From OpenAI to Malaysia, Masayoshi Son Unveils SoftBank's 'ASI Blueprint'! $50 Billion Investment Plan Ignites Bullish Sentiment in Semiconductors
In 2000, Masayoshi Son made his first investment in Alibaba, putting in about $20 million. By holding the shares for a long time, he watched the e-commerce company grow in value at a record pace. Alibaba went from losing large amounts of money in its early days to becoming a major name in global financial markets. Many people see this as a legendary investment. Today, Son believes OpenAI, which he has strongly supported, and other leading companies that provide AI computing power also have huge growth potential. He sees them becoming major ways for people to access digital services and important technology platforms for businesses.
Zhitong Finance APP has learned that SoftBank Group Corp., led by investor Masayoshi Son, is expanding its AI investment strategy. The plan focuses on OpenAI, a leading AI application company, and ARM, a major chip design company. It is also expanding into Southeast Asia's AI computing infrastructure and AI application markets. Son is moving faster to put his AI plans into action and build a complete AI system. This system covers chip designs, AI computing infrastructure, cloud computing infrastructure, energy, and AI applications.
According to recent media reports, Malaysian Prime Minister Anwar Ibrahim said Masayoshi Son plans to invest $50 billion in Malaysia. The proposed investment would cover artificial intelligence, emerging technologies, and data centers. Possible locations include Kuching in Sarawak and the Malay Peninsula. The plan supports Malaysia's goal of becoming an influential AI nation by 2030. It also offers a possible new source of demand for semiconductor companies in global stock markets, which have recently faced heavy pressure. Major investment firms and cloud computing companies are continuing to expand their AI computing infrastructure into more locations. Whether future orders for key AI infrastructure components, such as AI chips, memory chips, and data center CPUs, will be higher than expected depends on whether project funding proceeds smoothly, reliable electricity is available, and customers fully use and pay for the computing power being built.
SoftBank plans to invest $50 billion in Malaysia, which aims to become an AI nation.
Malaysian Prime Minister Anwar Ibrahim said SoftBank Group plans to invest $50 billion in the country.
According to the report, Anwar said on Saturday that SoftBank Group founder and CEO Masayoshi Son had chosen possible sites in Kuching, Sarawak, and on the Malay Peninsula after a video conference earlier that week.
According to reports, the Malaysian prime minister said the proposed investment would focus on artificial intelligence, advanced emerging technologies, and the construction of large AI data centers.
The investment plan comes as Malaysia works toward becoming an influential AI nation by 2030. Earlier this year, the government launched a National Artificial Intelligence Action Plan to strengthen the country's AI skills and infrastructure.
Earlier this month, SoftBank signed framework agreements with Singapore-based Grab and Malaysia's Sarawak Oil Company to explore the development of AI and digital infrastructure in Sarawak. The areas under discussion include AI computing services and advanced technologies.
SoftBank is not the only company planning major technology investments in Malaysia. In 2024, Alphabet, Google's parent company, announced that it would invest $2 billion to build its first large-scale AI data center and a major AI cloud computing region in the country. Amazon Web Services (AWS), the world's largest cloud computing service provider, said it plans to invest $6.2 billion by 2038. Microsoft also recently announced an investment of about $2.2 billion to expand traditional cloud computing, AI computing infrastructure, and other large digital projects.
According to Malaysia's Investment Development Authority, the country attracted about RM385.7 billion in investment linked to large-scale AI data centers from 2021 through the first half of 2026. This is equivalent to about US$94 billion. A SoftBank spokesperson did not immediately respond to an initial media request for comment.
Masayoshi Son is expanding his AI plans to include infrastructure, energy, and applications.
Son's strategy is taking shape as a system with several connected layers. It combines large AI models, chip-based AI computing infrastructure, energy systems, and leading AI applications, such as AI agents that can carry out tasks.
On October 1, SoftBank announced that it had completed its final $10 billion follow-on investment in OpenAI. This brought its total investment commitment to $64.6 billion and gave it an approximately 13% stake. In chip technology, ARM's CPU architecture and its own data center CPUs, together with the acquired companies Ampere and Graphcore, are expanding SoftBank's computing capabilities. For infrastructure, the company is working through Stargate, its large AI data center project in the United States; SB Energy, which supports energy needs; and SB Neo, which plans to offer cloud-based computing services similar to those of CoreWeave. These businesses are involved in building data centers, supplying energy, and providing computing power. In Japan's business AI market, SB OpenAI Japan is developing AI agent applications tailored to companies.
From the wider AI industry's point of view, this strategy aims to help SoftBank benefit from the development of super artificial intelligence (ASI). It seeks to gain value from rising ownership stakes in major technology platforms, growth in AI chip and infrastructure businesses, and income from renting out or selling AI cloud computing resources.
SoftBank Group has moved beyond making separate investments and is building a more complete ASI system. At the lowest level, it has ARM's chip architecture and its own data center CPUs. The middle layer includes Graphcore Ltd., a UK-based AI chip design company that SoftBank fully acquired, and Ampere, which it recently completed acquiring. The upper layer includes OpenAI, Stargate, and a business-focused cloud AI computing platform developed jointly with Oracle. SB Neo is another new player in its cloud computing plans. Masayoshi Son has publicly said that SoftBank aims to become the world's largest provider of AI computing power and core application platforms over the next decade, as the era of Super Artificial Intelligence (ASI) develops. He is taking a highly aggressive, strongly optimistic approach to investing in OpenAI, the company behind ChatGPT and a global leader in large AI models. Oracle mainly works with Stargate as its cloud and infrastructure partner. SoftBank's joint platform for business customers is SB OpenAI Japan, which it established with OpenAI.
