BlackRock (NYSE: BLK), Goldman Sachs (NYSE: GS), Apollo Global Management (NYSE: APO), Blackstone (NYSE: BX), Brookfield Asset Management (NYSE: BAM), and KKR (NYSE: KKR) are in discussions with Nvidia (NASDAQ: NVDA) to participate in a massive new round of AI infrastructure spending that could ultimately total as much as $500 billion.

According to the Financial Times, the planned collaboration centers on funding expanded computing capacity, power supply, and data center construction. An announcement of the deal could come as early as next Monday.

The involvement of six of the most influential names in global asset management and private equity underscores how AI infrastructure has moved from a niche technology investment into one of the largest capital-allocation themes in global finance. If the reported figure is reached, the program would rank among the largest private-sector infrastructure commitments in history, comparable in scale to national-level energy or telecommunications buildouts.

Following the news, NVDA shares declined 1.4%, wiping out more than $70 billion in the company's market valuation.

Nvidia's Expanding Role in AI Infrastructure Financing

The talks highlight Nvidia's increasingly central position in financing the physical infrastructure underpinning artificial intelligence. Valued at approximately $5.25 trillion, the company already supplies the graphics processors used to train and operate most of the largest AI systems in the United States, along with related software and computing tools. Nvidia's near-dominant share of the AI training chip market has made it a gatekeeper for the entire buildout: data center operators, cloud providers, and AI developers typically cannot scale without first acquiring its hardware.

However, GPUs alone are insufficient to sustain AI operations. Operators also require large-scale server facilities, power, cooling systems, and long-term financing. AI workloads consume substantially more electricity than conventional data center computing, and industry estimates project that data center power demand could more than double by the end of the decade, placing strain on electrical grids and driving interest from infrastructure-focused investors experienced in energy projects.

Nvidia has progressively stepped in to help clients secure this funding, even investing its own capital directly into companies that purchase its hardware.

Mark Cuban Warns the Financing Model Could "Crumble"

Billionaire investor Mark Cuban has publicly expressed concern over the volume of debt and financial engineering now supporting the AI spending surge. His core argument is that Nvidia has moved beyond simply selling chips — it is actively financing the very customers placing orders.

In July, Cuban drew a parallel between the current landscape and the funding frenzy of the dot-com era, writing: "instead of IPOs, Nvidia is the IPO, funding everyone and anyone."

Nvidia's financing support has taken multiple forms, including direct capital injections, revenue-sharing agreements, and arrangements guaranteeing minimum income levels for data center operators and emerging cloud providers. These structures enable customers to order GPUs and begin construction before their own operations generate sufficient cash flow.

During the first half of 2026, Nvidia invested more than $40 billion across its AI initiatives. Recipients and related entities included OpenAI, Corning (NYSE: GLW), and IREN (NASDAQ: IREN).

Several individual deals are notable for their scale. OpenAI has been linked to a proposed $100 billion data center program. Financing connected to xAI has utilized special-purpose vehicles (SPVs) valued in the billions of dollars.

Complexity and Risk in AI Infrastructure Financing

Some of these arrangements involve pledging GPUs as loan collateral. They may also include long-duration leases and revenue projections based on the performance of early-stage AI companies — projections that could prove inaccurate if actual earnings fall short of the figures assumed when contracts were signed.

Legal experts have observed that AI infrastructure financing has grown increasingly intricate. A single project may combine private lending, debt securitization, and SPVs established specifically to hold assets or borrow without appearing on a balance sheet. Participants can include banks, insurance companies, and pension funds. The entry of major private equity and asset management firms into the current round of talks reflects a broader trend in which alternative-asset managers have become leading lenders and infrastructure investors, stepping into roles traditionally held by commercial banks.

Hardware depreciation presents another layer of risk. Nvidia has been releasing new generations of AI chips on an annual cycle, meaning a data center could still be paying off one generation of processors when a more advanced version reaches the market. Consequently, older GPUs may lose value more quickly than lenders initially assumed — a critical concern when the hardware itself serves as loan collateral.

Banks and private credit facilities frequently extend loans for equipment carrying very high costs at the time of installation. If overall AI spending slows, operators that built capacity for demand that fails to materialize could face significant financial strain. More frequent hardware refreshes could render expensive installations obsolete sooner than anticipated, and competing chips from rivals — including AMD's Instinct accelerators and custom silicon developed by major cloud providers such as Google, Amazon, and Meta — could reduce demand for Nvidia-based systems.

Any of these factors could leave operators repaying substantial debts against facilities that generate less revenue than projected — a dynamic that extends well beyond traditional technology equity investing and into the realm of leveraged finance applied to data centers, GPUs, and energy projects.