NVIDIA's 'Compute-for-Revenue' model hits a snag. According to a Wall Street Journal report, NVIDIA has paused deals where it provided credit support to AI cloud providers in exchange for revenue sharing. Two main reasons: 1. Internal concerns that the plan might trigger antitrust scrutiny; 2. In the first few weeks of the plan, NVIDIA sought control: chips could only be rented to NVIDIA-approved customers. NVIDIA stated that it hopes to distribute capacity among multiple smaller AI companies rather than leasing it all to one major client. This, however, faced resistance from some cloud providers, who believe they should have the right to choose their own customers. In simple terms, a chip company that's already big enough to worry about antitrust issues tried to use a guaranteed revenue-sharing model to organize the downstream cloud market. Within two weeks to two months, they realized that 'telling customers how to do business' wasn't going to work, so they hit the brakes. Looking at the bigger picture, the timing is sensitive: aside from pushing the compute-for-revenue model, NVIDIA recently arranged two even larger deals: 1) Helping arrange approximately $500 billion in client financing involving major U.S. financial institutions; 2) Agreeing to provide up to $105 billion in guarantees to help OpenAI lease ultra-large-scale data centers. This has raised suspicions of circular transactions. From a personal perspective, NVIDIA pausing some smaller-scale deals is likely a proactive move to reduce contingent liabilities on its balance sheet and limit regulatory exposure. Consider it a risk-control brake. Of course, NVIDIA officials also emphasized that the model is still evolving, suggesting they’re more likely to tweak the terms (lower revenue-sharing ratios, reduce customer approval rights, shorten guarantees) rather than completely abandon 'helping customers finance.' This post is sponsored by @bitget_zh: 'Bitget Buy U.S. Stocks: Instant Entry, Smooth Trading'.