# Nvidia’s $500 billion financing plan is a framework, not a loan book > Source: > Published: 2026-08-13 09:35:04.376502+00:00 # Nvidia’s $500 billion financing plan is a framework, not a loan book - Nvidia’s August 10 announcement covers memorandums of understanding and independent financing platforms, not a completed $500 billion loan. [[1]](https://nvidianews.nvidia.com/news/nvidia-partners-with-apollo-blackrock-blackstone-brookfield-goldman-sachs-and-kkr-to-establish-ai-compute-infrastructure-financing-platforms-to-mobilize-over-500-billion-of-third-party-capital) - The six named financial firms are Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR; the eventual lenders and fund investors have not been disclosed. [[1]](https://nvidianews.nvidia.com/news/nvidia-partners-with-apollo-blackrock-blackstone-brookfield-goldman-sachs-and-kkr-to-establish-ai-compute-infrastructure-financing-platforms-to-mobilize-over-500-billion-of-third-party-capital) - Nvidia said it may provide residual-value support covering up to 25% of an opportunity, but it did not publish a standard guarantee, borrower list or collateral schedule. [[2]](https://www.axios.com/2026/08/12/nvidia-ai-financing-blackrock-goldman) - Existing AI infrastructure loans show the likely structures: special-purpose borrowers, delayed draws, GPU and infrastructure collateral, pledged customer cash flows, and long-term offtake contracts. [[3]](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000236/crwv-20260515.htm)[[4]](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000222/crwv-20260331.htm) - Analysts disagree over whether Nvidia’s role lowers the cost of productive infrastructure or reinforces a circular financing loop in which the chip supplier helps fund demand for its own products. [[2]](https://www.axios.com/2026/08/12/nvidia-ai-financing-blackrock-goldman) Nvidia’s latest financing initiative is large enough to reshape how AI data centers borrow, but it is not yet a financing transaction that investors can underwrite. On August 10, the chipmaker announced memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to create independent platforms capable of mobilizing more than $500 billion of third-party capital over time. [[1]](https://nvidianews.nvidia.com/news/nvidia-partners-with-apollo-blackrock-blackstone-brookfield-goldman-sachs-and-kkr-to-establish-ai-compute-infrastructure-financing-platforms-to-mobilize-over-500-billion-of-third-party-capital) The announcement named the capital providers and described the intended asset class. It did not name borrowers, specify how much each firm would commit, set interest rates or maturities, identify a completed project, or disclose the collateral that would secure a loan. Nvidia said the platforms would be developed for its customers and that the arrangements still require further documentation. [[1]](https://nvidianews.nvidia.com/news/nvidia-partners-with-apollo-blackrock-blackstone-brookfield-goldman-sachs-and-kkr-to-establish-ai-compute-infrastructure-financing-platforms-to-mobilize-over-500-billion-of-third-party-capital) ## What Nvidia actually announced The transaction is best understood as an origination and capital-formation framework. Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR are expected to help assemble pools of long-duration capital for AI factories, data centers and related infrastructure. Nvidia described the platforms as independent, which suggests that the financial firms, rather than Nvidia alone, would conduct the formal underwriting and hold the resulting credit exposure. [[1]](https://nvidianews.nvidia.com/news/nvidia-partners-with-apollo-blackrock-blackstone-brookfield-goldman-sachs-and-kkr-to-establish-ai-compute-infrastructure-financing-platforms-to-mobilize-over-500-billion-of-third-party-capital) That does not mean Nvidia has removed itself from the risk. Nvidia said it may provide a residual-value support mechanism for up to 25% of an opportunity, assessed project by project. The wording leaves open whether the support would cover GPUs, lease payments, equipment resale values or another defined loss layer. It also leaves open whether the support would be a guarantee, a contractual purchase obligation, a put-like arrangement or a combination of tools. [[2]](https://www.axios.com/2026/08/12/nvidia-ai-financing-blackrock-goldman) The distinction between a platform and a funded loan is important. Until final agreements are signed and a project closes, the announced $500 billion is a mobilization target rather than debt outstanding. Nvidia’s announcement did not identify a completed financing, a drawdown schedule or a binding aggregate commitment from the six firms. [[1]](https://nvidianews.nvidia.com/news/nvidia-partners-with-apollo-blackrock-blackstone-brookfield-goldman-sachs-and-kkr-to-establish-ai-compute-infrastructure-financing-platforms-to-mobilize-over-500-billion-of-third-party-capital) ## The borrowers and lenders are still mostly undefined The likely borrowers fall into several categories: AI cloud providers that buy and