Masayoshi Son adds $30 billion in funding for OpenAI, credit rating may be downgraded

SoftBank is in talks to add an additional $30 billion investment in OpenAI, having already sold $5.8 billion worth of NVIDIA stock and reduced holdings in T-Mobile for liquidity. OpenAI’s valuation is expected to reach $750 billion, surpassing Arm to become SoftBank’s largest holding (accounting for over 30%). S&P has warned that the BB+ credit rating faces downward pressure. Following the news, SoftBank’s stock price rose by 8.8%. OpenAI has launched a research tool called Prism.

SoftBank Asset Restructuring: Selling NVIDIA to Buy OpenAI

According to multiple informed sources, SoftBank is currently evaluating the possibility of investing up to an additional $30 billion in OpenAI. Negotiations are ongoing, and the final amount and terms have not yet been finalized. If this financing deal goes through, SoftBank’s stake in OpenAI will increase further. In December last year, SoftBank invested $22.5 billion in OpenAI, raising its stake to about 11%, making it one of OpenAI’s largest external shareholders.

According to The Wall Street Journal, this round of additional investment is part of a larger fundraising plan by OpenAI, which is trying to raise between $50 billion and $100 billion from global investors, with a target valuation of up to $750 billion. OpenAI is considering an initial public offering (IPO) and plans to raise funds from Middle Eastern sovereign wealth funds and other venture capital funds. Current investors include Thrive Capital, Khosla Ventures, and the UAE fund MGX.

For SoftBank, this is not just a financial investment but a strategic bet. Over the past few years, Masayoshi Son has repeatedly emphasized that “AI will reshape everything,” and he views OpenAI as the company most likely to be at the core of this transformation. To raise enough capital to continue investing in OpenAI, SoftBank is undertaking a series of aggressive asset adjustments.

The Wall Street Journal reports that to fund its investment in OpenAI, SoftBank has sold its NVIDIA shares for $5.8 billion (approximately RMB 40.29 billion). At the same time, SoftBank has reduced holdings in assets like T-Mobile and used Arm stock for financing operations. Previously, SoftBank also paused negotiations to acquire U.S. data center operator Switch, focusing funds on AI-related projects.

This asset reallocation logic is intriguing. NVIDIA is the absolute leader in AI computing infrastructure, with its stock price increasing tenfold over the past two years. Selling at a high point to realize gains is understandable. However, from a strategic perspective, Masayoshi Son’s choice to “sell tools and buy applications” reflects a bet that the upstream AI value chain (OpenAI’s models and applications) has more long-term value than downstream hardware (NVIDIA). This judgment may be based on his belief that NVIDIA’s monopoly position will be challenged by competitors like AMD and Google TPU, while OpenAI’s leadership in large language models is harder to surpass.

Over the past year, SoftBank’s investments in AI and automation have accelerated significantly: investing $6.5 billion to acquire U.S. chip design company Ampere Computing, $5.4 billion to acquire ABB’s robotics business, and continuing to increase its stake in OpenAI. These moves demonstrate Masayoshi Son’s effort to build a complete AI industry chain empire—from chip design (Ampere) to physical execution (ABB robots) to intelligent brains (OpenAI)—forming a closed loop.

Credit Rating Crisis and 30% Shareholding Risks

S&P Global has issued a warning that SoftBank’s active investments in AI, combined with Arm’s stock price volatility, are putting pressure on SoftBank’s credit rating. If OpenAI’s valuation constitutes a large proportion of SoftBank’s assets, it could further amplify portfolio risk.

Bloomberg industry analyst mentioned that if SoftBank Group invests an additional $30 billion in OpenAI, its BB+ credit rating could face downward pressure. Considering the incomplete nature of the transaction and assuming SoftBank’s current holdings in OpenAI are revalued upward, the loan-to-value (LTV) ratio could reach 35%, triggering a rating downgrade. To keep the LTV below 25% under disclosed accounting standards, SoftBank would likely need to raise at least $15 billion through asset sales and margin loans.

Three Risks Facing SoftBank

Concentration Risk: OpenAI’s valuation may surpass Arm, becoming SoftBank’s largest holding, accounting for over 30% of its total assets. Failure of this single target could severely damage the group.

Liquidity Risk: An LTV of 35% could trigger credit rating downgrades, increasing financing costs and affecting other investment plans.

Competitive Risk: Google Gemini, Anthropic Claude, and Chinese models like DeepSeek are rapidly catching up, challenging ChatGPT’s leading advantage.

Meanwhile, SoftBank’s portfolio risk will further increase. Based on Arm’s stock price as of January 27, OpenAI’s valuation could exceed Arm, making it SoftBank’s largest single holding, with its proportion of total assets possibly exceeding 30%. Japanese stock strategist Amir Anvarzadeh said, “Masayoshi Son has clearly gone all-in, betting everything on ChatGPT.”

From a competitive perspective, OpenAI’s prospects are no longer as certain as they were a year ago. Models like Google’s Gemini, Anthropic’s Claude, and China’s DeepSeek are rapidly catching up, shrinking OpenAI’s technological moat. The company also faces a stark reality: burning cash at an extremely fast rate, with training requiring massive computing power, inference costs continuously rising, and fierce competition among top researchers. These are all risk factors that SoftBank must consider in its additional investment.

Prism Tool Targets Scientific Research Market

Alongside the fundraising news, OpenAI has also released an important product signal. On January 27, OpenAI officially launched a free tool called Prism, positioned as an AI assistant for scientists, available to any user with a ChatGPT account.

Unlike traditional chat interfaces, Prism functions more like an “AI-enhanced scientific research writing and collaboration platform”: based on GPT-5.2, natively supporting LaTeX, usable for paper writing, editing, literature retrieval, supporting multi-user collaboration, and capable of quickly converting hand-drawn sketches into standard diagrams. OpenAI states that Prism is not meant to replace scientists but to accelerate their workflows. Company executives compare it to a “Cursor or Windsurf” in scientific research—tools deeply embedded in workflows that significantly improve efficiency.

OpenAI’s push of Prism into the scientific research field is not accidental. Data shows that ChatGPT receives an average of 8.4 million messages related to advanced science and mathematics weekly, and this number is expected to grow by 47% by 2025. OpenAI’s Vice President of Science Kevin Weil said, “I believe that 2026 will be for AI and science what 2025 is for AI and software engineering.” In his view, AI-assisted scientific research is at a similar “pre-explosion” moment for code assistants.

Looking at both funding and products, it is clear that OpenAI is following a very defined path: on one hand, securing model scale, computing power, and talent through massive funding; on the other hand, embedding AI deeply into high-value workflows such as scientific research, education, and enterprise through tools like Prism. SoftBank’s large investments in OpenAI are betting on the ultimate possibility of AI transforming industries.