Blackstone and Wall Street Giants Mobilize $500 Billion for AI Infrastructure

Six of the world's largest financial institutions are joining forces with Nvidia to mobilize more than half a trillion dollars in outside capital for artificial intelligence infrastructure. Blackstone, Apollo, BlackRock, Brookfield, Goldman Sachs and KKR have signed a memorandum with the chipmaker to build financial platforms designed to turn AI computing power into a new investable asset class.

The agreement takes the form of a memorandum rather than a binding financial deal, but the scale of the ambition is unprecedented. The platforms are meant to let hyperscalers, AI labs and enterprises finance Nvidia hardware purchases and data center construction without loading the risk onto Nvidia's own balance sheet. Instead, capital is meant to flow from institutional investors, for whom computing infrastructure is becoming an asset class similar to commercial real estate or toll roads.

Not a bubble, a shortage

Blackstone President and Chief Operating Officer Jon Gray laid out his reasoning in an interview with France's Les Echos Capital Finance, published September 4, 2026. In his view, the current AI investment boom differs fundamentally from classic speculative bubbles, in which built-up capacity outpaces real demand. Classic bubbles form when investors bet on speculative future revenue and the production capacity that gets built exceeds actual demand.

Despite real demand for computing power, every application of this intelligence and the productivity gains that come from it are constrained by available supply. That's the exact opposite of what we see in a classic bubble - Jon Gray, President and COO of Blackstone

With AI, Gray argues, it's the reverse of a classic bubble: it's the availability of capital, energy and specialized chips that limits the pace of deployment, not a lack of willing customers. The argument rests on physical bottlenecks that capital alone can't solve: stable power access, grid connections, transformers, cooling systems, land and administrative permits.

The physical limits of the race

In many locations around the world, the wait for a new power grid connection is now longer than the construction of the server hall itself. The unit Gray uses illustrates the scale of the challenge well: a single 1-gigawatt 'AI factory' costs around $35 billion in chips alone, before adding power, cooling and the building itself.

At those sums, Gray argues, capital availability becomes one of the main factors limiting how fast infrastructure can be built out. That capital shortage is precisely what's meant to justify bringing the world's six largest funds and investment banks together in one joint venture with Nvidia.

How the financial platforms would work

According to Nvidia's statement, the platforms are meant to treat computing power built on its chips as an investable asset class, characterized by low per-token cost, high revenue from hardware utilization, a long operating life and a broad user ecosystem built around the CUDA platform. Nvidia CEO Jensen Huang described it as a natural extension of the company's role.

We started by building chips, and today we're helping create a new class of productive, investable infrastructure: AI factories - Jensen Huang, CEO of Nvidia

The mechanism is meant to work by having the six financial institutions channel outside investors' capital to independent platforms building infrastructure based on Nvidia hardware, rather than putting that risk on Nvidia itself. The financing is meant to include long-term compute-usage agreements tied to revenue, making the investment more predictable for pension funds or insurers seeking stable returns. Gray cautioned, however, that not every AI company valued highly today will survive the current wave of consolidation, and not every location has equal access to cheap energy.

What it means for Poland

The scale of capital referenced in the Nvidia agreement is beyond the reach of individual Central European countries, but the direction is clear: global private capital now treats AI computing infrastructure as a distinct asset class, alongside real estate or energy infrastructure. For Poland, which is simultaneously seeking EU funding for AI gigafactories and seeing rising data center demand for electricity, this means growing competition for the same resources: grid access, land and administrative permits.

Polish energy companies and grid operators already warn that connecting new large power consumers, including data centers, requires years of investment in transmission infrastructure. If Blackstone's argument about a structural computing power shortage holds true, pressure on the same energy bottlenecks already affecting the Polish market will keep growing with every new wave of global capital flowing into the sector.