rent out GPU capacity, data-center developers building facilities for large tenants, frontier AI companies seeking dedicated compute, and enterprises or governments funding their own AI factories. Nvidia referred broadly to frontier AI labs, enterprises and AI clouds rather than identifying a first borrower or project. [[1]](https://nvidianews.nvidia.com/news/nvidia-partners-with-apollo-blackrock-blackstone-brookfield-goldman-sachs-and-kkr-to-establish-ai-compute-infrastructure-financing-platforms-to-mobilize-over-500-billion-of-third-party-capital) The named financial firms may also play different roles. Goldman Sachs could arrange or distribute credit; Apollo, Blackstone and KKR could supply private credit or infrastructure capital; BlackRock and Brookfield could invest through infrastructure, credit or insurance-linked vehicles. Those are reasonable possibilities, not disclosed assignments. The announcement does not allocate responsibilities, identify fund vehicles or name a lender syndicate. [[1]](https://nvidianews.nvidia.com/news/nvidia-partners-with-apollo-blackrock-blackstone-brookfield-goldman-sachs-and-kkr-to-establish-ai-compute-infrastructure-financing-platforms-to-mobilize-over-500-billion-of-third-party-capital) A separate Nvidia-OpenAI arrangement illustrates why investors are asking for more detail. The Wall Street Journal, as summarized by Tom’s Hardware, reported in July that Nvidia was discussing a guarantee of roughly $250 billion tied to OpenAI’s proposed lease of a 10-gigawatt Ohio campus, alongside a possible arrangement to finance up to $350 billion of chips for the site. The terms were unsettled and could fall apart; there is no public confirmation that the proposal is part of Nvidia’s new six-firm platforms. [[5]](https://www.tomshardware.com/tech-industry/data-centers/nvidia-weighs-250-billion-guarantee-so-openai-can-lease-softbanks-10-gigawatt-ohio-campus) Nvidia’s latest available public filing provides a useful baseline. Its April 26, 2026, Form 10-Q disclosed existing partner facility-lease guarantees with maximum gross exposure of $3.5 billion, reduced as partners make payments over five to seven years. Partners had placed $712 million in escrow, and Nvidia said the guarantees’ fair value was not material. A $250 billion proposed backstop would therefore be structurally and quantitatively different from the guarantee book Nvidia had disclosed, if it were ever finalized. [[6]](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000052/nvda-20260426.htm) ## What the credit structures look like in practice The clearest public precedents come from CoreWeave, an AI cloud provider that has raised debt against contracted compute deployments. In March, CoreWeave announced that a subsidiary had closed an $8.5 billion delayed-draw term loan. The company said the facility was non-recourse, received A3 and A-low investment-grade ratings from Moody’s and DBRS, and was secured by high-performance-computing infrastructure and an associated customer contract. Those are company-reported transaction details, although the ratings were assigned by the named agencies. [[7]](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000129/ex991.htm) CoreWeave’s May 2026 filing shows the mechanics in greater detail. CoreWeave Financing DDTL V, LLC was the borrower. Morgan Stanley Senior Funding acted as administrative agent, U.S. Bank Trust served as collateral agent, and MUFG and Morgan Stanley were coordinating lead arrangers and joint bookrunners. The loan could be drawn as capital expenditures were made, carried a margin of 4.50% over SOFR for floating-rate loans and matured on November 15, 2031. [[3]](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000236/crwv-20260515.htm) The structure relies on more than the resale value of a GPU. CoreWeave’s filings say its delayed-draw facilities are collateralized by the assets underlying contributed contracts and pledged contractual cash flows, generally from investment-grade counterparties. That combination gives lenders a claim on equipment, the revenue contract supporting it and the special-purpose borrower that owns the financed assets. [[4]](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000222/crwv-20260331.htm) Nvidia has also supported the demand side of that model. A CoreWeave filing describes a $6.3 billion Nvidia order form under which Nvidia can purchase residual unsold cloud capacity through April 13, 2032, subject to delivery and service conditions. That obligation can make a lender more comfortable that capacity will have a buyer if CoreWeave cannot place all of it with its own customers, although it is not the same as a blanket guarantee of CoreWeave’s debt. [[8]](https://www.sec.gov/Archives/edgar/data/1769628/000176962825000059/coreweave3q25earningspress.htm) Other public transactions point to a broader menu of structures. Lambda announced a $1 billion syndicated senior secured credit facility in May, arranged by J.P. Morgan, to expand its Nvidia accelerator fleet and data-center capacity. Alpha Compute