it is more conducive to IPO.
It is necessary to add that after being acquired by SoftBank, Arm has worked tirelessly to open up new markets for growth, and Arm has greatly expanded its team with the support of SoftBank. Arm's workforce has increased by 1,034, 101 and 751 respectively from FY2018 to FY2020, respectively, and Arm's workforce continues to increase by 7.9% in FY2021. As the size of the technical staff continues to rise, Arm's operating costs naturally rise. Compared to the company's ever-expanding staff, Arm's expansion into new markets has not been so smooth.
ARM Processor Market Share Source: Pacific Securities November 2020 Research Report
According to the November 2020 research report of Pacific Securities, Arm's share in the smartphone, modem, in-vehicle information device and wearable device markets is 99%, 99%, 95% and 90% respectively, in short, Arm's share in the mobile processor market has reached 90%, and the market is basically saturated. However, according to SoftBank's earnings report (SoftBank's fiscal year ends March 31 of the current year), from fiscal year 2018 to fiscal year 2021, Arm's revenue was 202.344 billion yen, 202.699 billion yen, 206.652 billion yen and 209.848 billion yen, respectively. Over the course of four fiscal years, Arm's revenue growth has been sluggish. At the same time, Arm only made a profit of 133.966 billion yen in fiscal 2019, and in fiscal years 2018, 2020 and 2021, Arm lost 31.380 billion yen, 42.819 billion yen and 33.873 billion yen, respectively. Arm's former CEO Simon Segars has mentioned that if Arm chooses an IPO, it will need to face short-term revenue growth and profitability pressure after going public. In terms of investment, expansion, innovation, etc., Arm's response speed will be affected.
Arm Q1 2017-Q3 2021 Quarterly Revenue (Source: Statista)
Arm FY2017-FY2021 Revenue and Revenue by Sector (Source: Statista)
2, the performance of customers is soaring, Arm in order to get growth is trying to find a way out
In September 2020, when Nvidia announced its intention to acquire Arm, it immediately caused an uproar in the Arm architecture ecology. In the PC field, if customers want to buy x86 architecture processors, only Intel and AMD are available; relative to the Arm ecosystem, there are hundreds of manufacturers that manufacture, design and sell Arm architecture processors, including Google, Microsoft and Qualcomm, which have expressed high concern about the acquisition. Other manufacturers, including Apple, Samsung, and even MediaTek, even Intel and AMD in the x86 architecture camp, are licensed to Arm, which may be affected by the acquisition.
Since SoftBank itself is not engaged in chip design and is not a semiconductor manufacturer itself, the main difference between arm's open licensing model and the x86 model is that Arm itself is not engaged in any chip manufacturing work, and therefore, the Arm camp (chip designers) has been able to crush the x86 architecture in the field of mobile processors through in-depth cooperation with chip foundries such as TSMC, Samsung, GF, and UMC.
Arm provides both architectural licenses and IP licenses. With apple, AMD, Qualcomm, Samsung and even Intel including such architecture licensing customers, after obtaining architecture licenses from Arm, they can create Arm processors through customized microarchitectures developed by customers themselves, and Apple's A-series processors are like this. Some no longer emphasize the licensing customers of customized processors, including Samsung's CPU and Qualcomm's AP part development, and even directly purchase Arm's own standard chip design licenses, including Cortex-A9, Cortex A-72, Cortex-A75 and so on. Of course, the premium for architecture licensing is much more expensive than standard chip design licensing, but Arm offers both services.
Arm's urgent need for capital injection is still an indisputable fact, and in the documents submitted to regulatory authorities, it is emphasized that if there is no investment from Nvidia, Arm's plans to expand in the data center market and compete with Intel and other x86 architecture camps are bound to be relatively weak. Especially in contrast to the soaring performance of major customers such as Apple, Qualcomm, and Amazon, Arm's industrial position is very disproportionate to actual revenue, and the increasing cost has gradually crushed the global semiconductor intellectual property (IP) leader that has been established for 30 years.
3, Arm focuses on the data center market, is the had to do
In fact, before obtaining Nvidia's offer to acquire Arm, SoftBank considered letting Arm go public in 2019 and 2020, and finally gave up due to assessing that the capital market could not effectively return SoftBank's investment. However, after the failure of the NVIDIA acquisition, SoftBank once again proposed a plan to let Arm IPO on its own, but from the perspective of the capital market, Arm still faces operational challenges - it must submit public financial report data every 3 months. The capital market focuses on the company's short-term performance growth and profitability, which may not be beneficial to Arm's long-term operation and development.
As mentioned above, Arm's biggest source of revenue is the mobile processor market, which is now saturated. In contrast, the data center market and the PC market sent by Arm's future investment are very difficult to enter. Moreover, Arm's entry into the data center market and the PC market has been shouting for 12 years, as a comparison, whether it is Intel or AMD and other manufacturers who are popular in the data center market, they are supported by the developer, software, system and peripheral ecosystems.
Whether it is SoftBank originally planned to sell Arm, or Nvidia had the intention to buy Arm, one of the main reasons behind it is to focus on entering the data center market and the PC market, but Arm has announced its entry into the data center market so far, and the progress it has made is actually quite limited, mainly through the way of licensing amazon, supporting the creation of customized server processors, and ampere Computing, the only new entrepreneur who provides Arm architecture data center CPUs. The vast majority of Arm architecture chips are used in smartphones and wearable devices, for now, Arm's chip sales in the data center market are far from being comparable to the mobile device and wearable device markets, but the huge revenue growth momentum of data centers is obviously something that Arm cannot ignore, or make Arm salivate.
