From selling chips to securing AI infrastructure! Nvidia (NVDA.US) wants to endorse the new cloud power Nscale to escort Anthropic's computing power expansion

The Zhitong Finance App learned that Nvidia (NVDA.US), the “AI chip superleader”, is actively negotiating to provide significant credit support for the large-scale AI computing power infrastructure project of the new cloud force nScale. According to a source familiar with the matter, Nvidia is discussing providing a credit guarantee for Nscale's large-scale AI data center type computing power infrastructure. Anthropic, the developer of the Claude series, a popular global AI application product line, plans to use the facility.

The insider, who requested anonymity due to undisclosed discussions, said that this credit support will help guarantee the power and construction costs of an AI data center park in West Virginia. The park will be equipped with a cutting-edge computing power cluster in the AI GPU series led by Nvidia. As a dominant supplier in the field of artificial intelligence processors, Nvidia has invested in nScale and other AI computing power service providers.

UK-based Nscale is stepping up the operation of a series of large data centers, including this flagship campus in West Virginia. The company is also moving ahead with an initial public offering, but some investors are objecting to its $30 billion IPO target valuation.

A representative for Nscale declined to comment, while Nvidia did not immediately respond to requests for comment. The Information previously reported discussions between Nvidia and Nscale, saying that the credit guarantee amount could reach several billion dollars.

Nvidia CEO Hwang In-hoon said earlier this month that he had asked the British company to “work around the clock” to overcome the challenges faced in putting some of its data centers into operation.

In recent weeks, the overall market where so-called “new cloud” type AI computing power infrastructure-related technology companies are located has continued to be in turmoil — such companies include nScale, which mainly provides AI computing power. Another new cloud company, the Australian company Firmus Grid, which received investment from Nvidia, abandoned its 30 billion US dollar IPO plan in Australia; South Korea's Elice's performance on the first day of listing was also quite bumpy.

Some media previously reported that Nscale has discussed a second $45 billion deal with Anthropic, involving a second building in the West Virginia campus.

For investors focusing on Nvidia's long-term bull market curve, the potential incident involving the chip giant's guarantee of Nscale not only has the significance of stabilizing the industrial chain, but also increases the risk that investors need to examine. In the short term, if implemented, credit support can help maintain GPU and supporting equipment procurement; in the medium to long term, valuation support depends on whether the project can be put into operation on time, whether the customer can continue to pay, and whether the guarantee exposure matches the supplier's cash strength.

In the future, if financing is blocked by new cloud forces such as CoreWeave, it will inevitably push upstream to assume more credit responsibilities; if terminal cash returns are still insufficient, the pressure may extend further to core suppliers. Nvidia's plan to provide guarantees highlights that leaders in the AI computing power industry chain are taking the initiative to fix major gaps in AI financing, but guarantee news alone cannot judge that the upward trend in AI financing costs has been lifted, or that the trend of the AI bubble with pessimistic pricing has burst.

From selling chips to providing credit support: Nvidia is trying to break the financing bottleneck in AI order fulfillment

Nvidia is discussing providing credit guarantees for Nscale's West Virginia data center to be used by Anthropic, focusing on power and construction costs. The core significance of this arrangement is to use the credit strength of chip suppliers to help computing power operators cross the financing threshold before the project is completed, electrified, and put into operation. Against the backdrop of credit pressure on new cloud enterprises and the cloud giant's cash flow being strictly scrutinized by investors, the AI industry chain is facing a more specific problem: demand and contracts already exist; who will bear the upfront capital and project risks required to turn these orders into deliverable computing power?

The reported deal is still being discussed, and the guarantee amount and specific terms have not been announced. The multi-billion dollar credit support claimed by The Information cannot be equated with the second $45 billion computing power deal that Nscale and Anthropic are discussing; the latter is the size of a potential commercial contract, not an amount guaranteed by Nvidia. The $30 billion target valuation dispute that NScale faced when seeking an IPO indicates that its expansion plans are facing both equity capital pricing and project financing tests at the same time.

