SoftBank's $5.8B Nvidia Exit to Fund Its OpenAI Bet

What SoftBank sold, and the reason it gave

SoftBank disclosed at its November 11, 2025 earnings call that it had sold all 32.1 million shares of Nvidia it still held, for $5.8 billion, with the trades executed over the course of October. It was SoftBank's full remaining position: after the sale, SoftBank owned no Nvidia stock at all. [1][2]

CFO Yoshimitsu Goto stated the reason without hedging: "This year our investment in OpenAI is large, more than $30 billion needs to be made. For that, we do need to divest our existing assets." He added that the decision had "nothing to do with Nvidia itself," describing it as ordinary portfolio rotation rather than a view on the chipmaker's prospects. SoftBank's board had separately approved a $22.5 billion second installment toward OpenAI that October, and the company expected its total OpenAI commitment to reach $34.7 billion by the end of the year. [1][2]

Selling before the top

SoftBank's disclosed numbers, $5.8 billion for 32.1 million shares, work out to an average sale price of roughly $182 a share. Nvidia spent the back half of October climbing past $200 a share and, on October 29, closed above a five trillion dollar market capitalization for the first time, the highest valuation any public company had reached. [5][3]

Atlas interpretation: SoftBank has not disclosed which trading days in October the block actually moved on, so it is not possible to say it missed the exact October 29 close by some number of days. What the arithmetic does show is that its average print sits meaningfully below where Nvidia finished the month: an all-in exit executed steadily across October, rather than concentrated at the end, would have captured less of the run than shares held into the $5 trillion close would have. The sale was disclosed twelve days after that peak, once the shares were already gone. [5][3]

Selling the chipmaker to fund the customer

Atlas interpretation: The specific irony is that Nvidia's chips are the thing the entire AI capital expenditure boom is priced on, the boom that had just pushed Nvidia itself past five trillion dollars, and SoftBank chose that month to exit the stock entirely in order to fund OpenAI, one of the largest buyers of exactly those chips. SoftBank was not diversifying away from AI. It was consolidating an AI bet, moving proceeds from the supplier whose valuation the boom had already re-rated into the customer whose future spending is supposed to justify that valuation going forward. [1][3]

Atlas interpretation: The loop has another turn worth naming without overstating it. Nvidia had separately signed a letter of intent to invest up to $100 billion into OpenAI the previous month, money that was itself explicitly tied to OpenAI buying Nvidia hardware. SoftBank's Nvidia sale did not fund that specific arrangement; it funded SoftBank's own, separate OpenAI stake. But the two events together mean that in the same autumn, both Nvidia and one of Nvidia's largest shareholders were independently moving capital toward OpenAI, one by investing directly, the other by liquidating Nvidia stock to do it. Two different mechanisms arriving at the same conclusion is not proof of a single circular scheme, but it is a fair description of how concentrated the money behind one company had become. [1]

Son's pattern, and this exact trade before

Masayoshi Son's record is built on scaled, leveraged, high-conviction bets rather than diversified ones. His roughly $20 million into Alibaba in 2000 was worth an estimated $58 billion when Alibaba went public in 2014, the trade that anchors his reputation. WeWork went the other way, ending in a collapse that cost SoftBank's Vision Fund tens of billions of dollars. Both outcomes came from the same instinct: commit heavily to one conviction and let it run, rather than hedge it. [1][6]

This was also not SoftBank's first full Nvidia exit. It built roughly a $4 billion stake starting in 2017 and sold out completely in 2019, near the bottom of a chip downturn and years before the generative AI rally, for about $3.6 billion. Son later called those shares, which would have been worth more than $150 billion had SoftBank held them, "the fish that got away" at a 2024 shareholder meeting, an unusually candid admission of regret from an executive not known for them. [1][3][7]

Atlas interpretation: Son sold the same stock in the same way twice, both times to fund a different high-conviction bet rather than out of caution, and both times ahead of a run higher in the shares he gave up. His stated defense of that pattern, offered around this sale, was blunt: "SoftBank's position is that the risk of not investing is far greater than the risk of investing." That is a real articulation of his long-term case for OpenAI over Nvidia, not a rationalization applied after the fact, but it is also the identical logic that produced the 2019 exit he later called his biggest regret. [6]

How the market read SoftBank's balance sheet

SoftBank's own stock fell as much as 10 percent in Tokyo trading the day after the disclosure, closing 3.5 percent lower, despite the same earnings report showing quarterly net profit more than doubling on the back of AI-related valuation gains. Nvidia's shares dipped nearly 3 percent in the US on the news itself. The larger move landed on SoftBank, not Nvidia, which is a signal the market was reacting less to what the sale said about Nvidia's prospects and more to what repeated asset sales say about SoftBank's own financing position. [4][3]

MST Financial analyst David Gibson quantified the concentration concern directly: SoftBank had committed roughly $113 billion across its investments against only about $58.5 billion in funding capacity, a gap that pushes the company toward margin loans against remaining holdings and further asset sales rather than fresh outside capital. The proceeds from this Nvidia sale were earmarked not only for the OpenAI installment but also for other October-quarter commitments, including a stake in chipmaker Ampere. [1][4]

Atlas interpretation: The concentration risk analysts were pointing to is straightforward: a company that just converted its remaining stake in the world's most valuable chipmaker into a bigger stake in one AI lab has replaced a liquid, publicly priced, diversified-by-customer holding with an illiquid, single-counterparty one. Nvidia's value depends on many buyers; OpenAI's depends on one company's execution and continued access to financing of its own. SoftBank's earnings that quarter looked strong on paper because Nvidia's rally and OpenAI's rising private valuation both flowed through its books, the same concentration that had just produced the gain being the thing the stock sell-off was pricing as risk going forward. [4][1]