separately announced a proposed $31.9 million non-recourse senior secured facility backed by servers containing Nvidia B300 GPUs; because that transaction was described as a binding term sheet with definitive documentation still in process, it should not be treated as a completed loan. [[9]](https://www.businesswire.com/news/home/20260507872879/en/Lambda-Closes-%241-Billion-Senior-Secured-Credit-Facility-to-Meet-Gigawatt-Scale-AI-Infrastructure-Demand)[[10]](https://www.sec.gov/Archives/edgar/data/1095435/000117184326002823/exh_991.htm) ## Why investors are focused on valuation and credit risk The credit question is whether a financed AI factory can generate enough contracted cash flow before its equipment loses economic value. A GPU may be physically resalable, but its lending value depends on utilization, power availability, networking compatibility, software demand and the pace of new chip releases. A lender that seizes hardware may still face storage, redeployment and obsolescence costs. Those risks are different from the relatively standardized collateral pools found in many mortgages or auto loans. The financial structure can reduce some of that uncertainty. Delayed draws match borrowing with equipment deployment. Special-purpose borrowers can isolate project assets and revenue contracts from a parent company’s wider balance sheet. Long-term customer commitments can give lenders visibility into debt service. Nvidia’s proposed residual-value support could absorb part of the loss if a project’s hardware or capacity is worth less than the financing assumed. These mechanisms can make credit available to smaller operators that cannot borrow on hyperscaler terms. [[2]](https://www.axios.com/2026/08/12/nvidia-ai-financing-blackrock-goldman)[[3]](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000236/crwv-20260515.htm)[[4]](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000222/crwv-20260331.htm) But the same mechanisms can blur where demand originates. Nvidia sells the GPUs, sometimes invests in infrastructure providers, may purchase unused capacity and is now helping assemble capital for customers that use Nvidia systems. UBS analysts said the strategy raises questions about circular AI financing, while Seaport Global analyst Jay Goldberg warned that investors could eventually question whether the spending cycle makes economic sense. Cantor Fitzgerald analyst CJ Muse offered the strategic counterpoint: financing can help customers use Nvidia GPUs rather than competing chips. [[2]](https://www.axios.com/2026/08/12/nvidia-ai-financing-blackrock-goldman) The market’s judgment will depend on disclosures that the August 10 announcement did not provide: who bears first losses, how residual values are calculated, whether Nvidia support is capped in aggregate, which customer contracts qualify, how lenders treat new GPU generations, and whether borrowers remain liable beyond the financed assets. Those details will determine whether the platforms function like ordinary project finance with specialized collateral or like vendor-supported credit whose performance remains closely tied to Nvidia’s own sales cycle. For now, investors have a concept, a list of six financial partners and a potential support mechanism. They do not yet have a loan book to value. The next evidence will come from final agreements and the first completed platform-financed projects, including their borrowers, collateral packages, customer contracts, pricing and guarantees. [[1]](https://nvidianews.nvidia.com/news/nvidia-partners-with-apollo-blackrock-blackstone-brookfield-goldman-sachs-and-kkr-to-establish-ai-compute-infrastructure-financing-platforms-to-mobilize-over-500-billion-of-third-party-capital) ## Companies mentioned ## Further sources [[1] Nvidia’s August 10, 2026 primary announcement describes MOUs with Apollo, Black… ↗](https://nvidianews.nvidia.com/news/nvidia-partners-with-apollo-blackrock-blackstone-brookfield-goldman-sachs-and-kkr-to-establish-ai-compute-infrastructure-financing-platforms-to-mobilize-over-500-billion-of-third-party-capital) [[2] Axios reported Nvidia’s possible residual-value support of up to 25%, analyst c… ↗](https://www.axios.com/2026/08/12/nvidia-ai-financing-blackrock-goldman) [[3] CoreWeave’s SEC filing identifies the $3.1 billion DDTL 5.0 borrower, agents, a… ↗](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000236/crwv-20260515.htm) [[4] CoreWeave’s first-quarter 2026 filing describes non-recourse delayed-draw facil… ↗](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000222/crwv-20260331.htm) [[5] Tom’s Hardware summarized reports that Nvidia was discussing a possible $250 bi… ↗](https://www.tomshardware.com/tech-industry/data-centers/nvidia-weighs-250-billion-guarantee-so-openai-can-lease-softbanks-10-gigawatt-ohio-campus) [[6] Nvidia’s April 26, 2026 Form 10-Q disclosed $3.5 billion of maximum gross expos… ↗](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000052/nvda-20260426.htm)+4 more The stories that matter, in one email. Free — unsubscribe anytime.