Nvidia has corralled six of the world’s largest alternative asset managers and Goldman Sachs into a financing consortium that could direct more than $500 billion of third-party capital into data centers built exclusively on its hardware. The company signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, KKR and Goldman Sachs to create independent platforms that will lend to AI labs, cloud providers and enterprises buying Nvidia-based infrastructure. Nvidia itself is not putting up the money; it is arranging the pipeline.
The structure
The six firms will raise and manage the pools, targeting long-duration funding at rates the group describes as attractive. The source does not disclose fund sizes, target returns, management fees, carry structures or any break-fee provisions. It also does not say whether the capital will be deployed as senior debt, mezzanine, preferred equity or a blend. What is clear is that every dollar is earmarked for Nvidia compute, a contractual exclusivity that turns the financing into a de facto sales facility for the chipmaker.
The pitch
Goldman Sachs chief executive David Solomon called the moment “pivotal” and said the bank’s investment and distribution roles reflect confidence in Nvidia’s leadership. The consortium’s thesis is that AI data centers behave like long-duration infrastructure assets rather than conventional IT equipment because Nvidia compute can generate revenue over extended periods and retain value across workloads and operators. That characterization is a claim, not a demonstrated credit history. The source notes the firms do not explicitly argue that financing AI factories carries lower credit risk than financing traditional IT deployments.
The circularity risk
The arrangement weakens the natural brake on overbuilding: the price and availability of capital. By helping its own customers secure funding, Nvidia introduces a circular dynamic that echoes the vendor-financing loops of the late-1990s telecom and dot-com booms. The source acknowledges this does not prove a bubble exists, demand for Nvidia hardware is genuine and enormous, but it does mean the supply of capital is no longer an independent constraint on deployment.
The asset-life problem
The largest unstated tension sits in the hardware cycle. Nvidia and its peers introduce new, faster accelerators every year, which devalues the previous generation. The source flags that GPUs have short and uncertain economic lives, yet the financing is pitched as long-duration. If the collateral depreciates faster than the debt amortizes, the infrastructure-asset thesis gets tested in ways the MOUs do not address.