What Wall Street is looking for from Nvidia’s earnings report
Access to China, gross margins, the Blackwell ramp, and sovereign AI will be in focus.
Nvidia, the reason why earnings season seems never-ending, releases its fiscal 2026 first-quarter results after the close on Wednesday.
Analysts polled by Bloomberg are looking for adjusted earnings per share of $0.88 on revenues of $43.4 billion, with more than 90% of sales tied to its data center business. Gross margins, a spot of bother in its Q4 earnings report, are expected to come in at about 71%. Management’s guidance is for revenues between $42.14 billion to $43.86 billion for the quarter, with an adjusted gross margin between 70.5% and 71.5%.
What we “know,” thanks primarily to hyperscalers’ earnings reports we got about a month ago, is that the AI boom rolls on. Tokens (data processed by AI models) govern how much demand there will be for chips to support generative-AI capabilities (and more!).
“Every hyperscaler has reported unanticipated strong token growth,” Morgan Stanley analysts led by Joseph Moore wrote. “But our conviction is not driven by that, it’s driven by the fact that literally everyone we talk to in the space is telling us that they have been surprised by inference demand, and there is a scramble to add GPUs.”
We’ll see how that’s reflected in any Q2 guidance, where Wall Street is looking for adjusted earnings per share of $1.01, sales of $46.275 billion, and adjusted gross margins of 72%.
The Great Chipwall
In the past, management has downplayed its exposure to China; lately, CEO Jensen Huang is hyping up the opportunity set in the world’s second-largest economy, and reportedly has a tailor-made AI chip for China slated for mass production next month. The H20 export restrictions were a gut punch for the company, so it’ll be interesting to see if renewed access to the Chinese market is more of a “nice to have” or a “need to have” when it comes to sustaining incredible profit growth. That Huang is talking more and more about China points to the latter.
“AMD recently suggested its CQ2 would have ~47% or $70 million of the $1.5 billion total calendar year 2025 China restriction impact. Applying that same 47% proportion to NVDA’s $15 billion full-year China headwind implies a $7 billion FQ2 headwind to the unaffected (pre H20 ban) consensus $48 billion sales,” Bank of America analysts led by Vivek Arya wrote. “In other words, NVDA could guide FQ2 to as low as $41 billion, below recently lowered ~$46 billion consensus.”
“There is simply no offset to” the loss of H20 sales, wrote Morgan Stanley’s team, who agreed that this headwind to future sales may not be factored into estimates at present. “Blackwell demand is very strong... but they are supply constrained, and lost H20 does not result in more Blackwell supply.”
Grossed Out
As mentioned, gross margins (that is, sales less cost of goods sold, divided by total revenues) were on the soft side in Q4, which management attributed to the Blackwell ramp. CFO Colette Kress said adjusted gross margins would be back to the “mid-70s” later this year.
The unrelenting forward march of technological progress in general, and Nvidia’s product road map specifically, strongly imply that this will not be the last new product ramp for the firm, which raises the questions: are future generations going to need the same kind of expansion of manufacturing capabilities? Does being the best in AI inherently require somewhat of a recurring drag on gross margins in order to stay ahead of the pack?
“We await management’s confidence in gross margin recovery back to target mid-70s level in 2H (vs. consensus 73%/74% in FQ3/FQ4), as a sign of demand strength and Blackwell execution/rack-level product yields,” BofA’s Arya wrote.
Racks on Racks on Racks
You might remember overheating issues from early this year when it came to housing Blackwell chips in racks for use in data centers. Solving those logistical challenges and then turning those fixes into readily available products is a process that takes time.
“Because GB200 racks have been slow to get off the ground (UBSe <1k racks shipped from ODMs in CQ1:25), we believe the vast majority of Blackwell shipments in FQ1 were B200 (HGX platform, the same platform as Hopper) with B200 comprising nearly 70% of the Blackwell unit mix as customers took HGX servers/boards rather than waiting for the NVL72 racks,” a UBS Securities team led by Timothy Arcuri wrote.
Teasing out whether any potential sales softness in Q1 means the rest of the year will be stronger than anticipated, or whether hyperscalers are saying one thing and doing another (less likely), may become a key point of debate, as the near-term revenue profile could be a touch volatile in light of rack ramping obstacles.
“Our data points would suggest that in recent weeks the full year rack forecasts have started to be revised upwards by 50%+,” Morgan Stanley’s team wrote.
I’m from the Government, and I’m Here to Help
Even as it’s become harder for Nvidia to sell into China, it’s become easier to sell into the Middle East, as evidenced by its recent deal with Saudi Arabia’s Public Investment Fund to build “AI factories of the future.”
Private sector spending on AI will inevitably slow at some point (or not, I guess), and one key question is how much government investment is waiting in the wings.
“We look forward to hearing from Jensen about this new demand trajectory from the Middle East and what this could do around the future/growth of the AI Revolution,” Wedbush Securities analyst Dan Ives wrote. “We also note that with Stargate and other AI initiatives in the US this will be the start of a massive AI spending initiative in the Beltway over the coming years in more private/public partnerships to build out the US AI infrastructure over the coming years with Nvidia a key player.”
