From OpenAI to Malaysia, Masayoshi Son Rolls Out the "SoftBank ASI Blueprint"! US$50 Billion Investment Plan Ignites Bullish Sentiment in Semiconductors

SoftBank Group Corp., led by legendary investor Masayoshi Son, is extending its AI investment footprintcentered on AI application leader OpenAI and chip design leader ARMfurther into Southeast Asia's AI computing cluster infrastructure and application markets. Son is accelerating the rollout of his "AI ambitions" toward and promotion of a "full-stack AI ecosystem" encompassing underlying chip architecture, AI computing infrastructure, cloud computing infrastructure, energy, and AI applications. According to the latest media reports, Malaysian Prime Minister Anwar stated that Son plans to invest US$50 billion in the country, covering artificial intelligence, new technologies, and data centers, with sites including Kuching in Sarawak and Peninsular Malaysia. The plan aligns with Malaysia's goal of becoming an "influential AI nation" by 2030, and also provides new demand clues for the global stock market's semiconductor sector, which has come under heavy pressure recently: major investment institutions and cloud computing giants are still expanding the geographic footprint of AI computing infrastructure construction. Whether subsequent procurement of core AI infrastructure equipmentincluding AI chips, memory chips, and data center CPU systemscan exceed expectations depends on the comprehensive implementation of project financing progress, power supply guarantees, and customer demand for computing resources. In 2000, Son first invested approximately US$20 million in Alibaba Group Holding Limited Sponsored ADR, and subsequently through long-term shareholding shared in what can be called an "investment myth"the record-setting massive value growth of a Chinese e-commerce platform as it moved from the initial stage of huge losses to the spotlight of global capital markets. Today, OpenAI, which Son has backed with huge sums, and a host of AI computing leaders, in his view, similarly possess unparalleled growth potential to become large-scale user gateways and enterprise technology platformsthat is, betting on the mutual reinforcement among the most cutting-edge AI chip components, AI large-model capabilities, developer ecosystems, paid services, and enterprise workflows, with future B-end and C-end commercialization space set to extend comprehensively as the scope of AI applications expands. SoftBank plans to invest US$50 billion in Malaysia, which aims to become an "AI nation" Malaysian Prime Minister Anwar Ibrahim said SoftBank Group plans to invest US$50 billion in Malaysia. According to the report, Anwar said on Saturday that after the two held a video conference earlier this week, SoftBank Group founder and leader Masayoshi Son had selected sites in Kuching, Sarawak, and Peninsular Malaysia for the proposed investment. The Malaysian Prime Minister said, according to the report, that the proposed investment would focus on artificial intelligence, new frontier technologies, and large AI data center construction projects. The investment plan comes as Malaysia seeks to become an "influential AI nation" by 2030. The government launched a national AI action plan earlier this year to strengthen the country's AI capabilities and infrastructure. Earlier this month, SoftBank signed a framework agreement with Singapore-headquartered Grab and Malaysian Sarawak oil company PETROS to explore developing AI and digital infrastructure in Sarawak, including AI computing services and advanced technologies. SoftBank is not the only company seeking technology investment in Malaysia. Alphabet Inc. Class C parent Alphabet previously announced in 2024 an investment of US$2 billion to build its first large AI data center and large AI cloud computing region in the country; AWS Cloud Technology, the world's largest cloud computing service provider under e-commerce and cloud computing giant Amazon.com, Inc., said it plans to invest US$6.2 billion by 2038. Microsoft Corporation also recently announced an investment of approximately US$2.2 billion for traditional cloud computing capacity and AI computing infrastructure construction as well as other large digital projects. According to data from the Malaysian Investment Development Authority, from 2021 to the first half of 2026, the country recorded investment associated with large AI data centers of approximately RM385.7 billion, equivalent to about US$94 billion. SoftBank representatives did not immediately respond to media requests for comment. Masayoshi Son extends AI ambitions to AI infrastructure, energy, and applications Son's layout has formed a multi-layered structure connecting AI large models, chip-based AI computing infrastructure, energy, and flagship AI applications such as AI agents. On October 1, SoftBank announced the completion of its final US$10 billion additional investment in OpenAI, bringing cumulative investment to a staggering US$64.6 billion and a stake of about 13%. On the chip side, ARM's central processor architecture and ARM's self-developed data center CPUs, together with the acquired Ampere and Graphcore, expand the computing capability layout; on the infrastructure side, Stargate (the US "Stargate" large AI data center project), SB Energy, and SB Neowhich plans to emulate new cloud providers such as CoreWeave in leasing computing cloud servicesundertake data center construction, energy supporting facilities, and computing supply; on the enterprise AI application side in the Japanese market, SB OpenAI Japan lays out customized agent applications. From the perspective of the AI industry structure, this comprehensive layout aimed at super artificial intelligence (ASI) is intended to simultaneously capture equity appreciation of super platforms and revenue from AI chip-based AI infrastructure business and AI cloud computing resource leasing and sales. The AI super platform SoftBank is now building is no longer a single-point bet, but an increasingly complete "ASI comprehensive stack"that is, at the bottom layer, Arm architecture and Arm's self-developed data center CPUs; at the middle layer, the UK AI chip design company Graphcore Ltd., fully acquired by SoftBank Group, and Ampere, formally acquired not long ago; at the upper layer, OpenAI, Stargate (the "Stargate" project), and the enterprise-level cloud AI computing platform layout jointly with Oracle Corporation, plus SB Neo, a new cloud force (Neocloud); and Son himself has long publicly stated that SoftBank's goal is to become the largest