NVIDIA (NASDAQ: NVDA) announced on August 10 that it has signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish separate AI compute financing platforms. These platforms are designed to mobilize over $500 billion in third-party capital for artificial intelligence infrastructure in the future. The company emphasized that this figure represents the aggregate upper limit of capital that can be mobilized across all platforms, not NVIDIA's current revenue, nor a single fund commitment. The agreements are still pending final documentation.
Market participants will look for more details on demand, guidance, and the financing arrangements when NVIDIA reports its fiscal 2027 second-quarter results after the US market close on August 26. According to official explanations, the six institutions will each form their own capital pools, independently evaluating customers, utilization rates, cash flows, and residual values on a project-by-project basis. They will provide long-term financing at relatively competitive interest rates to frontier labs, enterprises, and cloud providers for the purchase and deployment of NVIDIA's full-stack compute solutions.
NVIDIA provides the "AI factory" platform, while financial institutions bring long-term capital and underwriting capabilities. Jensen Huang added on his personal account the same day that the company can provide residual value support of up to 25% of the opportunity size on a project basis. This represents limited, case-by-case credit enhancement and does not replace independent due diligence. PitchBook estimates that based on the 25% cap, the potential contingent exposure could be up to approximately $125 billion, though no funds have actually been raised yet and the legal structure remains undetermined.
Goldman Sachs CEO David Solomon stated that the move aims to "create a market for credit collateralized by NVIDIA compute." BlackRock's Larry Fink emphasized connecting long-term capital with critical infrastructure. Huang's framing is that the company has shifted from making chips to turning "AI factories" into investable productive assets. In his logic, compute itself generates revenue, and the CUDA ecosystem enables equipment to be transferred across customers, extending its economic life.
NVIDIA will release results for the quarter ending July 26 on August 26. The company previously guided to revenue of $91 billion, plus or minus 2%, with non-GAAP gross margin of approximately 75%. Wall Street consensus stands at roughly $92 billion in revenue and about $2.08 in earnings per share. The prior quarter delivered $81.6 billion in revenue, with data center revenue at approximately $75.2 billion. Some sell-side analysts expect this quarter's revenue to reach $94 billion to $95 billion, with even higher numbers projected for the following quarter.
The Vera Rubin platform has entered mass production, and sell-side analysts are focusing on its revenue contribution in the third and fourth quarters, as well as the smooth transition from Blackwell. The company has expressed confidence in approximately $1 trillion in combined revenue from Blackwell and Rubin over calendar years 2025 to 2027. If the financing platforms can ease customers' one-time capital expenditure pressure, they could theoretically help convert orders into recognized revenue. However, the August 26 earnings call is more likely to address orders, capacity, and gross margins first, rather than disclosing memorandum progress item by item.
On the day of the announcement, NVIDIA shares fell approximately 1% to 3%. Market concerns at the time centered on whether the $500 billion figure represented intent or locked-in capital, whether residual value support would expand the company's contingent liabilities, and whether this constituted refinancing of its own demand. Supporters argue that treating GPU clusters as rent-generating infrastructure helps channel private equity and insurance funds into data center construction, alleviating balance sheet pressure on hyperscale cloud providers and emerging cloud companies. Rental prices for products like H100 have at times risen rather than fallen, suggesting that residual values are not a one-way decline.
Skeptics, however, point to cross-shareholdings within the ecosystem, prepayments, and long-term offtake agreements: suppliers injecting capital into or guaranteeing customers, who then use those funds to purchase chips from the same suppliers, potentially exaggerating the independence of end demand. The $500 billion platform is viewed by some commentators as the institutionalized version of this approach. The key variables are whether third-party funders genuinely lend independently based on utilization rates and cash flows, and the actual utilization rate of NVIDIA's residual value commitments. Memorandums are not contracts. Whether funds arrive as planned, whether pricing proves "attractive," and how residual value sharing works when non-performing projects emerge, all require validation from the first batch of projects on the ground.
Comments