Written by: Rita

NVIDIA is launching a new type of cloud (neocloud) architecture, offering credit guarantees to new cloud service providers in exchange for a continuous share of their future cloud revenues.

On August 14, Morgan Stanley maintained NVIDIA's overweight rating and a target price of $288 in a research report, believing this model is expected to bring NVIDIA recurring revenues in the tens of billions of dollars, with limited downside risk. However, the cost of the credit guarantee is that NVIDIA needs to assume financing risks. Morgan Stanley predicts that this model will become the next core controversy in the market for long and short sides, clearly standing on the optimistic side.

Compute Gap Forces New-Type Clouds, NVIDIA Exchanges Credit for Revenue Share

The demand for computing power is still accelerating, but the expansion of traditional hyperscale manufacturers is hindered by cash flow and physical resources.

Morgan Stanley pointed out that the growth of tokens far exceeds the supply of computing power, with some clients even forced to redeploy computing resources originally intended for training to inference tasks. Meanwhile, U.S. hyperscale manufacturers face shortages of land, electricity, and shells domestically, while encountering sovereign demands when expanding overseas. New cloud service providers are an important force in filling this gap, but their capital strength is far less than that of hyperscale manufacturers.

NVIDIA's solution is: to help these clients obtain financing through credit guarantees in exchange for a continuous revenue share from their cloud income. The specific mechanism is that NVIDIA provides credit endorsement for new cloud service providers by committing to pay GPU hourly rental fees, enabling them to finance at rates close to investment grade. In exchange, NVIDIA obtains a revenue share above the agreed price.

This structure is similar to NVIDIA's existing collaboration model with CoreWeave but increases the upside potential. Morgan Stanley believes that if this model is fully rolled out, NVIDIA will establish a recurring revenue source with a 100% gross margin beyond chip sales.

This Model's Pros and Cons

Optimists believe that the demand for computing power is real, and new-type clouds are a complement to traditional hyperscale manufacturers, not a duplication of construction.

The growth of tokens far exceeds the supply of computing power; the expansion of traditional hyperscale manufacturers is constrained by cash flow and physical resources, and the entry of new clouds is addressing this supply-demand imbalance. Morgan Stanley pointed out that some sovereign countries have begun to realize that their land, electricity, and shell resources can serve as strategic assets to attract AI infrastructure investments. These countries tend to support local new cloud enterprises, and NVIDIA's credit guarantee model can accelerate this process.

The revenue share will also create recurring income for NVIDIA on the scale of tens of billions of dollars, enhancing the predictability of earnings. Morgan Stanley estimates that if the new cloud ecosystem reaches a scale similar to the top four hyperscale manufacturers (about 25GW), based on an industry rental level of $20 million per MW, annual revenues could reach $500 billion. If NVIDIA can secure a quarter of that revenue as a 100% gross margin share, it could enhance FY28 EBIT by about 60%.

Pessimists, however, believe that this essentially involves NVIDIA "creating customers," stimulating demand that did not previously exist through credit guarantees, which poses a self-reinforcing risk. Additionally, if future supply of computing power becomes excess, NVIDIA will face significant downside exposure. However, Morgan Stanley believes that as long as the demand for computing power continues to exceed supply, this model has commercial rationality.

Limited Downside Risk, Considerable Upside Potential

Morgan Stanley acknowledges that this seems like "creating demand" or even "self-reinforcing," but the core premise is that the demand for computing power is real.

If there is an oversupply of computing power (which currently appears to be very unlikely), NVIDIA has backup plans. Credit guarantees are not purely cash-backed securities, rather NVIDIA commits to using the GPUs of new clouds for its own R&D. NVIDIA has a large annual R&D budget for renting GPUs, and this expenditure can be fully transferred to new cloud partners.

NVIDIA can keep its direct exposure from the previously announced $500 billion investment fund under 25%; the main investment decision-makers are not involved in self-reinforcing issues, further reducing downside risk.

For new cloud service providers, the appeal of partnering with NVIDIA lies in obtaining computing power in a tight supply environment, receiving full stack architecture support, and achieving faster deployment speeds. Morgan Stanley estimates that the premium paid for gaining computing power a year in advance can offset a 30% price decline over the next five years, indicating that clients are willing to pay a premium for early access to computing power, which is precisely the key in the current market.

Morgan Stanley lists NVIDIA as the preferred target in the semiconductor sector, maintaining an overweight rating and a target price of $288. The Vera Rubin product cycle will alter the current supply-demand dynamics, prompting customers to compete for allocation rights, which is the source of NVIDIA's pricing power. The new cloud revenue sharing model is a new variable in this narrative. Morgan Stanley concludes: downside risk has an upper limit, while upside potential has no ceiling.

Disclaimer

This article is a compilation and interpretation of a third-party brokerage research report (Morgan Stanley, August 14, 2026) by Chaoxiang Research, combined with publicly available market information. The ratings, target prices, earnings forecasts, and related judgments quoted in the text are the views of the analysts of that brokerage and represent the position of their institution, not the view of Chaoxiang Research, nor do they constitute any investment advice.

The market has risks, and decisions should be made independently. This article should not be used as the basis for buying or selling any securities.