The cloud giant's latest earnings report explains exactly why the market is starting to raise financing requirements. In the third quarter, Google Cloud's revenue increased 103% year on year to 30.709 billion US dollars, and operating profit of 11.121 billion US dollars, but Alphabet Group's free cash flow for the quarter was negative 9.065 billion US dollars; Amazon AWS revenue increased 39.3% year on year to about 46 billion US dollars, and operating profit was about 17.8 billion US dollars, but Amazon's free cash flow for the past 12 months was negative 42.4 billion US dollars. The growth and profit of the cloud business is real, and the consumption of cash from expansion investment is also real — the group's negative free cash flow cannot be directly equivalent to loss in cloud business operations, but it means that the construction speed exceeds the internal cash generation capacity during the same period, and more capital is needed.

The market is repricing around this funding gap. On October 29, Alphabet Class A shares closed down 12.6%, and the Philadelphia Semiconductor Index fell about 1.4%, ending four consecutive trading days of gains; after Amazon's earnings report was announced, the stock price fell about 7% after the market. In addition to cash consumption, Alphabet's advertising business performance and Amazon's overall performance and guidelines also affected transaction results. Therefore, a more accurate judgment is that strong cloud growth is no longer sufficient to support valuations alone, and investors demand that growth, capital investment, and cash returns be tested at the same time.

The pressure on the credit market is becoming more divided, and the yield on credit debt and CDS interest spreads linked to AI infrastructure continue to surge. CoreWeave's yield on bonds maturing 2031 reached 15.76% on October 22; although its 5-year CDS narrowed to 1041.4 basis points on October 27, it is still high. Using a simplified model with a 40% recovery rate and constant risk intensity, the spread corresponds to a 5-year cumulative risk-neutral implied default probability of about 58%, including a risk premium, and cannot be regarded as an actual default prediction. Meanwhile, Alphabet's five-year CDS rose 5.3 basis points to 84 basis points on October 29, and Meta and Broadcom's default protection costs also rose on the same day. Instead of assuming that these companies have the same risks, the market is asking together: When can expanding AI investments reduce their dependence on external financing?

Guarantees can protect orders and accelerate the extension of customer risk to AI chip suppliers

Judging from the project financing mechanism, the role of credit guarantees is to add a layer of payment security to creditors or other counterparties within the agreed scope. If Nvidia eventually provides a legally binding guarantee, relevant financiers may reduce their reliance on Nscale's separate credit qualifications, thereby improving financing availability, interest rates, or collateral requirements. However, the guarantee itself does not equal a cash injection, nor does it mean that Nvidia has repaid the debt for Nscale; how much risk it ultimately bears depends on the guarantee limit, coverage obligations, duration, trigger conditions, and recourse arrangements.

The reason why this support falls on the power and construction sector has a clear industrial logic. GPU orders can only be converted into continuous billable computing power services after power supply, cooling, network, and computer room are completed. For large model training and inference clusters that operate intensively, the lack of any key support may prevent expensive equipment from forming the intended output. Construction delays may also lengthen the time gap between interest expenses and the commencement of revenue. Therefore, when Nvidia helps customers resolve construction financing, it actually increases the probability that its own chip orders will be converted into operational clusters, customer revenue, and subsequent purchases.

Nvidia has previously adopted a similar arrangement. Its 10-Q disclosure for the second quarter of fiscal year 2027, including the SB Energy arrangement signed in August, was the maximum total exposure of relevant guarantees of US$108.5 billion, of which the SB Energy portion was US$105 billion, and guarantees relating to land, electricity and construction entities for other AI cloud projects were US$3.5 billion. The SB Energy arrangement takes effect after the conditions are met in accordance with the project phase, and events such as specific tenant default are used as payment triggers. This figure represents the maximum potential exposure under the contract. It is not cash payments or anticipated losses, nor is it the guarantee amount for this Nscale project.

For investors focusing on Nvidia's bull market prospects, the key change is the deepening correlation between return and risk. Chip suppliers originally mainly bear the risks of product competition, customer procurement and repayment; after increasing equity investment and project guarantees, they may also bear losses caused by customer construction delays, financing difficulties, and contract performance. If the same round of AI spending slows down while weakening chip sales and customer solvency, suppliers may experience the double impact of revenue pressure and rising guarantee risk.