Recently, two events have occurred in the semiconductor industry.

The first event is the dispute between Arm and Qualcomm. As Qualcomm launched several self-developed Oryon processors, Arm issued a mandatory notice to Qualcomm, demanding that it resolve the dispute within 60 days or face the cancellation of their architecture licensing agreement. Qualcomm responded by accusing Arm of its usual tactics, while Arm threatened legal action in December.

The story dates back to 2021 when Qualcomm acquired the chip design company Nuvia for $1.4 billion. The issue lies in the fact that Nuvia had previously obtained an ALA (Architecture License Agreement) from Arm. Arm offers two types of licenses: ALA, which allows chip manufacturers to develop their own chips based on Arm's instruction set, and TLA (Technology License Agreement), which provides a reference design for manufacturers to build upon.

For a long time, Qualcomm has relied on the TLA license, which has limited its ability to innovate and upgrade its chips. To obtain an ALA license, Qualcomm would need to renegotiate with Arm and pay additional fees. Instead, Qualcomm took a "shortcut" by acquiring Nuvia, which already had an ALA license. Although Nuvia's Arm license was terminated in March 2022, Qualcomm had already integrated Nuvia's technology into its product line.

As a major chip seller, Qualcomm estimates that replacing Arm cores with Nuvia-designed cores would save 40% to 50% in patent fees.

Naturally, Arm was not pleased and sued Qualcomm in 2022 for infringement. However, the legal process is lengthy, and Arm has also been accused of bundling sales, requiring chip manufacturers to use Arm's TLA for CPUs, NPUs, GPUs, and ISPs if they want to use Arm's architecture. This has led to a protracted legal battle between the two companies.

The second event involves Masayoshi Son, who has once again spoken out in favor of AI. At a recent FII forum, Son stated that he is focusing all his efforts on AGI (Artificial General Intelligence) and predicts that by 2035, superhuman AI (ASI) will be achieved, contributing over $9 trillion to the global GDP annually. He also believes that the market has underestimated NVIDIA's potential.

Recently, rumors have circulated that SoftBank is discussing with Apollo the establishment of a large AI fund to invest in data centers, chip factories, and other AI-related projects, potentially exceeding $20 billion. This fund may involve industry partners like Arm. Additionally, SoftBank participated in OpenAI's latest $6.6 billion funding round, contributing $500 million.

In his book "The Gambler: The Wild Ride of Masayoshi Son," FT Editor-in-Chief Lionel Barber wrote that in 2022, the 65-year-old Son lamented his life, saying, "I often see my face on Zoom video calls, and I hate it. What an ugly face. I'm just getting old..."

However, in 2023, the 66-year-old Son made a comeback with Arm, which had been deemed "unsellable." In 2024, at the age of 67, Son aims to raise $100 billion to launch a new AI chip project called "Izanagi," aiming to compete with NVIDIA's Jen-Hsun Huang.

Now, the "all-in" Son is ready to gamble again.

So, what is the future of Arm, and what are Son's plans?

What supports Arm's high market value?

Since going public, Arm's share price has nearly tripled, with a current market value of approximately $148.3 billion, compared to Intel's $99.9 billion. While NVIDIA is considered the biggest winner in the AI boom, Arm has also benefited unexpectedly. In the massive data centers built by Microsoft and OpenAI to train the next generation of ChatGPT, CPUs based on Arm architecture are working alongside NVIDIA's latest Blackwell series.

The development of AI is crucial to both NVIDIA and Arm. Not only does Son emphasize this, but Arm's CEO Rene Haas has also publicly stated that doubting the future of AI is akin to doubting the internet in 2000 or automobiles in 1907.

This view runs counter to the prevailing narrative of an AI bubble. However, even if the future of AI is bright, under the current business model, Arm will always be a secondary player to NVIDIA. Moreover, Arm is not particularly profitable. Despite 99% of smartphones using Arm-based chips, covering approximately 5.6 billion people globally, Arm only earns from licensing fees and royalties, which amount to a mere 0.5% to 2% of the chip's value.

In comparison, Arm earns far less than Qualcomm. In 2023, Qualcomm's revenue was $36.29 billion, with a net profit of $7.76 billion. To increase revenue, Arm plans to adjust its business model by charging royalties based on the value of the device rather than the chip. However, this change has met with resistance from customers, leading some to explore alternatives like the open-source RISC-V architecture.

