A look at OpenAI's tangled web of dealmaking

Sam Altman, the CEO of OpenAI, is making headlines as his artificial intelligence company reaches a staggering valuation of $500 billion. Amidst its rapid expansion, OpenAI has secured monumental deals worth tens of billions with various infrastructure partners, even as it faces significant cash expenditures. This influx of investments is reshaping the financial landscape, contributing to record highs for the Nasdaq and S&P 500, particularly after Nvidia's commitment to invest up to $100 billion in OpenAI. This investment follows a previous agreement with Oracle worth $300 billion related to the Stargate initiative, a vast $500 billion infrastructure project also backed by SoftBank. The financial commitments continue to pour in, with CoreWeave recently announcing it will provide OpenAI with up to $22.4 billion in AI infrastructure, a substantial increase from their earlier agreement of $11.9 billion. Additionally, Broadcom has recently indicated a new $10 billion customer, with analysts suggesting that OpenAI is at the center of this development. While OpenAI emphasizes that scaling is essential for driving innovation and unlocking future advancements in AI, some investors and analysts are starting to question the sustainability of these enormous financial commitments and the company's dependency on its intricate web of infrastructure partners. For instance, OpenAI secured a $350 million stake in CoreWeave prior to its IPO in March. Furthermore, Nvidia solidified its financial relationship with OpenAI by participating in a $6.6 billion funding round this October. Reports indicate that Oracle is investing around $40 billion in Nvidia chips essential for powering one of OpenAI's Stargate data centers. CoreWeave also disclosed an order valued at a minimum of $6.3 billion from Nvidia earlier this month. With Nvidia’s $100 billion stake in OpenAI, the tech giant is set to earn both equity and revenue from this partnership. OpenAI is projected to generate approximately $13 billion in revenue this year, according to CFO Sarah Friar, who stressed the necessity of bold investments in infrastructure during technological booms. Altman has expressed a willingness to operate at a loss to enhance growth and strengthen investments. However, analysts are voicing concerns that OpenAI's dealings with Nvidia mirror vendor financing practices that contributed to the dot-com bubble burst in the early 2000s. Nvidia has emerged as a primary beneficiary of the AI surge, producing the critical graphics processing units (GPUs) needed for training models and executing large-scale AI operations. Bespoke Investment Group highlighted the concerning self-referential nature of the current AI ecosystem, noting that if Nvidia must fund its own revenues to sustain growth, the entire sector could be at risk. Peter Boockvar, CIO at One Point BFG Wealth Partners, likened the situation to companies from the late 1990s, while acknowledging that the scale of the current transactions is unprecedented. For this ambitious endeavor to succeed without incurring significant losses, OpenAI and its counterparts must generate substantial revenues and profits to meet their financial obligations while delivering returns to their investors. An OpenAI representative referred to Altman and Friar's recent comments, emphasizing the pursuit of a historic opportunity that demands equal ambition. According to Bain & Company's 2025 Technology Report, the demand for computing power could soar to 200 gigawatts by 2030, necessitating an annual investment of around $500 billion in data centers. This implies that AI firms would need to collectively generate $2 trillion in annual revenue to cover these costs, leaving an $800 billion shortfall if investments in cloud and data centers are maximized. Despite these challenges, Altman remains resolute, dismissing concerns about overextending infrastructure investments. "This is what it takes to deliver AI," he stated, highlighting the unique demands of this technological revolution compared to past iterations of the internet.