Berkshire holds $397.4B cash while SoftBank bets $64.6B on OpenAI amid 163.4T yen debt

On May 2, 2026, the atmosphere in Omaha shifted palpably as the annual Berkshire Hathaway shareholder meeting drew only slightly more than half of its 18,000 available seats, a stark contrast to previous years where demand saturated local hotels. At 95 years old, Warren Buffett did not preside over the event; instead, new CEO Greg Abel faced investors, fielding inquiries regarding the conglomerate's unprecedented cash hoard of nearly $400 billion.

Concurrently, 6,000 miles away in Tokyo, Masayoshi Son's team executed a divergent strategy, packaging SoftBank Group's unprofitable AI assets into a new entity, Roze AI, targeting a $100 billion valuation for a planned US IPO in late 2026. This aggressive maneuver is driven by the necessity to fund a $64.6 billion commitment to OpenAI, a figure that could eventually swell to nearly $100 billion. One entity sits on $397.4 billion in cash awaiting a market crash, while the other shoulders a parent company debt of 163.4 trillion yen, betting heavily that the market will not correct.

The magnitude of Berkshire's liquidity is transformative, representing nearly 40% of its total market value and doubling the average cash levels maintained by Buffett over the past two decades. Of this total, $339.3 billion is deployed directly in U.S. short-term Treasury bonds, positioning Berkshire as one of the largest non-government creditors to the U.S. Treasury. This accumulation was not passive but the result of active divestment over fourteen consecutive quarters. Apple, historically the largest holding, saw its position reduced for four straight quarters starting in Q3 2024, with nearly 688 million shares sold to realize over $100 billion. Data compiled by Woofun AI shows that Buffett's rationale remains consistent: he cannot find attractive valuations, describing the current market in a 2024 letter as unappealing and likening recent sentiment to a casino attached to a church. This judgment has persisted for over a year, resulting in Berkshire's stock underperforming the S&P 500 by approximately 40 percentage points over the last twelve months, a relative drawdown comparable only to the 1999 internet bubble peak.

Greg Abel, a Canadian accountant who rose through heavy asset sectors like utilities and railroads, now faces a dilemma distinct from his predecessor. He acknowledges he is not Buffett, emphasizing continuity and decentralization in his first shareholder letter while rejecting any notion of splitting the group. His challenge lies in the inability to deploy the massive cash pile due to high market valuations, coupled with investor dissatisfaction as they vote with their feet. If the market continues its ascent over the next five to ten years, Abel will eventually confront the question of capital return, potentially through special dividends or the breakup of the conglomerate's 60+ subsidiaries. While Berkshire will not face sudden bankruptcy due to its diversified assets and low debt, its identity as a compounding machine may erode in an environment where valuations consistently exceed 30 times earnings and traditional models fail.

In sharp contrast, SoftBank's strategy is defined by aggressive leverage and high-stakes betting. On February 27, 2026, the company confirmed a cumulative investment in OpenAI of $64.6 billion for a 13% stake, marking the era's most expensive single bet. To finance this, SoftBank's interest-bearing debt surged from 121.4 trillion yen in March 2025 to 163.4 trillion yen by December 2025. The parent company's reported cash of 38 trillion yen includes nearly one-third in unused committed credit lines rather than liquid funds. To bridge the gap, SoftBank borrowed $20 billion against Arm holdings and raised $7.7 billion by pledging shares of its Japanese telecom subsidiary SBKK. On March 27, 2026, an unprecedented $40 billion bridge loan was secured from a consortium including JPMorgan, Goldman Sachs, and Mizuho, with $30 billion earmarked specifically for OpenAI co-investment. Woofun AI notes that this loan carries a 12-month term, requiring full repayment by March 2027.

The desperation to fund this venture has led to a liquidation of core assets. Son admitted to selling Nvidia shares in October 2025, cashing out $5.8 billion despite his reluctance, stating he cried while executing the trade. Further divestitures included 56.9 million T-Mobile shares for $12.7 billion in the first three quarters of fiscal 2025, followed by another 12.5 million shares for $2.3 billion in Q4. Holdings in Deutsche Telekom and Alibaba were fully divested, and by April 2026, SoftBank arranged a $10 billion margin loan secured by OpenAI equity at an interest rate near 8%. Simultaneously, the company issued bonds at record costs: a 500 billion yen bond at 3.98% in November 2025, followed by a 418 billion yen subordinated bond at 4.97% in April 2026, the highest coupon in Japanese non-financial corporate history. Woofun AI analysis suggests these rising costs signal growing market skepticism, evidenced by SoftBank's five-year credit default swaps surging to 355 basis points in early March, an 11-month high.

SoftBank's survival hinges on three correlated triggers that could precipitate a violent collapse. The first is OpenAI's IPO timeline; while CEO Sam Altman targets Q4 2026, CFO Sarah Friar suggests a delay until 2027, creating internal uncertainty that could derail the liquidity event needed to service the $40 billion bridge loan. The second trigger is Arm, currently valued at $200 billion with SoftBank holding 87%, which serves as the primary collateral for a $20 billion mortgage. Although Arm's royalty revenue grew 26% year-on-year in Q3 2026, a reversion from its current 70 times forward P/E ratio to a normal semiconductor valuation would collapse the loan coverage ratio. The third trigger is refinancing; with bond coupons rising from 3.98% to 4.97%, the cost of rolling over debt increases annually. While each trigger individually may seem manageable, their high correlation creates a systemic risk where a delay in OpenAI's IPO or a repricing of Arm could trigger a chain reaction, forcing SoftBank to sell assets at distressed prices or face insolvency.

Ultimately, the trajectories of these two giants represent a bifurcation in capitalist strategy. Berkshire's "death" will be a gentle erosion of its mythos as a value investing fortress, slowly transforming into a cash-heavy holding company under Abel's rational management. SoftBank, conversely, faces a binary outcome: if the AI bubble holds, Son will be deified with a $100 billion Roze AI valuation and a realized ASI narrative by age 70; if it bursts, the leverage will trigger a catastrophic failure. The market has effectively placed a bet on which philosophy will survive the next decade, with Berkshire waiting for the crash and SoftBank betting against it.