For example, chips built for data center servers are priced at thousands of dollars; even entry-level chips for artificial intelligence AI cost thousands of dollars. In contrast, a smartphone only needs a SoC processor and costs about a few tens (40) US dollars, and in the context of the global mobile phone market becoming saturated, and from the perspective of revenue growth, in theory, Arm's expected benefits of choosing to enter the data center market will be quite large.
4, Arm has publicly named, RISC-V architecture is the number one competitor
The competitor that Arm has publicly named is the RISC-V camp, which also runs a licensed IP business. Previously, customers were worried that Arm might be acquired by NVIDIA, so they turned to alternatives including RISC-V, but now that NVIDIA's acquisition of Arm has failed, the situation of customers switching to the RISC-V camp should be slowed down. But it is undeniable that as Arm moves towards the path of IPO independent operation, the RISC-V camp may have more opportunities to become an alternative to Arm.
In terms of business model, RISC-V is similar to Arm, but due to Arm's exclusive nature for chip design, customers basically cannot change the instruction set, and can only strengthen or expand it. In contrast, RISC-V opens up the source code instruction set, allowing members to change the architecture itself. Compared with the Arm camp, the RISC-V camp has a greater tolerance, or can be compared to the early openness of the Android camp in the mobile phone business. Once Arm moves to independent operations, it is expected to face greater challenges from the RISC-V Consortium.
List of senior members of the RISC-V International Foundation
Recently, Intel officially joined the RISC-V International organization (formerly known as the RISC-V Foundation), alongside Google, Witten, SiFive, and Jingxin Technology, becoming the top members of the organization. Since 2021, under the leadership of Kissinger, Intel's layout actions can be described as frequent, actively recruiting foreign troops, and has also invested tens of billions of dollars to significantly expand factories to improve chip manufacturing capacity - transforming into chip foundry service providers. In addition to the x86 and Arm architectures, it has even begun to invest heavily in and lay out RISC-V processor-related technologies. Established in 2015, RISC-V International has members covering major software and hardware manufacturers, including Google, IBM, Samsung, Qualcomm, MediaTek, Nvidia and Witton, as well as up to 2,000 members of major open source communities, organizations and individuals. Although Intel is not the first server CPU manufacturer to join, with the addition of Intel, it is undoubtedly an injection of a shot in the arm for the development of RISC-V.
5, Intel is ambitious, its chip foundry services or cover Arm, RISC-V and x86 three major architectures
In the future, Intel will contribute its expertise and design optimization experience in the x86 architecture, as well as providing an open source software development platform, to help drive the development of RISC-V technology and ecosystem to provide chip foundry services including supporting RISC-V products. Intel's ambitions don't stop there, with Intel Foundry Services intending to license hybrid chips that combine Arm, RISC-V, and x86 architectures.
Bob Brennan, vice president of Intel's chip foundry service business, recently pointed out that Intel conceived a small chip design, including x86, Arm and RISC-V architecture related IP can coexist on the same silicon at the same time, from the perspective of long-term development of the market, the three camps of various IP are invested in a specific kind of optimal use, through the integration of design, and then combined with small chip packaging technology, although it has not yet fully developed a specific strategy, but Intel believes that this has long-term market potential.
Intel and x86 can be said to be synonymous with two sides of the same coin, and Intel's chip foundry service announced support for both RISC-V and Arm architectures, in fact revealing a whole new milestone. From 2021 onwards, Intel has released a number of hybrid x86 architecture CPUs that combine low power and high-performance cores on the same silicon, and the aforementioned x86, Arm and RISC-V architecture chips are packaged together and connected to the cores through small chips, which can be said to be an extension of the same concept.
For a long time, chip manufacturers have not designed and built SoCs in such a way that microarchitectures have not been mixed and matched. But Intel's vision is that by mixing and matching CPU instruction sets and connecting hardware through small chips, conceptually similar to gpu developments, the workload can be further enhanced and amplified to a position comparable to CPU computing.
Three mainstream CPU design architectures
6, NVIDIA fully promotes arm architecture data centers, pointing to the intel core market
At this point, it is not difficult to infer that Kissinger's intention to acquire Arm is not a lie, which is a piece of the puzzle that must be strengthened in the process of strengthening chip foundry services after Intel expands out of the CPU core business. Even if Intel combines the acquisition alliance to obtain a considerable proportion of the equity in the process of Arm's listing, from the perspective of Arm's business growth, Intel may not be able to provide the same financial support as Nvidia promised.
On the other hand, despite the failure of NVIDIA's acquisition of Arm, CEO Huang Jenxun recently stressed that its first Arm architecture CPU will still be released on time in the first half of 2023. Huang Jenxun stressed that NVIDIA has obtained a 20-year-long Arm architecture license, which will fully use the instruction set to develop the CPU required for various applications, from robot applications to high-end processors of supercomputers, as for the advent of the Arm architecture Grace CPU in 2023, it is only the beginning of a large range of Arm architecture CPU product series.
So far, Apple's iPhone smartphones, iPad tablets and Mac computers have all used self-developed Arm architecture processors, but have not shown any intention to develop Arm architecture processors for data centers or supercomputers. NVIDIA hopes to develop a variety of Arm architecture chips from top to bottom, specialized development for a variety of different application markets, somewhat similar to Apple's A series SoC chips for the mobile processor market, and M series SoC chips for Mac computers.
From Arm's point of view, in the data center market, Arm architecture products are bound to compete with x86 architecture products. However, based on its own operational considerations, Intel will not let go of the opportunity to invest in Arm, and Arm's other customers will not miss the possibility of investing in Arm, and even NVIDIA may become one of the investors.