A look at OpenAI's tangled web of dealmaking (www.cnbc.com)

OpenAI has spent the last year signing enormous infrastructure deals—Nvidia agreed to invest up to $100 billion and take equity, Oracle reportedly committed to a roughly $300 billion Stargate arrangement (part of a $500 billion program also backed by SoftBank), CoreWeave raised its OpenAI deal to $22.4 billion, and Broadcom has been tied to a new large OpenAI-linked customer. Those partnerships give OpenAI access to huge GPU capacity and data‑center buildouts while funneling massive orders and revenue opportunities to suppliers; OpenAI itself forecasts roughly $13 billion in revenue this year even as executives say sustained losses and aggressive capital spending are necessary to scale models and drive innovation. The scale matters: GPUs and data centers are the binding constraint for large models, and Bain projects compute demand could hit ~200 GW by 2030—implying roughly $500 billion a year in data‑center costs and a need for roughly $2 trillion in combined industry revenues to sustain that buildout. Analysts warn the financing structure—vendor-backed capital, equity-for-infrastructure and interlocked contracts—resembles vendor‑financing dynamics from the dot‑com era and could be unsustainable if revenues don’t materialize. OpenAI’s leadership argues the bets are essential to deliver future AI breakthroughs, but the deals concentrate market power around a few suppliers (notably Nvidia) and raise systemic risk if growth or profits fall short of these unprecedented commitments.