The change in NVIDIA's business model is' computing power for revenue '. According to reports, Nvidia is signing revenue sharing agreements with startups, which will allow customers to exchange a portion of their future revenue for access to computing power. From a personal perspective, this is not as simple as traditional "selling chips", nor is it strictly a bank loan. It should be a combination of "computing power credit+manufacturer financing+revenue sharing+ecological locking". Will it be another kind of 'computing power loan'? It's worth discussing in detail There are key details to be corrected: many reports are written as' future profits', but Nvidia's official statement is closer to cloud revenue share, which is cloud revenue sharing, rather than net profit sharing. This may be heavier for Nvidia than 'profit sharing', as revenue sharing usually occurs before profits. Nvidia received two revenue streams: One segment is the revenue from selling GPU, system, network and other products; The other paragraph is that after these GPUs are rented out to AI customers by cloud vendors, Nvidia continues to receive a share based on cloud revenue or usage. The official definition of this is "standard product revenue+cloud revenue sharing+recurring revenue linked to usage". 1. Why did Nvidia do this? From a personal perspective, we should address the two bottlenecks in the AI industry: who can finance the construction of computing power, and who can sustain high utilization rates to digest computing power. 1) Nvidia stated in the announcement that AI is moving from model development to production inference, and the demand for computing power is shifting towards AI factories that operate continuously and generate tokens on a large scale. In this mode, whether the GPU cluster can achieve long-term high utilization is more critical than selling hardware at once. 2) Nvidia wants to expand its customer base from large cloud vendors to AI native companies, regional clouds, enterprises, research institutions, and sovereign AI projects. Nvidia split its data center into Hyperscale and ACIE in its first quarter financial report; ACIE includes AI Clouds, Industrial, and Enterprise, indicating that it is expanding its growth narrative from a few mega cloud vendors to a broader AI factory ecosystem. 3) This is an ecological strategy to counter customer self-developed chips and platform substitution. Large cloud vendors have the ability to independently develop AI chips, while startups and Neocloud do not have the same capability. NVIDIA pre binds these emerging customers to its GPU, network, software, and DSX architecture through credit support, computing power points, and revenue sharing. In other words, it's not just making hardware money once, but embedding itself into the future AI application revenue chain. 2. Has the revenue structure of Nvidia changed? The new revenue sharing model, although strategically significant, should have a small scale and may not immediately change the overall financial report structure in the short term. But what it changes is the quality of income and risk exposure. 1) NVIDIA has shifted from "one-time hardware revenue" to "hardware revenue+future computing power revenue sharing". This is a bit like selling chips to a mining site and continuing to extract revenue from the mining site's computing power. 2) It shifts Nvidia's revenue from a pure supply chain cycle to a usage cycle. The benefit is that if the demand for AI inference really explodes in the long term, Nvidia can continue to share downstream growth; The downside is that if downstream demand is insufficient, GPU rental prices decline, and customers default, Nvidia becomes not only a seller, but also a bearer of credit risk and utilization risk. 3) Financial processing will be more complex. Nvidia's annual report states that product sales revenue is usually recognized when control of the product is transferred, and customer items such as customer incentives, rebates, and market development funds are treated as revenue reduction items. So in the future, it depends on the details of the contract: are token credits fees? Is it income deduction? Is it customer motivation? Is it prepaid computing power? When is revenue sharing recognized? All of these will affect the judgment of 'income quality'. 4) The valuation logic may be repackaged. Hardware companies usually look at it based on cyclical profits; Companies with recurring, usage tied revenue may be willing to offer higher multiples in the market. But the premise is that this part of the revenue comes from real terminal demand, rather than demand that Nvidia has "pulled" through credit support. 3. Is this considered a 'computing power loan'? Strictly speaking, it does not seem like a loan in the legal sense; In essence, the economy is very similar to computing power loans: 1) If the contract states that a fixed amount or minimum usage fee must be paid in the future, regardless of whether it is profitable or not, then it is more like debt or financial leasing. 2) If the contract states' you will only distribute your future income proportionally to Nvidia ', it is more like revenue sharing financing. 3) If Nvidia provides guarantees, repurchases, and credit enhancements to AI cloud vendors, making banks willing to lend to data center projects, it would be more like supplier financing support. So personally, I define it as: The financing of "computing power loan" manufacturers, rather than standard loans. Its essence is to transform GPU computing power from a one-time purchase commodity into a financial asset that can be prepaid, installment, mortgaged, securitized, and revenue shared. 4. Is this necessarily all a good thing? Not entirely Easy to form circular financing. Suppliers invest or guarantee customers, and customers buy products from suppliers. Suppliers confirm revenue, and the market uses this revenue growth to prove strong demand for AI. This cycle may not necessarily be fake, but it will make it more difficult for external investors to distinguish how much is real terminal demand and how much is demand driven in advance by credit, subsidies, points, and income sharing. And the financial structure is becoming increasingly opaque So overall: This can be considered a very strong industrial financial innovation, which is very effective when demand truly grows: startups gain computing power, AI cloud vendors obtain financing, Nvidia obtains hardware revenue and future revenue sharing, ecological expansion accelerates, and it binds more small and medium-sized customers. But its risks are not small either: It will tie together the cash flows of Nvidia, AI startups, creditors, and end customers. Once AI application revenue is not fast enough or thick enough, or GPU rental prices decline, risks will be transmitted in the opposite direction along these contract chains. Of course, it was okay in the early stages because the physical strength was also not strong. In the future, we will focus on four key indicators: 1) The actual utilization rate of downstream AI clouds; 2) GPU rental price trend; 3) Does the revenue sharing contract have a minimum commitment or repurchase clause; 4) And whether Nvidia's investment/guarantee/accounts receivable growth is faster than terminal cash flow growth. As long as these indicators are healthy, then this is a good industrial financial innovation. But once these indicators deviate significantly, the risk of so-called 'computing power loans' will shift from a narrative issue to a financial issue. This article is sponsored by @ bitget_zh, titled 'Bitget Buying US Stocks: Instant Entry, Smooth Trading'