OpenAI Uses Complex and Circular Deals to Fuel Its Multibillion-Dollar Rise (www.nytimes.com)

OpenAI has financed its rapid expansion with a web of large, often circular deals that blur the line between investment and vendor payment. Microsoft invested more than $13 billion from 2019–2023 and then received much of that back as cloud-compute contracts; when Microsoft capacity fell short, OpenAI contracted Oracle, CoreWeave and others. Recent agreements include more than $22 billion in multi-year commitments to CoreWeave (paired with $350 million in CoreWeave stock), a SoftBank-led $40 billion investment plus fundraising to build data centers, Oracle pledges to spend some $300 billion building facilities OpenAI will then lease, a roughly $20 billion G42 data-center build in the UAE, Nvidia’s stated intent to invest $100 billion, and an AMD option letting OpenAI buy ~160 million shares at a penny apiece (about 10% of AMD). OpenAI earns significant revenue from ChatGPT and developer tools but still runs at a loss. For the AI/ML community this financial engineering matters as much as hardware or models: it’s enabling massive, near-term increases in specialized compute (chips, GPU clusters, bespoke data centers), accelerating model training and capabilities, but concentrating risk across a few vendors and highly leveraged data-center builders. If AI-driven revenue growth continues, these bets could support long-term infrastructure scale; if progress stalls, partners carrying heavy debt (especially smaller providers) could face insolvency, creating supply shocks for GPUs, raising costs for researchers and startups, and potentially amplifying systemic financial fallout.