Key Takeaways:
Nvidia is underwriting neocloud data centers in exchange for a slice of their revenue, a model Morgan Stanley says could add 60% to fiscal 2028 profit.
Nvidia is shifting from chip seller to AI infrastructure financier, guaranteeing GPU lease prices for neocloud builders in exchange for a share of revenue above that floor — a structure Morgan Stanley estimates could unlock a $500 billion annual revenue pool.
"This is AI-era financial innovation," Joseph Moore, semiconductor analyst at Morgan Stanley, said in a note that kept an overweight rating and $288 price target on the stock, which trades near $225.30.
Under the model, Nvidia negotiates a floor price per GPU-hour with each cloud provider, which uses that guarantee to secure near-investment-grade financing for data center construction. Nvidia then takes a cut of lease revenue above the floor, with the split varying by deal. Morgan Stanley's base case assumes a 35 percent share, adding $37 billion to $65 billion in annual revenue per 5 gigawatts of capacity at lease prices of $10 to $16 an hour. In some cases the backing is not a cash credit guarantee but an exchange for GPU access for Nvidia's own research, similar to its earlier CoreWeave arrangement.
The model turns Nvidia's stock into a bet on recurring, annuity-like income rather than one-off chip sales. If neoclouds scale to 25 gigawatts — comparable to the largest hyperscalers — the revenue pool reaches $500 billion a year, and a quarter of that profit at full margin would lift fiscal 2028 earnings before interest and taxes by 60 percent.
The strategy responds to a financing gap among the hyperscalers that buy most of Nvidia's chips. Management argues that Microsoft, Amazon and Alphabet face free-cash-flow pressure and tightening land, power and space constraints in the U.S., forcing them to divert compute from training to inference. Geopolitics is accelerating the shift: as U.S. clouds face policy pressure building data centers abroad, some governments now treat local land and power as strategic AI assets and are backing sovereign or regional neoclouds such as Yotta, YTL and SK Telecom.
Nvidia has been seeding this ecosystem since its 2023 investment in CoreWeave. It now holds more than 10 percent of CoreWeave and has committed to buy all unsold compute through April 2032, owns roughly 10 percent of Nebius, holds a $3.4 billion five-year cloud contract with IREN plus $2.1 billion in warrants, and has taken private stakes in Nscale, Crusoe, Lambda and Firmus. On the software side, its DGX Cloud Lepton platform routes developer demand to whichever cloud partner has spare capacity. In hardware, Nvidia and partners are building 16 new fiber routes spanning 8,000 miles, and it is reportedly signing a 15-year, $200 billion lease for HUT 8's 1-gigawatt Beacon Point campus. A newly announced $50 billion investment fund will back long-term usage-linked returns, with direct exposure capped at 25 percent per deal.
Bulls argue compute supply still trails token demand, that minority stakes across multiple clouds make earnings more predictable, and that the model locks in Nvidia's roughly 85 percent GPU market share ahead of the Vera Rubin cycle — smaller clouds still choose the most complete ecosystem. Bears counter that the backing creates artificial demand, that some capacity would not exist without Nvidia's guarantee, and that the structure is exposed to a compute glut. Morgan Stanley said transparent disclosure is the condition for market trust.
In a downturn, Nvidia has a hedge: GPU access it receives can feed its own model development. Its flagship open-source Nemotron 3 Ultra has 550 billion parameters, and the next Nemotron 4 is expected to cross one trillion.
The model also pressures existing neoclouds. More competitors entering the market is a long-term negative, though operators with a track record of on-time delivery can win incremental investment. Nvidia's backing solves financing, but builders still must complete construction and convert capacity to revenue amid supply-chain and geopolitical friction.
For investors, the question is whether the market prices Nvidia as a chipmaker or a financier. Morgan Stanley forecasts 82 percent revenue growth in fiscal 2026 and 52.4 percent in fiscal 2027, with fiscal 2029 earnings per share of $17.63 — about 22 times the current price. Mark Cuban has compared the strategy to a dot-com-era IPO machine "funding everyone and anyone," a warning that the model's upside depends on demand staying ahead of the capacity Nvidia is helping to build.
This article is for informational purposes only and does not constitute investment advice.
