The artificial intelligence revolution is transforming into a high-stakes financial battle where balance sheets matter more than algorithms. As Silicon Valley's AI rivalry intensifies, the alliances built around Sam Altman's OpenAI are showing dangerous cracks, with investors questioning the sustainability of their massive bets.
Partners Under Pressure
Recent market movements reveal growing investor anxiety about OpenAI's key partners. SoftBank's shares have plunged 40% since late October, while Oracle's stock has surrendered all gains made since early September. The September rally came after Oracle announced a staggering $300 billion computing deal with OpenAI, suddenly giving the legacy database company an AI halo.
The three companies are collaborating on the ambitious $500 billion Stargate AI infrastructure project, which aims to construct data centers across the United States. Additionally, in March, SoftBank Chairman Masayoshi Son secured a venture capital agreement with Altman, committing to invest $30 billion by year-end.
The Valuation Question
Investors are now reevaluating OpenAI's dominance following Alphabet's release of its multipurpose Gemini 3 model, which received glowing reviews. The central concern revolves around how a potentially weakened OpenAI might affect its partners' financial health.
SoftBank's situation appears particularly precarious. The Japanese conglomerate values OpenAI at $500 billion, with its stake accounting for just over 20% of its net asset value. However, if OpenAI fails to maintain this valuation in future funding rounds, SoftBank faces substantial losses. The company reported its best earnings in three years during the September quarter, thanks largely to a $12.8 billion fair value gain from its OpenAI shares - gains that would need reversal if the startup's valuation drops.
As an investment management company, SoftBank is assessed based on NAV, with investors closely monitoring portfolio liquidity and cash distribution capabilities. According to Bloomberg Intelligence, private holdings, including the 11% OpenAI stake, now constitute about 36% of SoftBank's NAV, up from 21% at the beginning of 2024.
Costly Strategic Shifts
To secure his position at Altman's table, Son has been aggressively selling liquid assets. He has completely exited his $5.8 billion stake in Nvidia and sold $9.2 billion worth of T-Mobile US shares. Combined with new bond sales, SoftBank raised $30 billion - precisely the amount needed to complete the OpenAI deal.
Market skepticism about Son's investment strategy is evident, with SoftBank's stock currently trading at a 32% discount to its NAV.
Oracle's position might be even more concerning. The fundamental question of how OpenAI will pay for its $300 billion computing contract remains unanswered. The five-year deal commencing in 2027 coincides with OpenAI's projected $60 billion in sales for that year - a rosy outlook that investors are questioning given Google's Gemini 3 emergence.
Meanwhile, Oracle is making massive capital commitments to meet the projected demand. For the Stargate project alone, Ellison has committed over $100 billion to lease data-center shells, with these obligations appearing on financial statements over the next three years. The company's debt-heavy approach reflects its strategy to save upfront costs by leasing land while concentrating spending on chip purchases. As of August, Oracle carried $105 billion in total debt.
The Cash Game Reality
The AI competition is increasingly becoming a financial endurance test. While Alphabet maintains negative net debt despite repeated bond sales this year, generating approximately $150 billion in annual operating cash flow with nearly $100 billion in cash reserves, OpenAI's allies are taking different paths.
OpenAI requires continuous venture capital fundraising to cover expenses, SoftBank sold quality assets to invest in the unicorn, and Oracle turned to debt-financed land acquisitions to accelerate data-center construction. In this high-stakes environment, smart money appears to be favoring established cash cows over ambitious but financially strained newcomers.
The situation highlights the enormous capital requirements of AI infrastructure development. Altman recently stated he doesn't want US guarantees for his startup's data centers but suggested the government might build and own its own AI infrastructure - a telling acknowledgment of the staggering costs involved in this technological race.
