NVIDIA and six of the world’s largest asset managers say they will build financing platforms meant to steer more than $500 billion of third-party capital into AI data-center construction, reframing GPU clusters as a financeable infrastructure asset rather than a purchase order for servers (NVIDIA says six asset managers will steer more than $500 billion into AI data-center construction). The partners named were Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR.
The pitch reframes NVIDIA’s full stack — accelerated chips, networking, systems software, AI frameworks, and a global developer base — as productive infrastructure. One factory, the company argues, can serve many customers and many workloads, and software updates keep installed hardware earning for longer. NVIDIA points to its A100, introduced in 2020, as still in active commercial use six years later, with customers committing multi-year capacity that stretches its economic life toward a decade (NVIDIA argues the A100 from 2020 is still in active commercial use six years later). NVIDIA summed up the thesis in a single line: “In AI, compute is revenue.” The partners — Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR — are among the world’s largest asset managers, each with the balance-sheet scale to underwrite data-center construction at the gigawatt level. Apollo and Blackstone both run dedicated infrastructure and real-assets funds; Brookfield and KKR have deep histories in energy- and infrastructure-intensive assets; BlackRock’s Aladdin risk platform and Goldman Sachs’s balance sheet give the group financial firepower that stretches well beyond a single vendor.
The company cites live rental markets to support the “asset” claim. One-year H100 rental rates rose from about $1.70 per GPU-hour in October 2025 to about $2.35 in March 2026, while cross-provider on-demand median pricing climbed from roughly $2.00 to $2.70 per GPU-hour over the same window, and Blackwell B200 cloud rates were reported at roughly $5.30 to $7.05 per GPU-hour (NVIDIA itself reports that one-year H100 rental rates rose from about $1.70 to $2.35 per GPU-hour between October 2025 and March 2026). Those figures, NVIDIA says, show compute economics that behave like a durable, income-producing asset rather than a depreciating line item.
NVIDIA built explicit guardrails into the plan (NVIDIA’s AI-factory compute post described these guardrails). The $500 billion is aggregate third-party capital the platforms are designed to mobilize over time — not NVIDIA revenue, a single fund, or a commitment to one customer (NVIDIA stresses the $500B is aggregate third-party capital, not its own revenue or a single fund) — and each financial partner will independently underwrite demand, utilization, cash flow, and residual value. Where NVIDIA steps in, it may offer residual-value support for up to 25% of an opportunity, assessed carefully on a project-by-project basis (NVIDIA may offer residual-value support for up to 25% of an opportunity, assessed project by project). That is deliberately narrower than other compute-financing arrangements, the company notes, an answer to the obvious “circular financing” question.
The financing push arrives as NVIDIA expands its physical and developer footprint. The company is promoting GTC Berlin, a three-day AI infrastructure conference scheduled for October 20–22, 2026, where CEO Jensen Huang will deliver a keynote (NVIDIA opened registration for its GTC Berlin AI infrastructure conference on June 26, 2026). The same full-stack platform is also moving into national programs: NVIDIA, Indosat, and Universitas Gadjah Mada recently opened Indonesia’s first university AI center to train local talent (NVIDIA, Indosat, and Universitas Gadjah Mada opened Indonesia’s first university AI center to train local AI talent).
For a closer look at how NVIDIA is extending that stack into sovereign AI programs, see our coverage of Indonesia’s first university AI center built with NVIDIA (zBrandco covered Indonesia’s first university AI center built with NVIDIA).
