Masayoshi Son’s Admission of “Selling NVIDIA Shares with Tears” Reveals the Truth About AI Investment
Editor
Masayoshi Son, Chairman of SoftBank Group, admitted to “selling NVIDIA shares with tears” at the FII Priority Asia Forum held in Tokyo on December 1, 2025, causing a significant stir in the AI investment market. As one of the world’s largest investors, Son’s decision to sell his entire stake in NVIDIA, the company with the highest market capitalization, and his strong rebuttal of the AI bubble theory vividly illustrate the complex situation of the current AI investment market.
Last month, SoftBank secured a massive $5.8 billion, approximately 8.5 trillion won, by selling its entire stake in NVIDIA. This amount is comparable to the annual revenue of a major domestic corporation. However, according to Son, this was not due to a lack of trust in NVIDIA but rather an inevitable choice to secure funds for new AI investment opportunities, particularly in OpenAI.
His statement, “If I had unlimited money, I wouldn’t have sold a single share,” accurately reflects the dilemma of the current AI investment market. While there are abundant promising investment opportunities, even investors with vast capital cannot seize every opportunity simultaneously. In fact, SoftBank’s Vision Fund was established by raising $45 billion from Saudi Arabia’s Public Investment Fund (PIF), yet it is still insufficient to cover all AI investment opportunities.
More noteworthy is Son’s strong rebuttal of the AI bubble theory. Using the somewhat provocative expression, “Those who claim it’s a bubble are not smart enough,” he presented specific economic grounds. His logic is that if AI eventually generates 10% of the global GDP, the current cumulative investment of trillions of dollars will be more than recouped.
The Current State and Scale of the AI Investment Market
Examining Son’s claims with specific figures reveals an interesting picture. Considering that the global GDP in 2024 is approximately $105 trillion, AI generating 10% of the GDP means creating more than $10 trillion in economic value annually. Even if the cumulative investment in the AI sector so far is in the trillions, considering such economic ripple effects, it is not an excessive investment.
In fact, NVIDIA’s market capitalization as of December 2025 exceeds $3 trillion, equivalent to about 12% of the U.S. GDP. The fact that a single company’s value has reached this level can be seen as an indicator of the AI market’s potential. Notably, NVIDIA’s data center revenue recorded $22.6 billion in the fourth quarter of 2024, a 427% increase compared to the same period last year, suggesting that this growth is based on substantial demand rather than a bubble.
Looking at the domestic situation, memory semiconductor companies like Samsung Electronics and SK Hynix are direct beneficiaries of the AI boom. For Samsung Electronics, the memory division’s operating profit turned positive in the third quarter of 2024 compared to the same period last year, and SK Hynix’s stock price rose more than 30% since the beginning of the year due to a surge in demand for HBM (high-bandwidth memory). These tangible achievements serve as evidence that AI investment is not just a bubble.
However, despite Son’s optimism, there are clear risks in the AI investment market. The biggest issue is the time lag between investment and actual profit generation. In the case of OpenAI, its revenue in 2024 is estimated at $3.4 billion, but the cumulative investment exceeds $13 billion. The profitability relative to investment has not yet been clearly demonstrated.
Competitive Landscape and Market Dynamics
Looking at the current competitive landscape of the AI market, NVIDIA holds a dominant position in hardware, but there is fierce competition in the software and service sectors. Google’s Gemini, Microsoft’s Copilot, Meta’s Llama, and OpenAI’s GPT series each show competitiveness in different areas. This diversified competitive structure can be interpreted as a positive sign of market health.
Particularly noteworthy is that each company is pursuing different strategies. NVIDIA focuses on hardware infrastructure, OpenAI concentrates on developing general AI models, and Google and Microsoft focus on integration with existing services. These differentiated approaches increase the likelihood of creating unique value in each area.
Son’s decision to sell NVIDIA shares and invest in OpenAI can be understood in this context. While hardware infrastructure has reached a certain level of maturity, the application software and service sectors are still in the early stages. In fact, OpenAI’s ChatGPT secured 100 million monthly active users within two months of its launch, becoming the fastest-growing consumer application in history.
However, there are differing opinions on whether this growth is sustainable. The computing costs required for AI model training and operation are increasing exponentially, making profitability difficult to achieve. In OpenAI’s case, it is estimated that monthly operating costs reach $700 million, making revenue growth and cost efficiency improvement crucial tasks.
Domestic companies’ response strategies are also varied. Naver provides Korean-specialized AI services through HyperCLOVA X, Kakao focuses on AI technology development through Kakao Brain, LG Electronics integrates AI functions into home appliances, and Hyundai Motor uses AI for autonomous driving technology development. Attempts to leverage AI in their areas of strength are increasing among companies.
For Son’s prediction that “AI will generate 10% of GDP” to become a reality, AI utilization in these various areas must lead to substantial productivity improvements. So far, achievements have mainly appeared in areas like content creation and customer service automation, but full-scale adoption in traditional industries such as manufacturing, healthcare, and finance is still in the early stages. Success in these areas will be a key factor in determining the long-term profitability of AI investments.
Ultimately, Son’s admission of “selling with tears” symbolically represents the complex reality of the current AI investment market. Opportunities are limitless, but resources are limited, and while long-term prospects are bright, short-term uncertainties remain significant. Nevertheless, his strong rebuttal of the AI bubble theory can be interpreted as a demonstration of confidence in the long-term potential of this field. Watching how these investments translate into substantial economic value over the next few years will be crucial.
This article was written after reading a Seoul Economic Daily article, with personal opinions and analysis added.
Disclaimer: This blog is not a news outlet, and the content reflects the author’s personal views. Responsibility for investment decisions lies with the investor, and no liability is accepted for investment losses based on the content of this article.
