A quarter of Nvidia's business next year comes from labs it is financing
Nvidia has invested nearly $50 billion into AI labs purchasing its chips and partnered with six major investment firms (Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, KKR) to raise over $500 billion in outside capital for data center construction The arrangement creates a circular financing loop where Nvidia invests in labs, labs build data centers filled with Nvidia chips, and purchases are recorded as Nvidia revenue, though the company rejects this label OpenAI's compute commitments
Analysis
TL;DR
- Nvidia has invested nearly $50 billion into AI labs purchasing its chips and partnered with six major investment firms (Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, KKR) to raise over $500 billion in outside capital for data center construction
- The arrangement creates a circular financing loop where Nvidia invests in labs, labs build data centers filled with Nvidia chips, and purchases are recorded as Nvidia revenue, though the company rejects this label
- OpenAI's compute commitments through 2030 are estimated at ~12 gigawatts, with an unnamed second lab receiving credit support for nearly 2 gigawatts, all hosted on SB Energy facilities designed exclusively for Nvidia equipment
- Nvidia guides to $108 billion in revenue this quarter with ~70% growth expected through January 2028, though memory price inflation is pressuring margins down to 71-72% by Q4
- Jensen Huang claims AI has "tipped over" to being mostly agentic, with agents requiring 15-100x the computing power of human users, driving unprecedented demand for Nvidia's platform
Why It Matters
This represents a fundamental shift in how AI infrastructure is financed, with Nvidia effectively acting as both supplier and financial backer to the labs driving AI development. The circular financing model raises important questions about concentration of risk, market sustainability, and Nvidia's exposure if demand softens or labs fail to deliver on commitments.
Technical Details
- Nvidia has signed partnerships with six investment firms to create financing platforms targeting $500 billion in outside capital; these partnerships remain subject to definitive agreements with no binding contracts yet signed
- SB Energy facilities will host exclusively Nvidia equipment, with Phase 1 supporting 4.25 gigawatts already allocated to OpenAI
- Nvidia's revenue model includes dual income streams from equipment sales and rental income sharing with smaller cloud operators who receive guaranteed floor payments
- Jensen Huang positioned Nvidia's platform as cloud-agnostic across the full AI system lifecycle, contrasting it with rival chips designed for single services like OpenAI's Jalapeño processor
- Memory scarcity driven by AI buildout is causing price increases faster than expected, with margins guided to bottom at 71-72% in Q4
Industry Insight
- The circular financing model creates significant counterparty risk for Nvidia; if any backed lab fails, the company loses both its investment and the chip sale simultaneously, though Nvidia argues hardware can be redirected while demand exceeds supply
- The $500 billion figure represents intention rather than committed capital, suggesting the actual scale of infrastructure buildout may be materially lower than implied by current announcements
- Nvidia's dual revenue model (equipment sales plus rental income share) creates misaligned incentives where the company benefits regardless of whether labs achieve operational profitability, potentially encouraging overbuilding
- The claim that AI has "tipped" to being mostly agentic lacks supporting data and appears designed to justify continued massive capital expenditure commitments from investors and partners
Disclaimer: The above content is generated by AI and is for reference only.