Now, Arm is taking Qualcomm to court, which many analysts see as a power play to demonstrate Arm's influence in the industry. However, even if Arm's licensing model changes in the future, and it has already made inroads into cloud computing processors, automotive chips, and AI, it lacks a major end-device to drive the market rapidly. Without such a device, Arm will struggle to achieve the same level of dominance in other markets as it has in the smartphone market.

In this context, if the AI bubble bursts, Arm's high market value could also collapse, which is something Son desperately wants to avoid.

Son: Correct Instincts, Wrong Timing

Known as the "Instinct Master," Son believes that the achievement of all goals comes from unfounded belief! Contrary to the perception of being slow, Son's instincts have long been focused on AI. Since 2012, SoftBank has been investing in AI companies globally. From 2017 to 2022, Son mentioned AI over 500 times in quarterly and annual reports.

However, when OpenAI emerged as a trendsetter, Microsoft was the lead investor, leaving Son on the sidelines. Son attributes this to having the right instincts but choosing the wrong timing. During the Vision Fund era, many AI companies were small, in early development stages, or not yet in the public eye. During the COVID-19 pandemic, Son was stuck in Tokyo. By early 2022, when restrictions were lifted, SoftBank was facing significant losses.

If Son had not invested indiscriminately in over 500 companies, he could have acquired promising AI companies at lower valuations during the high-interest rate environment. Instead, due to indiscriminate spending, Son was forced to sell his NVIDIA shares in 2019 for $3.3 billion.

Now, with SoftBank's finances stabilizing, Son is focused on returning to his core mission – profiting from the AI era. For example, he wants Arm to compete with NVIDIA in AI chips. A former Arm executive expressed concern about the cost implications of such a move.

Why does Son believe Arm can compete with NVIDIA? Primarily because he believes the AI chip market is massive, and NVIDIA and other giants cannot dominate the entire pie. During the FII forum, Son hinted at his plans.

Dialogue 1

Host: You've admitted that you haven't fully invested in AGI due to lack of funds.

Son: That's right. Now I'm saving hundreds of billions of dollars to prepare for the next big move.

Host: Do you know what the next step will be? I mean, you might not even know these companies; they might not even exist yet. Do you know where this is going?

Son: I can't tell you that. I have my own ideas. Arm has nearly monopolized the mobile chip market and 90% of high-end AI and IoT chips. So, I'll soon be the boss of an AI-centric chip company. However, I'm particularly interested in AI robotics, not traditional robotics. The combination of ASI and robotics will create a great product.

Dialogue 2

Son: To achieve ASI that is 10,000 times smarter than humans, how much investment is needed?

I estimate that AI data centers will consume 400 gigawatts of electricity, more than the entire United States! They will also require 200 million chips! The total investment needed is $9 trillion! Many people think this investment is too huge and will never be recovered. But I think it's not enough, it might even be too little.

Why is $9 trillion not enough? If AI replaces 5% of global GDP in ten years, that would be $9 trillion! In other words, AI will generate $9 trillion in value annually! Compared to that, the initial $9 trillion investment is negligible and will be recovered within a year.

In reality, Arm's CEO Rene Haas admits that the company can manufacture NPU chips, but their performance is "not yet at NVIDIA's level" as they are primarily used in edge devices, not data centers. He adds that the company is already investing heavily to expand into this area, not just for edge devices.

Ironically, if Arm had not shelved its internal AI chip development project five years ago, Son's AI plans could have advanced further.

In 2016, SoftBank acquired Arm for $31 billion and faced pressure to return to the stock market. This meant Arm had to focus on profitable, mature businesses to boost investor confidence. At that time, major Arm customers like Apple, Qualcomm, and Graphcore poached Arm engineers to strengthen their own chip design capabilities.

Arm redirected many engineers from an emerging AI chip project to continue CPU development, leaving the company without a standalone NPU product suitable for data centers to this day.

Admittedly, SoftBank cannot compete with NVIDIA or giants like Amazon, Google, and Microsoft that plan to manufacture their own chips.

Nevertheless, Arm needs a new narrative, and AI is the perfect core for that story. How this "movie" will end remains to be seen. Perhaps we can once again trust in Son's "luck.""Some call Son lucky, while others say SoftBank is "too big to fail." Whether as a "pioneer of the times" or as someone emerging from a prolonged downturn, Son, the disruptor of the world, should never be underestimated.