
Bank of America: After accounting for AI financing risks, NVIDIA is still undervalued by 34% to 50% based on free cash flow
Bank of America analyst Vivek Arya maintains a buy rating on NVIDIA with a target price of $350. The research report points out that even accounting for approximately $300 billion in vendor financing and off-balance-sheet commitment risks, NVIDIA's stock price is still undervalued by 34% to 50% compared to its estimated fair value, presenting a positioning opportunity. The current stock price of $219 implies an upside potential of 55%-60%. The market is concerned that its role as a "financial intermediary" may affect earnings quality, but the analyst believes the discount overestimates the risks
Bank of America analyst Vivek Arya maintained a buy rating on NVIDIA with a target price of $350 in his latest research report, providing a segmental sum calculation based on free cash flow: even including vendor financing and off-balance-sheet commitments, the stock price is still discounted by 34% to 50% relative to its estimated fair value. Arya believes that the discount may overestimate the associated risks, thus constituting an attractive allocation opportunity. On Tuesday, NVIDIA's stock fell about 2.5% to $219, participating in the general pullback of the artificial intelligence sector; it has risen about 18% year-to-date, but is still about 7% lower than its peak in May. The target price implies an upside potential of about 55% to 60% relative to the stock price at that time. The earnings report is scheduled for August 26, and the bank expects the company to increase disclosures regarding off-balance-sheet commitments.

The market discount is directed at "chip manufacturers also acting as financial intermediaries."
Recent core concerns among investors are not about NVIDIA's chip market share, but rather its provision of capital for the artificial intelligence cycle. Bank of America estimates that the company has committed approximately $300 billion in support to ecosystem partners, including about $70 billion in equity investments and approximately $230 billion in remaining value guarantees or backstops. Publicly reported projects include: approximately $40 billion invested in leading U.S. laboratories, with about $30 billion for OpenAI and about $10 billion for Anthropic; up to $105 billion in related support for a data center park in Ohio to be leased by OpenAI; and collaboration with six large asset management firms to mobilize about $500 billion in third-party capital for data centers and related projects. The company also disclosed substantial equity holdings in SpaceX and Intel.
Arya described these investments as necessary to accelerate the cycle, especially for leading laboratories and emerging cloud providers, but acknowledged that they may reduce profit quality and suppress trading multiples. Regarding the Ohio project, some reports indicate that the guarantees cover the remaining gap upon re-leasing or selling after a default, rather than the entire rent of the lessee, with a cap of $105 billion. Even so, off-balance-sheet risks are "significant but difficult to value precisely," which is precisely why multiples are compressed.
Valuation framework: Even after applying a 50% discount to investment-related cash flows, there is still upside.
Calculations cited by Benzinga suggest that NVIDIA could generate approximately $470 billion in free cash flow over the calendar years 2026 and 2027. The current stock price is roughly equivalent to 18 times the free cash flow for 2027 and 15 times for 2028, while comparable AI semiconductor companies average about 38 times and 25 times. Even applying a 50% discount to investment-related cash flows, under the bank's conceptual framework, the company could still be about 50% higher than the current price in the calendar year 2027 and about 34% higher in 2028 The source of this set of numbers, "34% to 50% discount," is not simply the target price minus the stock price, but rather a comparison of cash flow multiples after accounting for financing risks.
Arya stated that the current environment makes the aforementioned strategy still worth pursuing: rental prices for graphics processors remain strong, computing power is still scarce, and the company's free cash flow generation capability is among the industry leaders. He also provided a counter-condition: if demand for artificial intelligence slows, growth rates and balance sheets may come under pressure simultaneously. In other words, being bullish is based on the premise that the cycle continues, while being bearish views guarantees as contingent liabilities in the event of a demand downturn.
The repricing path given by banks is to increase buybacks, rather than deny risks.
The research report believes that the most effective countermeasure to alleviate concerns about profit quality and promote multiple recovery is to increase the proportion of free cash flow used for buybacks. Larger buybacks can directly return cash and partially hedge against the narrative of "chipmakers becoming cyclical lenders." Therefore, the earnings call on August 26 has a dual pricing function: first, to verify demand and guidance, and second, to disclose the scale, triggering conditions, and accounting treatment of off-balance-sheet commitments. Prior to this, the market could only price risks using a rough estimate of approximately $300 billion in commitments, resulting in a naturally wide discount range.
Peer performance serves as a comparison. Reports indicate that both Intel and AMD have seen their stock prices more than double this year, while NVIDIA has only increased by about 18%. Some funds have shifted risk premiums from "growth stocks" to "challengers with simpler balance sheets." Bank of America holds the opposite stance: ecological investment is a tool to maintain the cycle, and the discount has overly compensated. This is the analysts' viewpoint, not an already realized valuation convergence.
Implications for trading positions and unresolved questions.
If one accepts Bank of America's framework, the logic for buying around $219 is that after guarantees are accounted for as worst-case scenarios, the core business cash flow is still significantly underestimated. If one rejects this framework, then the $300 billion commitment creates a cycle of financing between the company and its customers and lessees. If rental rates or capital expenditures for model training decline, guarantees may turn into real cash outflows, and the existing multiple discount may still be insufficient. Both interpretations cannot be falsified before August 26.
Matters that need to be verified separately include: the precise legal structure of the $105 billion arrangement, how the fair value fluctuations of the $70 billion equity enter profit and loss, the actual risk share borne by NVIDIA in the $500 billion third-party capital, and whether buyback guidance will be raised. Before these numbers are disclosed by the company, the 34% to 50% discount is merely a conceptual model from a sell-side perspective. Whether the stock price approaches $350 depends on whether the market agrees to view ecological financing as a cycle acceleration rather than hidden leverage
