Nvidia Corp‘s (NASDAQ:NVDA) AI story is entering a new phase, but investors appear increasingly focused on what could go wrong.
Memory costs, margins, custom chips and Nvidia’s investments in AI customers have all emerged as concerns. Bank of America, however, thinks investors may be overlooking what could drive the chipmaker’s next leg higher.
In a note shared Monday, Bank of America analyst Vivek Arya called Nvidia a “top pick" with a $350 price objective.
The target implied 56.3% upside from Nvidia’s $223.96 price at the time.
The key is Nvidia’s upcoming Rubin architecture.
Nvidia Stock And The Rubin Era
Nvidia is expected to report its second-quarter results on Aug. 26.
BofA expects revenue of $94 billion–$95 billion, about $3 billion–$4 billion above Nvidia’s $91 billion guidance.
It also expects third-quarter guidance of $107 billion–$108 billion, above the roughly $104 billion consensus estimate.
But Arya anticipates that the start of Vera Rubin shipments, new Vera CPU ramps and continued cloud capital spending will trigger a "multi-quarter upgrade cycle."
That could make Rubin more important than the immediate earnings reaction.
Nvidia’s next-generation systems are also becoming significantly more valuable.
BofA estimates Vera Rubin NVL racks could cost roughly $7 million–$8.5 million, compared with about $4 million for Blackwell Ultra. That creates an important cushion against rising component costs.
Why Memory May Not Break Nvidia’s Margins
Memory inflation has become one of the biggest concerns surrounding the AI trade.
BofA estimates that higher memory costs create only about a 60-basis-point gross-margin headwind for Nvidia’s Rubin compute racks.
The bank still expects long-term gross margins around 73%–74%, versus roughly 75% currently.
Why so little impact? Nvidia has pricing power, preferential memory sourcing and the ability to adjust prices as component costs change.
The economics look even stronger when GPU rental prices are considered.
BofA’s data show A100, H100 and B200 rental prices near historical highs as of August. That suggests customers continue to place significant value on scarce Nvidia compute capacity.
For investors, that matters more than the absolute cost of memory.
If Nvidia can raise system prices faster than memory costs rise, Rubin could expand the company’s revenue opportunity without destroying its margins.
The Valuation Is The Real Bull Case
Nvidia trades at roughly 16 times forward earnings, which BofA describes as the “cheapest valuation” in about a decade. Yet BofA forecasts Nvidia’s EPS rising from $4.55 in calendar 2026 to $9.09 in calendar 2027, nearly doubling in one year.
That creates an unusual setup.
Nvidia is entering another major AI product cycle while its valuation has compressed sharply relative to its historical growth profile.
The Rubin era, therefore, is not simply another Nvidia product launch. It could be the catalyst that forces investors to rethink both Nvidia’s earnings power and its valuation.
Image: Shutterstock
Login to comment