원문 보기
October 1st, 2026, New York. Shares of ARM, the chip design company, closed at $292. Soft Bank Group owns about 920 million ARM shares. At that price, they are worth roughly $270 billion. The next day in Tokyo, SoftBank Group's own share price put the whole company's value at about 36 trillion yen or roughly 228 billion.
Its arm stake alone is worth more than SoftBank itself. 10 years ago in 2016, Masayoshi's son bought this company for about 3.3 trillion yen and he paid in cash without issuing a single new share. After the announcement, SoftBank stock fell about 10% in a single day. Was it too expensive? Where did the money come from?
And what is ARM to SoftBank today? Using the announcements and financial results of the time, let's trace the ARM acquisition, June 2016. Soft Bank was selling off the shares it held one after another. At least $7.9 billion worth of Alibaba stock, and its stake in Supercell, the Finnish game maker, sold to China's Tencent for $7.3 billion.
The company said it was to pay down debt. June 23rd, Britain voted to leave the EU, and the pound plunged. In early July in the Turkish harbor town of Marmmoris, Sonn met ARM's chairman Stuart Chambers. It was there that Son first said he wanted to buy ARM. July 18th, London, Soft Bank announced the deal, £17 a share, about 24 billion in total.
That was a 43% premium to the last closing price. At the press conference, son revealed that the talks had begun only two weeks earlier. It's true by the way that I have met with the chairman of ARM only two weeks ago as the first time. Okay. And that's the that's the day I said I'm interested in uh acquiring.
Did he buy because the pound had fallen? This was son's answer. Well, uh, Brexit did not affect, uh, okay. So, I decide for the viewing the paradigm shift as the opportunity. This is the beginning of IoT. I would have made this decision at this timing regardless of Brexit happened or did not. Earlier in the same press conference, he explained why SoftBank had succeeded.
If I say only one reason, I'm not investing into the distressed asset. I am investing at the beginning of paradigm shift. The market was not impressed. When Tokyo reopened, SoftBank stock fell about 10%. ARM was founded in 1990 in Cambridge, England. Backed by Apple and others, it first designed power efficient processors for small computers.
That low power design spread quickly with mobile phones and then smartphones. ARM doesn't make chips itself. What it sells is the blueprint for a chip's brain, the CPU. Chip makers pay a license fee to use the design and then a royalty on every chip they make. It is the opposite of Intel, which makes and sells its own chips.
In 2015, about 15 billion chips using ARM designs were shipped. Almost every smartphone in the world ran on an ARM design. That year, ARM's revenue was $1.49 billion. A small company, but with an operating margin above 40%. So I am one of the very few person who bet with a big size of cash. Uh not just talking you know talking is easy right?
Talking to say oh UK will be still good great country it's easy to say but I'm proving that with big size cash. So, how did Soft Bank pay33.3 trillion yen? ARM shareholders received about 24 billion pounds all in cash. At the press conference, Son explained how it was funded. How do we uh bring the money for that?
2/3 of that is uh our cash on hand and the remaining one/ird is the debt. So 70% our own cash, 30% is the debt. This debt is the bridge loan. The cash on hand started with about 2.5 trillion yen at the end of March. Added to that was the money from three share sales in June and July. Okay. The reason is because I got the money finally now.
I didn't have the money to invest uh right before that. Okay. And uh uh I would have uh uh you know uh wanted to do this but I was waiting for the cash to come on hand. First Alibaba. Since investing in 2000, SoftBank had not sold a single share in 16 years. This time it raised $10 billion in total, but only $3.4 billion of that was an outright sale.
The remaining $6.6 billion came from turning a promise to deliver shares 3 years later into cash today. A newly created trust raised $6.6 billion from investors who received 5.75% interest a year. The trust then paid about $5.4 $4 billion upfront to the SoftBank side. In return, SoftBank promised to deliver Alibaba shares 3 years later.
Investors took the loss if the price fell, while Soft Bank kept gains of up to 17 12%. In June 2019, Soft Bank settled the deal by delivering 73 million Alibaba shares. Next, Supercell, the Finnish game maker. Soft Bank sold its 72.2% 2% stake to a 10-cent le group for about $7.4 billion. But the money came in installments.
By September 5th, when the ARM deal closed, only $3.5 billion had arrived. The other 3.7 billion came in October. Soft Bank also sold about 72 billion yen of shares in the game company Gungho. The money would arrive later than the payment was due and under British takeover rules, the cash had to be certain before the deal was announced.
That gap was filled by a 1 trillion yen bridge loan from Mizuo Bank. Son explained it this way. The cash is coming next month and a few months after that. So for that we are going to use uh bridge loan because uh the cash supposed to come from next month. Uh for this transaction we had to prepare the uh the cash uh before the announcement for the certain fund as the UK u exchange board uh required.
So we have uh utilized the uh bridge loan from the uh Japanese bank. It was unsecured and guaranteed by SoftBank Corp. the domestic telecom unit. The term was 2 years and it had to be paid down early as cash came in from share sales such as Alibaba. In November 2017, the bridge loan was replaced together with the Sprint acquisition loans by a 2.65 trillion yen long-term loan from a bank syndicate.
Another tool was the hybrid bonds issued in September. 400 billion yen for individual investors and 71 billion yen for institutions. A 25-year term at 3%. They rank behind other debt and interest payments can be deferred. Because of those features, credit rating agencies counted half of the money raised as equity.
Without issuing new shares, it had a similar effect to thickening its equity, at least by the ratings yard stick. The company itself said it had the rating agency's equity credit in mind. It was the first public hybrid bond offering to individual investors by a Japanese non-financial company. At the end of July, SAN stated flatly that Soft Bank would not raise money by issuing new shares.
