Concerns about Nvidia have dragged the stock's valuation near historically low levels - and that suggests big gains could be in store, according to a BofA analyst.
Shares of the chip maker (NVDA) are currently trading at a forward price-to-earnings multiple of about 25. That's similar to previous recent "troughs" in October 2023 and July 2022, BofA's Vivek Arya said in a Monday note.
The good news for Nvidia investors is that in the prior instances where the stock's multiple reached these levels, its forward P/E "usually rebounded" to a range between 30 and 40 in the following three to six months, Arya said. Over the past five years, that's led to a median forward price-to-earnings multiple of 37, he noted.
See also: Alphabet and Nvidia alone make up a third of the S&P 500's gains this year, in a sign of Big Tech's dominance
Additionally, Nvidia shares are trading at their widest-ever discount to shares of Broadcom $(AVGO)$ - about 40%, according to Arya, who pointed out that usually Nvidia shares trade at about a 10% discount.
On a conceptual basis, this reflects consensus estimates "already implicitly" shifting at least 10 points of AI market share toward Broadcom when looking at the second half of 2026 and the whole of 2027, Arya said.
Investors increasingly are debating the impact that application-specific integrated circuits designed by players like Broadcom will have on demand for graphics processing units from Nvidia and Advanced Micro Devices $(AMD)$.
Alphabet $(GOOGL)$ $(GOOG)$ saw its shares rally last month after Google introduced Gemini 3, which it said outperformed top AI models from competitors such as OpenAI and Anthropic. The tech giant said it used its tensor processing units to train the AI model, leading to jitters in a market already anxious about whether Nvidia would be able to hold on to all its market share in a competitive industry. Google has worked on its TPUs with Broadcom for more than a decade.
However, Arya said tight chip supply and Nvidia's "leading allocation" from Taiwan Semiconductor Manufacturing Co. $(TSM)$ (TW:2330), which also manufactures chips for Broadcom, will likely make it harder for customers to change over the next year. And Google's custom chips "have been proven only in Google's data center," while Nvidia's chips have been more widely used.
Morgan Stanley analyst Joseph Moore said he expects Nvidia to keep "dominant market share," as recent worries over the threat of ASICs "are becoming overstated." However, Moore said in a Monday note that it's not clear what will be able to flip negative investor sentiment.
Moore's Morgan Stanley team is modeling for Broadcom and AMD to see revenue from AI chips grow "slightly faster" than Nvidia's in 2026, "but that mostly just reflects supply-chain limitations of a $205 [billion] run-rate revenue stream," Moore said. Supply of key AI components is expected to be constrained through next year.
Over the next 12 months, Moore said "customers' biggest anxiety" revolves around getting enough of Nvidia's products, especially with its upcoming Rubin AI platform.
"Of course, everyone wants an alternative, and those alternatives will have good economics for some applications," Moore said. "TPU is by all accounts a solid alternative, and one that has meaningfully contributed to several key models."
However, Nvidia's $51 billion in data-center revenue in the last quarter is about 14 times Google's TPU revenue, Moore noted, and the $10 billion that Nvidia just logged in sequential revenue growth is about three times that total TPU number.