AI cloud infrastructure starts competing for “power financing”

What happened

On 16/7, Industrial Development Funding (IDF) and alternative asset manager Oaktree said they will invest $1.7 billion in Nebius’s AI cloud infrastructure project to deploy Bloom Energy fuel cell systems. When the project is completed, Nebius will have “behind-the-meter power” — meaning power generated and used directly on the data center customer side, rather than waiting entirely for the public grid to expand.

The structure of the deal is also notable: IDF is serving as the project’s lead developer, Oaktree is participating as a minority equity investor; Morgan Stanley is handling tax equity and distribution financing, while MUFG Bank is providing senior debt financing. In other words, the funding target for AI infrastructure is shifting from servers, chips, and data center buildings to a layer of power assets that can stand on their own as financeable, buildable, and operable projects.

Previously, Nebius announced a partnership with Bloom Energy to apply fuel cell technology in the expansion of its AI infrastructure in the U.S. The company’s core rationale was deployment speed, power supply capacity, and the ability to support the performance and availability of AI workloads. This $1.7 billion project investment takes that technology partnership one step further at the capital-structure level.

Why it matters

The bottleneck for AI cloud services is no longer just “whether there are GPUs.” Model training and inference require high-density, continuous, and predictable power, while connecting large data centers to the grid is often constrained by interconnection queues, transmission buildouts, and permitting processes. Behind-the-meter power is designed to place part of the generation capacity directly next to the load, reducing dependence on new transmission capacity and the public grid’s delivery speed.

This is also a deal with market-structure implications. On 15/7, Reuters reported that Wall Street banks are seeing rising demand for equity, debt, and project financing tied to AI infrastructure. One day later, the Nebius project with IDF, Oaktree, Morgan Stanley, and MUFG shows another capital path: letting infrastructure capital hold the power asset, complete construction with tax equity and bank debt, and then serve AI cloud computing demand.

For investors, the “observation window” for AI capex has therefore expanded. In the past, the market often used GPU purchases, cloud providers’ capex, or data center valuations to gauge the sector’s heat; now, fuel cells, nuclear power, renewable energy, transmission, and project financing conditions can also become key variables determining whether computing capacity can come online on time. The willingness of capital to assign a separate value to “power arriving sooner” shows that time-to-power is becoming a scarce asset in AI infrastructure.

What to watch next

First, the specific construction progress, commissioning timeline, and actual power delivery scale corresponding to the $1.7 billion still need to be monitored closely. The press release confirmed the investment structure and financing roles, but did not disclose the project’s full return profile, debt tenor, fuel costs, or Nebius’s long-term power purchase terms.

Second, whether fuel cells can reliably meet the 24/7 load of AI data centers still needs to be proven in real operations. Power delivery speed is only the starting point; system availability, maintenance cycles, fuel supply, and the cost of electricity per unit will all ultimately be reflected in the gross margin of AI cloud services.

Third, it remains to be seen whether this project expands from a single deployment into a replicable financing template. If more behind-the-meter power projects backed by institutional capital emerge afterward, AI infrastructure could form a new combination of “compute rental + power assets + structured financing”; if the project runs late or proves less economical than expected, capital markets will also have to reassess leverage risks in the AI buildout cycle.

Source

• Industrial Development Funding and Oaktree announcement of a $1.7 billion project investment in Bloom Energy fuel cells for Nebius AI infrastructure build-out (primary source)
• Nebius and Bloom Energy announcement on powering AI infrastructure build-out
• Reuters: Wall Street banks see AI “super cycle” set to boost deals and financing