1. The so-called "circular AI transaction" refers to NVIDIA selling GPUs to AI companies, cloud providers, and data centers, while simultaneously providing financial support to customers through equity investment, financing guarantees, and computing

2026-07-28

1. The so-called "circular AI transaction" refers to NVIDIA selling GPUs to AI companies, cloud providers, and data centers, while simultaneously providing financial support to customers through equity investment, financing guarantees, and computing power purchase commitments. A typical closed loop is: NVIDIA invests in or guarantees customers → customers obtain financing → purchase NVIDIA GPUs or lease related computing power → NVIDIA recognizes revenue → revenue and valuation increase, leading to further investment in more customers. 2. NVIDIA is therefore no longer just a chip supplier, but is gradually becoming a customer shareholder, financing supporter, computing power purchaser, and guarantor of unsold computing power. This model facilitates the rapid expansion of AI infrastructure and is not inherently fraudulent. GPUs, data centers, and computing power services may all be real, and related revenue may comply with accounting rules. 3. The core risk is that some demand may be primarily driven by financing rather than the actual cash flow of end customers. Without continued support from NVIDIA and the capital markets, customers may not maintain the same scale of GPU purchases. High levels of cross-shareholding, cross-purchasing, and cross-financing among industry players can amplify systemic risks. If AI revenue falls short of expectations or financing tightens, a negative feedback loop could form: project cancellations – declining GPU orders – valuation drops – credit contraction, with Nvidia at the heart of this storm. 4. "Big Short" Michael Burry previously warned that this model is similar to supplier financing during the dot-com bubble: the technology may be real, but demand may have been prematurely exhausted by credit. He therefore shorted Nvidia and the AI infrastructure chain. 5. The most crucial indicator is not the total transaction amount, but whether the actual revenue and cash flow generated by AI applications can ultimately cover data center, chip, interest, and leasing costs. (Source: Compiled from publicly available media reports)