Rating agencies were divided. S&P and Moody's kept their ratings while Japan Credit Rating Agency put SoftBank under review for a possible downgrade the day after the announcement. The deal also used a distinctly British procedure, a scheme of arrangement. If shareholders holding at least 3/4 of the shares approve, and a court sanctions it, every share changes hands at once, including those of shareholders who voted against.
At the August 30 meeting, 95% of the shares voted in favor. The court approved it on September 1. Soft Bank pledged to at least double ARM's UK workforce within 5 years. On September 5th, the deal closed. The final consideration was 3.367 trillion yen. Of that, 2.65 trillion yen was goodwill. A price paid not for factories or cash, but for future earning power, about 80% of the total.
At the end of July earnings briefing, son said that in 5 or 10 years, many people would see ARM as SoftBank's core business. At the same briefing, he also said a trillion chips would be scattered around the world in about 20 years. Yeah, we are going to enhance the security and also connectivity and uh we are we are going to uh uh ship 1 trillion AR.
But after the acquisition, ARM's revenue barely grew. In Soft Bank's accounts, ARM's revenue was 202 billion yen for the year to March 2018 and still 210 billion yen 3 years later. ARM hired more engineers and poured money into research and development. Profits thinned. In 2017, SoftBank launched the roughly $100 billion vision fund and committed up to $28 billion of its own.
$8.2 $2 billion of that was paid not in cash but in ARM shares. A quarter of ARM moved over 2017 and 2018. Even so, the fund was run by SoftBank. Its control of ARM did not change. Then in September 2020, SoftBank announced it would sell ARM to Nvidia for up to $40 billion. Soft Bank had paid about $32.4 billion.
After 4 years, that was about 1.2 times or roughly 5% a year. That spring, amid the COVID market crash, SoftBank had just decided to sell or monetize up to 4.5 trillion yen of assets. Son called Nvidia the perfect partner for ARM. But regulators around the world stepped in. ARMs designs are used by Nvidia's rivals, too.
If one company owned it, competition could be distorted. In December 2021, the US Federal Trade Commission sued to block the deal. Britain, the EU, and China were also still reviewing it. In February 2022, the sale was called off. SoftBank kept the $1.25 billion prepayment. Around the same time, ARM's leadership changed with Renee Hos becoming chief executive.
SoftBank chose to take ARM public instead. On September 14th, 2023, ARM listed on NASDAQ at $51 a share. It closed its first day at $6359, 25% above the offer price. Soft Bank sold some shares and raised $4.87 87 billion while keeping about 90%. Before the listing, it bought back the shares held by the vision fund for $16.1 billion.
Shares handed over as $8.2 billion worth came back for 16.1 billion. After the listing, the world entered a race to build AI data centers. NVIDIA's GPUs do most of the computing, but ARM designs spread to the server CPUs that work alongside those GPUs. In January 2025, the Stargate AI data center project was announced at the White House.
ARM was named as a key technology partner. The same year, SoftBank announced it would buy Ampere, an American maker of ARMbased server CPUs, for $6.5 billion. ARM's revenue reached $4.92 billion in the year to March 2026, more than three times the year before the acquisition. In April to June 2026, its data center royalties more than doubled from a year earlier.
In March 2026, for the first time in its 35-year history, ARM unveiled a chip of its own design. Its first customer was Meta. A company that sells blueprints now sells chips as well, which also means competing with some of its own customers. On June 18th, 2026, the stock closed at $439, more than eight times the IPO price.
By October 1st, it was at $292, down about a third in three and a half months. ARM came to play another role for Soft Bank, collateral for borrowing money. The limit on that borrowing grew from 13.5 billion to 20 billion and was reported to reach $25 billion in September 2026. 72% of ARM's outstanding shares are pledged as collateral.
The borrowed money is reported to be funding AI bets such as SoftBank's investment in OpenAI. When ARM's share price moves, SoftBank's borrowing power and its own market value move with it. In the 6 weeks from June 18 to July 29, the value of SoftBank's ARM stake swung by about $200 billion. So, how far would ARM share price have to fall before something happens?
If this year's 6-w week drop happened again, how high would the loan to collateral ratio rise? And how much room does the company's overall debtard stick have? We go through the numbers at each share price and how much buffer the company's own filing suggest in a membersonly follow-up. In 2016, Sun said ARM would become SoftBank's core business.
10 years later, ARM accounts for about 60% of the value of SoftBank's holdings. On that, he was right. A trillion chips in 20 years. By March 2026, cumulative shipments of ARMbased chips since its founding had passed 350 billion. A trillion is still a long way off. On the other hand, his plan to sell after 4 years fell through.
You could say ARM is worth what it is today because the sale failed. Soft Bank paid about $32.4 billion. Its stake is worth about $270 billion today. That's more than eight times, roughly 23% a year. In yen terms, helped by the weaker yen, it's more than 12 times. But this is the value of shares that haven't been sold.
If SoftBank sold a large amount, the price would likely fall, and many of those shares are pledged as collateral. A 3.3 trillion yen purchase decided in 2 weeks became in 10 years Soft Bank's biggest pillar. At the same time, that pillar is the collateral for borrowing for its next bets. sell the winners or borrow against them and put the money into the next bet.
Sun has repeated this pattern for more than 40 years. In this series, we'll follow his other big bets through the flow of money. The membersonly followup includes the calculations, figures, and sources behind the armbbacked loan. Find it through the link in the description.
Comments 0