The planned work in Malaysia is another example of how these businesses could support one another. SoftBank brings its AI and semiconductor network, Grab provides access to regional customers and real-world uses for the technology, and Sarawak Oil Company PETROS contributes to energy supply and supporting infrastructure. The business idea is to bring computing power, electricity, and actual customer demand together. This could make it more likely that data centers will earn steady income over time. So far, SoftBank has officially disclosed only the framework agreement signed on October 8 to explore the project. It has not yet announced the investment schedule, how the money will be allocated, or how equipment will be purchased for the planned $50 billion investment. The plan is therefore a possible source of new demand worth watching, but it cannot yet be counted fully as SoftBank's own spending or as confirmed orders for chipmakers.
Funding is still available, but it has become more expensive. For semiconductor stocks to make a major recovery, the industry will need a steady flow of orders and strong cash flow.
Subscription data for corporate high-yield bonds issued in September showed that early demand exceeded $20 billion. This suggests that credit investors are still willing to support Masayoshi Son's AI expansion, but they now want much higher returns for taking the risk.
This round of fundraising has reached its final stage. In September, SoftBank announced that it would issue $10 billion in U.S. dollar bonds and €1 billion in euro bonds, worth about $11.1 billion in total. The three U.S. dollar bond groups have annual interest rates of 8.625%, 9.25%, and 9.75%. The rate on the highest-paying group is close to 10%, a level the company had already tested before. Based on the size of the issue and the interest rates, the U.S. dollar bonds alone will cost about $941 million in interest each year. SoftBank later confirmed that it would use the money to complete its additional investment in OpenAI. The Malaysia project still needs its own funding arrangements. Early demand for the bonds shows that SoftBank can still raise money this way, while the final interest rates show how much that money really costs.
Stock and corporate bond markets are already reassessing the risks and possible returns of this expansion. Based on closing prices from September 30 to October 23, SoftBank Group shares in Tokyo fell from 6,430 yen to 4,276 yen, a total drop of about 33.5%. CoreWeave shares fell from $87.12 to $54.24, a drop of roughly 37.7%. On October 22, the Philadelphia Semiconductor Index fell 6.7% in a single day. This showed that worries about funding costs and investment returns had spread across the AI supply chain. Meanwhile, SoftBank's five-year credit default swap (CDS) rate rose to 432 basis points on October 20. This indicates that investors were demanding greater compensation for taking on the company's credit risk.
Yields on 10-year U.S. Treasury bonds have recently risen sharply and remained close to their highest levels in more than two decades. This has added pressure on CoreWeave (CRWV.US), a company at the center of both the AI computing sell-off and concerns about a possible AI bubble bursting. Reports on October 22 said that CoreWeave's bonds due in 2031 had a yield of about 15.76%, while its five-year CDS rate stood at around 1,115 basis points. Earlier, on October 15, media reports said market prices implied a cumulative default probability of nearly 63% over five years for the company's five-year CDS. However, this figure depends on assumptions about how much investors might recover after a default and on the extra returns investors demand for taking risks. It is not an actual forecast that the company will default.
The reason bond investors are pricing in such high risks is that large AI data centers need large amounts of money upfront to build facilities and buy equipment. Income from computing services arrives more slowly as equipment is delivered, customers begin using the services, and contracts are completed. When basic interest rates and the extra interest charged to riskier borrowers rise at the same time, new borrowing becomes more expensive and projects generate less cash after financing costs. SoftBank also faces a timing problem: even if the value of its OpenAI shares rises, that increase does not immediately provide cash to pay its debts.
At the infrastructure level, Muse and Astra, along with Meta's newly open-sourced Muse Spark Open, are expected to help AI agents become more widely used. These systems could create a huge increase in AI tasks that require models to generate answers, use tools, and keep track of information and task progress. They could also increase competitive pressure on some software companies. GPUs and TPUs handle the main calculations needed by AI models. High-performance server CPUs run browsers, virtual machines, product searches, database queries, and the coordination of transactions and tasks. High-bandwidth memory (HBM) and server DRAM store model data, context, and the working environments needed for many tasks to run at once. Enterprise-grade SSDs store product indexes, task logs, and information that must be kept over time. High-speed networks and optical connections allow different systems to exchange data quickly.
Based on this analysis, the Malaysian government's large AI infrastructure plan could provide a major boost to the semiconductor industry. It could increase chip purchases and create more demand linked to AI infrastructure. AI systems that run continuously need AI accelerators, including GPUs, TPUs, and application-specific integrated circuits (ASICs), to run models and generate answers. They also need data center CPUs to manage tools and coordinate tasks; HBM and server DRAM to store models and operating information; and high-performance enterprise SSDs to keep data. High-speed networks connect these parts and allow them to exchange information. If the project moves forward, the need for semiconductors worth billions of dollars could spread across the main parts of the AI computing supply chain. These include accelerated computing, memory, wafer manufacturing, advanced 2.5D and 3D chip packaging, high-performance networking equipment, and optical connections for data centers. Which companies benefit most, and how much of the market they capture, will depend on the technologies chosen. The most important development for investors will be whether the $50 billion plan gradually turns into actual orders for AI data center equipment, supported by funding, reliable power, and firm customer commitments. This would give suppliers a clearer view of future sales and help the market judge more accurately whether the large financing costs faced by developers can eventually be covered by cash earned from computing-intensive services.