AI computing and application-level foundational platform provider in the "super artificial intelligence" (ASI) era over the next decade, and has adopted a nearly "all in" aggressive investment posture toward ChatGPT developer and global AI large-model leader OpenAI. Oracle Corporation is mainly Stargate's cloud and infrastructure partner, while SoftBank's enterprise-facing joint venture platform is SB OpenAI Japan, jointly established with OpenAI. The Malaysia cooperation further reflects this industrial synergy: SoftBank provides the AI and semiconductor ecosystem, Grab provides the regional customer network and application scenarios, and Sarawak oil company PETROS participates in energy supply and supporting infrastructure. Its business logic is to combine computing power, power resources, and actual usage demand to increase the likelihood that data centers generate sustained revenue. At present, what SoftBank has officially announced is the exploratory framework agreement signed on October 8; the investment timeline, funding structure, and equipment procurement arrangements for the US$50 billion plan have not yet been disclosed, so this is a potential demand increment worth tracking and cannot yet be fully counted into SoftBank's own capital expenditure or chipmakers' orders. The financing window is still open, but more expensive: a semiconductor rally needs orders and cash flow to follow through September's corporate junk credit bond subscription data showed more than US$20 billion in preliminary subscriptions, indicating that credit capital is still willing to support Son's AI expansion, but the required returns have risen markedly. The financing has now reached a final result: in September, SoftBank finalized the issuance of US$10 billion and EUR1 billion in bonds, totaling about US$11.1 billion; the coupons on the three tranches of US dollar bonds were 8.625%, 9.25%, and 9.75%, respectively, with the highest tranche close to the previously tested 10%. Based on issuance size and coupons, the annual interest on the US dollar portion alone is about US$941 million. The company subsequently confirmed that the bond proceeds would be used to complete the additional investment in OpenAI; the Malaysia project still needs to secure its own funding arrangements. The heat of preliminary subscriptions shows that financing channels are available, while the final coupons reveal the true cost of using those funds. Stock and corporate bond markets are already reassessing the risk-reward ratio of this expansion. Based on closing prices from September 30 to October 23, SoftBank Group's Tokyo share price fell from JPY6,430 to JPY4,276, a cumulative decline of about 33.5%; CoreWeave fell from US$87.12 to US$54.24, down about 37.7%. On October 22, the Philadelphia Semiconductor Index fell 6.7% in a single day, indicating that financing and return concerns have transmitted along the AI industry chain. SoftBank's five-year credit default swaps (CDS) rose to 432 basis points on October 20, reflecting investors' demands for more credit risk compensation. As the recent rise in 10-year US Treasury yields keeps them near the highest levels in more than two decades, CoreWeave (CRWV.US)the stock at the very core of the "AI computing selloff storm" and "AI bubble burst" narrativefaces more direct pressure from AI financing costs: a October 22 report showed that its bonds maturing in 2031 yielded about 15.76%, and its five-year CDS was about 1,115 basis points; media reported on October 15 that the market had previously priced the five-year CDS at one point implying a cumulative five-year implied default probability of nearly 63%though this probability includes recovery rate assumptions and risk premiums and is not an actual default forecast. The mechanism behind this pessimistic bond-market pricing is that large AI data center construction and equipment procurement require cash paid in advance, while computing revenue gradually forms with delivery, customer usage, and contract performance; rising benchmark rates and credit spreads together raise new financing costs and compress project cash returns. SoftBank also faces another layer of time mismatch: the increase in the value of its OpenAI equity will not simultaneously become cash available to pay interest. From the perspective of the underlying architecture of AI computing infrastructure, Muse and Astra, as well as Meta's Muse Spark Open, are expected to jointly further expand the scope of AI agent use. The massive expansion of AI inference workloads they bring will simultaneously increase demand for model inference, tool execution, and state management, while deepening competitive pressure on some software companies. GPUs, TPUs, and other accelerators handle model computation; high-performance server CPUs run browsers, virtual machines, product search, database queries, and transaction orchestration; HBM and server DRAM carry model data, context, and concurrent working environments; enterprise-grade SSDs store product indexes, task records, and persistent state, while high-speed networks and optical interconnects support distributed data exchange. On this basis, the positive significance of the Malaysian government-led large AI infrastructure plan for semiconductors lies in further expanding the potential scale and sources of demand for semiconductor procurement closely related to AI infrastructure. Continuous operation of agents requires AI accelerators such as AI GPUs/TPUs/AI ASICs to complete model computation, data center CPUs to execute tools and task orchestration, HBM and server DRAM to carry models and working states, enterprise-grade high-performance SSDs to store data, and high-speed networks to exchange information. Once the project is implemented, these tens-of-billions-of-US-dollars-level semiconductor demands are expected to transmit along the deepest core AI computing industry chain links, including accelerated computing, memory, wafer manufacturing, 2.5D/3D advanced packaging, high-performance networking equipment, and data center optical interconnects; the specific beneficiary manufacturers and shares will still depend on technology choices. The most investment-valuable progress is the gradual conversion of the US$50 billion plan into AI infrastructure-related data center equipment orders backed by funding, power, and customers. This can both enhance suppliers' revenue visibility and allow the market to more accurately judge whether the persistently high financing costs borne by builders can ultimately be covered by computing operating cash flow.