AI infrastructure is increasingly being viewed not just as technology spending, but as a new category of investable infrastructure, with Nvidia and some of the world’s biggest financial institutions looking to channel hundreds of billions of dollars of institutional capital into data centers, power and computing capacity.

The shift reflects a broader change in how investors view AI infrastructure. As computing capacity becomes critical to businesses and AI workloads generate revenue, GPUs, data centers and related power infrastructure are increasingly being treated as assets that can be financed, valued and potentially securitised.

Nvidia CEO Jensen Huang and senior executives from Goldman Sachs, Apollo Global Management, BlackRock, Blackstone, Brookfield and KKR discussed the emerging investment opportunity in an exclusive interview with CNBC.

Nvidia has partnered with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent financing platforms aimed at bringing long-term institutional capital into AI infrastructure. The initiative is targeting more than $500 billion in third-party capital.

For investors, the significance goes beyond the size of the funding target. It points to the potential creation of a new infrastructure asset class built around the physical systems required to deliver AI — from power generation and data centers to GPUs, networking and compute capacity.

From technology products to financeable infrastructure

Huang said the investment model represents a fundamental change in how computing technology is being viewed by financial markets.

“This is really the first time that technology chips have become an investable asset class,” Huang said.

The shift is being driven partly by the economics of AI. As companies use AI to perform productive, revenue-generating work, demand for computing capacity is translating into a need for large, long-duration investments in physical infrastructure.

Huang said each gigawatt of AI infrastructure could cost around $50 billion-$60 billion to build, including energy, land, power, data-centre shells and computing equipment. Nvidia estimates that the US alone could require more than 70 gigawatts of power to support AI infrastructure.

That scale of investment is increasingly difficult for technology companies or individual investors to fund on their own, creating an opening for banks, private equity firms, private credit providers and other institutional investors.

Wall Street sees a multi-trillion-dollar opportunity

The $500 billion target being pursued by Nvidia and its financial partners may represent only the beginning of the capital required for the AI buildout.

BlackRock Chairman and CEO Larry Fink said the financing target should be viewed as an initial step, with the overall investment requirement potentially reaching trillions of dollars.

“We’re going to have to raise trillions of dollars over the coming years,” Fink said.

Fink also argued that the AI infrastructure buildout represents a broader economic opportunity, with investment in data centers, power and related infrastructure potentially supporting jobs and economic activity.

His comments underscore the scale of the transition. AI investment is no longer limited to software companies and semiconductor manufacturers; it is increasingly pulling capital into electricity generation, transmission, construction, land and physical computing infrastructure.

Why financiers believe AI infrastructure can be funded

Goldman Sachs CEO David Solomon said global capital markets have enough capacity to finance a significant portion of the investment required.

The US has around $9 trillion in money-market funds and more than $100 trillion in equities, Solomon said, arguing that the role of financial institutions is to channel capital towards productive investments.

“These are real assets. They have real value. You can put a tangible value on it,” Solomon said.

Solomon cautioned that the financing cycle would not move in a straight line, with some AI infrastructure projects likely to generate strong returns while others could fail. However, he expects productivity gains from AI to support stronger economic growth over the longer term.

Blackstone President and COO Jon Gray pointed to the widening gap between demand for computing and available infrastructure.

Gray said demand for large language models at Blackstone’s companies had increased sevenfold in six months, while data centers, power and computing capacity had failed to keep pace.

That imbalance could give investors greater visibility into the demand for the infrastructure they are financing.

GPUs could become infrastructure assets

The potential financialisation of computing goes beyond traditional data centers.

Apollo President Jim Zelter said AI infrastructure financing is moving into the mainstream as GPUs and compute become financial assets that can be funded through a wider range of capital markets.

The distinction is important because AI infrastructure combines traditional infrastructure characteristics with rapidly evolving technology.

Power plants and data-center buildings can have long economic lives, while GPUs can become technologically outdated much faster. Yet the ability of older generations of chips to continue generating revenue could allow computing assets to be financed and potentially securitised.

KKR’s global head of digital infrastructure, Waldemar Szlezak, highlighted the increasing fungibility of AI computing infrastructure, suggesting that compute assets could eventually be financed, securitised and sold to investors.

That could open up a financial model in which investors do not necessarily need to own an entire data center or power project to gain exposure to AI infrastructure.

The bottleneck is increasingly physical

The push to bring institutional capital into AI infrastructure is also being driven by supply constraints.

Huang warned that the AI infrastructure supply chain will remain constrained for some time, with bottlenecks extending beyond GPUs to memory, packaging, systems, photonics, connectors, land, power and construction capacity.

“We’re going to be constrained for some time,” Huang said.

Brookfield CEO Bruce Flatt said the fundamental problem was not a shortage of financing but the industry’s ability to build infrastructure quickly enough.

“We cannot build enough power. We cannot build enough compute for the demand,” Flatt said.

Brookfield has expanded from traditional investments in power infrastructure, including solar, wind and gas, into data centers and increasingly into AI computing infrastructure.

That shift illustrates why AI infrastructure is beginning to resemble a conventional infrastructure investment theme. Investors are not simply betting on the success of an individual AI model or company; they can potentially invest in the physical capacity required by the entire ecosystem.

A new infrastructure investment cycle

KKR’s Szlezak described the AI infrastructure opportunity as a “generational investment opportunity”, spanning power, data centers and the compute layer.

He compared the speed of the current buildout with earlier waves of internet and cloud infrastructure, while noting that computing capacity is now being added at a much faster pace.

For financial markets, the potential significance is that AI could create a new class of infrastructure assets with multiple layers of investable exposure.

At one end are power generation and transmission assets. Next are land, data centers and cooling systems. Above that sit networking equipment, servers and GPUs. The revenue generated from providing compute capacity could then provide the cash flows needed to support different forms of financing.

The model is still developing, and the risks remain significant. AI demand could change, technology can become obsolete rapidly, and projects built on aggressive assumptions about future computing demand could struggle to generate expected returns.

But the involvement of some of the world’s largest asset managers and banks signals that AI infrastructure is increasingly being evaluated through the same financial lens applied to other large infrastructure investments.

AI investment could extend far beyond technology

For Huang, the investment opportunity ultimately goes beyond the infrastructure required to run AI.

He said AI represents a fundamental change in computing that could affect virtually every industry.

“This is going to impact literally every single trade,” he said.

If that prediction plays out, the financial opportunity created by AI infrastructure could extend well beyond Nvidia and other semiconductor companies. Every additional AI application could create demand for computing, which in turn requires more power, data centers, networking and chips.

That creates the central investment thesis emerging from the discussions between Nvidia and Wall Street: AI may be the technology story, but the infrastructure required to run it could become a major new asset class in its own right.

And the $500 billion capital target could be only the first step in building that market.