Bank of America: Nvidia's Forward P/E Falls to 7-Year Low, Market Paying for a Non-Existent Risk

On July 9, Bank of America pointed out in a research report that Nvidia's current valuation has dropped to its lowest level since the start of the AI bull market—its forward P/E is only about 18 times, the lowest in nearly seven years, compared to an expected P/E of about 21.5 times for fiscal year 2027 and further down to 14.7 times for fiscal year 2028. This valuation is not only far below its historical range but also represents a discount of about 30%-35% compared to the average expected P/E of 22 times for 2027 and 19 times for 2028 among the five major tech giants: Apple, Microsoft, Google, Amazon, and Meta.

Bank of America's core judgment is that the current market pricing implies a downward expectation of 30%-35% for Nvidia's EPS in 2027/2028, and this assumption "is fundamentally unfounded." From a PEG perspective, Nvidia's PEG for 2027 is only 0.3 times, far lower than Apple's 2.7 times, Microsoft's 1.0 times, and Google's 1.9 times. So far this year, Nvidia's stock price has only risen by 3%, significantly lagging behind the Philadelphia Semiconductor Index's increase of 82%. Bank of America believes this significant divergence reveals an "enhanced buying opportunity," reiterating a buy rating with a target price of $350, implying over 70% upside from the current stock price.

Bank of America further dissected the two core concerns of the market. Regarding the impact of memory costs, Bank of America believes the market has overestimated cost pressures while underestimating Nvidia's pricing power—upgrading from Blackwell to Vera Rubin incurs an incremental HBM cost of about $200,000 to $300,000 per rack, but the entire rack's selling price is expected to rise by $2 million to $3 million to $6 million to $7 million. The driving force comes not only from memory but also from components that do not require HBM, such as Vera CPUs, NVLink, and Quantum Ethernet networks, as well as a series of software features that reduce inference costs. Nvidia has built a strong moat through over $119 billion in supply chain pre-purchase commitments, and gross margins are expected to remain in the mid-70% range. Regarding ASIC competition, Bank of America directly responded with data—since the launch of Google's TPU in 2015, Nvidia's GPU accelerator sales have grown by about 700 times, with sales to hyperscale cloud vendors increasing by 115% year-on-year, about twice the growth rate of cloud computing capital expenditures, indicating that its wallet share among hyperscale customers continues to expand.

Bank of America expects Nvidia to maintain a market share of over 65%-70% in AI computing capital expenditures in the long term, and the upcoming earnings call is expected to be a key catalyst for clarifying its pricing power and moat to the market.