Staff Reporter | Song Jianan
On August 10, local time, NVIDIA announced a series of strategic partnerships, joining forces with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish an independent compute financing platform. The plan is to gradually mobilize more than $500 billion in third-party capital to fuel AI infrastructure buildout.
NVIDIA has already inked memorandums of understanding with these six top-tier global financial institutions. Here’s how the platform works: these six firms will set up large-scale dedicated capital pools and offer financing support at competitive interest rates to cutting-edge AI labs, enterprises, and AI cloud service providers within NVIDIA’s ecosystem. With this third-party funding, customers can purchase hardware and build DSX-architecture AI factories without draining their own balance sheets. On the flip side, the capital providers earn long-term returns tied to actual compute usage—essentially turning AI compute projects into investment-grade assets, much like public utilities, that global institutional capital can confidently allocate to.
“NVIDIA has reached a significant milestone. We started out focused on chip design, and now we’re helping create an entirely new class of productive, investable infrastructure—AI factories,” said Jensen Huang, founder and CEO of NVIDIA.
In Huang’s view, “In the AI era, compute is revenue.” He noted that NVIDIA’s compute isn’t just widely adopted across various models and workloads—it also keeps improving through continuous CUDA software updates, extending lifecycles and optimizing economic returns over time. Behind it all sits a global ecosystem built by developers, customers, and downstream demand.
“That’s why we’ve brought together the world’s leading long-term capital institutions to independently provide credit support for AI infrastructure projects. These financing platforms will help customers access scarce compute at scale, build DSX AI factories, and empower industries and countries across the AI era,” Huang explained, articulating the rationale behind launching this financing platform.
The most intriguing aspect of this collaboration lies in how it redefines the asset class. For a long time, GPUs have been viewed as fast-depreciating electronic hardware—hardly ideal collateral for securing long-term, low-cost credit. But NVIDIA is attempting a fundamental cognitive shift here: positioning the entire CUDA-based compute cluster as an investable asset with stable cash flows that can move across customers.
Global demand for AI compute continues to explode, yet AI infrastructure is hitting a very real bottleneck—and the funding gap is the most glaring issue. AI data centers are quintessential heavy-asset projects: massive upfront spending on hardware, facilities, and power infrastructure, with painfully long payback periods. Big tech companies keep ramping up capital expenditures, but these intense investments are already squeezing free cash flow. Meanwhile, countless mid-sized AI firms and startup labs—even those with promising model research—simply can’t shoulder the enormous upfront cost of building a 10,000-GPU cluster, nor can they secure the long-term, low-interest loans needed to build out their own compute foundations.
At the same time, traditional financial institutions remain cautious about compute assets. Hardware iterates quickly, chip performance upgrade cycles keep shrinking, and the book value of hardware assets steadily erodes. If a project underperforms, disposing of or transferring the mortgaged hardware becomes a real headache. This is why massive pools of long-term institutional money sit on the sidelines, holding plenty of capital but struggling to flow into the AI compute track—creating this awkward mismatch where the industry is starving for compute while capital waits on the fence.
NVIDIA clearly wants to change that dynamic. Historically, the company made money by selling chips and server hardware. This new model leverages external capital to help customers complete hardware purchases and cluster builds, indirectly expanding the market ceiling for both hardware and software products.
The financial institutions involved each shared their own perspectives on the partnership.
Jim Zelter, President of Apollo, noted that modern compute has become a scarce and critical core asset with highly attractive investment characteristics, one that promises to drive long-term economic growth and substantial productivity gains.
“AI infrastructure buildout demands investment on an unprecedented scale, and it requires specialized expertise to convert capital into the hardware foundation that will support future growth,” said Larry Fink, Chairman and CEO of BlackRock. “BlackRock will combine NVIDIA’s leadership in accelerated computing with our ability to connect long-term capital with core infrastructure projects.”
For KKR Co-CEOs Joe Bae and Scott Nuttall, compute has already become a critical infrastructure asset. In building out digital infrastructure, the hardest part isn’t envisioning the blueprint—it’s delivering on the ground.
That said, these collaborations still need formal agreements to be signed before they become reality. The grand financing vision doesn’t mean industry risks have vanished. Project approvals, interest rate pricing, default handling, and hardware asset valuation all still require extensive fine-tuning. And once institutional capital floods in at scale, the market also needs to guard against overheated investment leading to reckless construction—avoiding a mismatch between compute projects and actual commercial demand.
Looking at the broader industry, the symbolic significance of this partnership outweighs the near-term capital deployment itself. It’s an attempt to bridge the gap between the tech sector and Wall Street’s long-term capital, pioneering a brand-new financing and investment paradigm for AI infrastructure. If the model successfully scales, it will lower the capital barrier for companies building AI factories and accelerate the pace of global AI infrastructure expansion.
But it’s also important to keep things in perspective: $500 billion represents the ceiling for long-term mobilizable capital, not an immediate shift in the current global compute supply-demand landscape. Where the industry ultimately heads still depends on how quickly AI commercialization lands and how these projects actually perform on the ground.
