🧠 Food for thought

Implications, context, and why it matters.

The $100B “investment” is still a letter of intent

• Nvidia announced a letter of intent to invest up to $100B in OpenAI, and cfo Colette Kress told a UBS event (UBS is a global investment bank) no definitive deal exists two months later 1.
• The plan centers on at least 10 gigawatts (GW, a measure of power capacity equal to 1 billion watts) of capacity 2. Enough for about 4 to 5 million GPUs, with funding released in tranches as systems come online 2.
• One gigawatt of data center build costs $50-60B 2. About $35B goes to Nvidia chips and systems, so the pledge is future hardware buys, not equity 2.
• Nvidia’s 10-Q filing (its quarterly report to the U.S. Securities and Exchange Commission or SEC) says completion is uncertain given build limits; power supply; regulation; capital access 1.

GPU clouds and data centers can meet OpenAI needs during talks

• OpenAI gets Nvidia GPUs through Microsoft and Oracle, which gives cloud partners room to deepen ties before any Nvidia pact 1.
• CoreWeave (a specialized GPU cloud provider; ClusterMAX Platinum rated) and Oracle Cloud Infrastructure (Oracle’s public cloud; ClusterMAX Gold rated) offer networking, managed services, enterprise security for large training 3. ClusterMAX is a benchmarking rubric where Platinum and Gold indicate higher performance tiers 3.
• Gigawatt campuses need $50-60B per site 2. Infrastructure funds and colocation operators (companies that lease power, cooling, and space in shared data centers) can fund it 2.
• GPU rental rates have fallen with more supply, so OpenAI can push for better terms across providers instead of a single Nvidia deal 3.