Nvidia CEO Jensen Huang speaks to members of media exterior a restaurant within the Hongdae district of Seoul, South Korea, June 5, 2026.
SeongJoon Cho | Bloomberg | Getty Photographs
Jensen Huang constructed the world’s most precious firm by pioneering the specialised laptop chips behind the bogus intelligence growth.
To maintain his imaginative and prescient for the long run inside attain, the Nvidia founder is now trying a special type of engineering: convincing Wall Road buyers that these chips are long-term monetary property akin to business actual property or toll roads.
His guess hinges on outpacing AI developments in China.
This week, Nvidia unveiled agreements with six of the world’s largest asset managers, BlackRock, Blackstone, Apollo, KKR, Brookfield and Goldman Sachs. The purpose was to assemble a $500 billion pipeline to finance the development of knowledge facilities and GPU clusters for corporations that lack the credit standing or money to purchase hundreds of thousands of {dollars} of silicon outright.
Key to his plan, which Huang introduced throughout a CNBC phase flanked by the leaders of all six Wall Road corporations, is one essential assumption: that Nvidia’s graphics processing items will maintain their worth over time, behaving extra like conventional onerous property than fast-depreciating shopper electronics.
“Nvidia’s AI manufacturing unit platform is admittedly an investable asset, an infrastructure asset,” Huang mentioned. “The explanation for that’s as a result of it is productive, it is income producing, it’s fungible, it is utilized by nearly each cloud service supplier, it runs each AI mannequin.”
In customary asset-backed finance, a financial institution lends cash as a result of if a borrower defaults, the financial institution can repossess the asset — like a constructing, a warehouse or a cargo ship — and promote it to get their a reimbursement. These bodily property have established secondary markets and might final a long time.
However the productive lifespan of cutting-edge GPUs is way from settled.
Whereas new chips energy frontier mannequin coaching, after just a few years they’re relegated to lower-margin inference work — a shift that instantly impacts their resale and collateral worth.
“Depreciation is the one key threat right here,” mentioned Ben Emons, founding father of FedWatch Advisors, who structured comparable asset-backed loans for IndyMac earlier than becoming a member of Pimco as a portfolio supervisor. Nvidia chips “may depreciate quicker than anticipated,” he mentioned.
Excessive-yield charges?
Particularly, Emons mentioned he believes the one greatest menace to Nvidia’s financing mannequin comes from China, which is quickly ramping up home compute capability and will select to flood the market with low-cost silicon in a worth conflict.
If Chinese language manufacturing pushes {hardware} costs right into a freefall, the collateral backing tons of of billions in personal loans may erode far quicker than the phrases of the debt itself, leaving buyers uncovered to losses, in line with Emons.
To compensate a minimum of partly for that threat, Emons estimates buyers will deal with GPUs as high-depreciation gear reasonably than actual property, demanding high-yield returns within the 11% to 17% vary relying on the place they sit within the capital construction.
On prime of that, the debtors are more likely to be non-investment grade corporations locked out of conventional debt markets, together with AI startups and neoclouds, in line with a Financial institution of America Securities word.
If these higher-risk debtors go below, Wall Road fund managers might be compelled to repossess and resell used chips right into a doubtlessly falling market.
No matter dangers China poses would not be realized anytime quickly. Huawei, the dominant supplier of Chinese language AI chips, has been on the U.S. Commerce Division’s Entity Record since 2019. And in Could, the U.S. authorities mentioned Huawei’s Ascend AI chips violate U.S. export controls, stopping any American firm from utilizing the chips.
Within the meantime, Nvidia stays by far the main provider of AI chips within the U.S., with upwards of 75% market share by most estimates.
And for now, the economics are nonetheless transferring in Huang’s favor. Pushed by shortage as hyperscalers race to construct out capability, rental charges for Nvidia’s H100 chips rose from roughly $1.70 per GPU-hour in late 2025 to about $2.35 per GPU-hour this 12 months, Huang famous.
Crucially, Nvidia argues its CUDA software program layer — which permits builders to run AI workloads on its GPUs — repeatedly improves {hardware} efficiency after deployment, permitting older chips to remain productive and generate yield longer than conventional accounting fashions predict.
The way forward for the AI buildout, and tons of of billions of {dollars} in investor cash, could rely on who is true.
— CNBC’s Ari Levy contributed to this report.
