NVDA Preview: Cutting Through the Noise, Street Estimates for Next Year Are Too Low

Gene Munster, Brian BakerAugust 22, 2025

The Grinch's rant about hating "the noise, noise, noise" will resonate with Nvidia investors next week. This quarter will have a lot of noise, much of it around how much of the China H20 catchup will be softened by Lutnick insulting China leadership, when a separate China chip based on Blackwell be available, and how the U.S. 15% China revenue share will impact margins. Cutting through the noise, the likely takeaway will be that Street growth rates for calendar 2026 are too low. The Street is currently expecting 27% top line growth next year compared to 55% this year. I expect the Street will revise the CY26 outlook to up 30-35%.

Key Takeaways

The noise. The degree of the China bounce back will add noise to the numbers. I expect that after Lutnick insulting China leadership and the U.S. revenue share, about $5B in catch-up revenue will fall into October. That upside will be dismissed by investors as one-time. In January I expect an 18% price increase on the new China Blackwell based chip which should offset the gross profit dollar hit from the U.S. revenue share.

The bar for July is higher than it looks at first pass when you factor in the impact of the China curbs. The Street is looking for similar growth rates (79%) in July as in April, despite the law of large numbers. If they meet Street expectations, the narrative should be that the company is bucking the law of large numbers.

Investors have known for a few weeks that the four key big tech companies meaningfully raised their capex outlook for next year following June earnings. All signs point to further increases in the coming quarters.

Networking is about 11% of the business (factoring in normal China demand). Y/y growth over the past year has lagged Data Center and GPU growth. Investors will be watching to see if long term the segment will add to growth.

The China Noise

For starters, the demand picture will significantly change for the October quarter as a result of the recent U.S. export license deal (the “15% revenue cut” arrangement) that Nvidia and AMD reached with the U.S. government. While most of this improvement is already factored in by investors, it’s worth outlining the levers in play.

The export license agreement means there will be a big catch-up in the October quarter given about $8B in revenue was lost in the July quarter, and much of that demand remains despite Lutnick comments.

My guess is that with two months of sales in October, China will add about $6B in revenue in the quarter. Investors will likely downplay the positive guidance related to China as a one-time catch-up. That number would be reduced by 15% given the U.S. government revenue share, so the net increase will be about $5B.

The topic of a new China Blackwell based chip and its price will be a focus on the earnings call. Going into the January quarter, I believe Nvidia will raise prices on the new China Blackwell based chip to offset the gross profit dollar impact from the revenue share. The math: maintaining gross profit dollars requires an 18% price hike, even though the government’s cut is 15%.

As for gross margins, Street estimates for next year are 71%. If Nvidia raises prices by 18% and preserves gross profit dollars, margins on the China business would drop to about 60% from 71%, assuming China H20 sales are 15% of revenue. This would pull overall gross margin to 69.3% from 71%.

The gross profit math is simple, but gross margin math is trickier because the government’s 15% cut also applies to the increased price. Margins will dip slightly, but I expect investors to focus on gross profit dollars rather than margins, given this is an external, policy driven pressure.

The July Bar

The bar for June is higher than it looks at first pass. The Street is expecting growth of 53% in July compared to 69% in April and 73% in January. Those numbers don’t tell the story given the noise from the China curbs.

In the April quarter, adjusting for the loss of H20 revenue to China, y/y growth would have been 79%. For July, adding back the estimated $8B impact in lost revenue from China (company estimate), the Street would be looking for 79% growth. Current Street consensus calls for $45.8B in revenue, but it would have been $53.8B without the curbs.

The bottom line is that the y/y growth rate expectation for July is similar to what it was in April, despite the law of large numbers, and these numbers are getting big. That means if the company meets Street expectations, the narrative should be that the company is bucking the law of large numbers. Some of that view should also be shifted to the October quarter and CY26.

For October, most analysts have added back H20 China revenue, which has yielded a growth rate expectation of 50%. The whisper number is calling for 50-60% growth, which at its midpoint would be a step down from July’s likely 79% rate. I believe guidance for October will be above the whisper number and likely around 60% y/y growth.

As for the upcoming stock reaction, it’s worth a look back at how NVDA traded the day after earnings, in combination with how the stock traded in the month leading into the report compared to the Nasdaq (recognizing NVDA has an outsized impact on Nasdaq performance).

Over the past four quarters, NVDA has rallied going into earnings. In the month prior to the report, NVDA has risen by an average of 12.9% compared to the Nasdaq up an average of 3.5%. Most of the earnings reports had a “buy on the rumor, sell on the news” feel, with the stock trading down an average of 2.8% on the day after earnings. Last quarter, shares traded up 3.2% following the print. Over the past month, NVDA is up 1.2% compared to the Nasdaq down 1.7%. All of this to say, stock action after the report historically has not been a leading indicator of future performance.

Hyperscalers Doubling Down: 101

Recent reports that Meta is reducing AI talent have stung some investors’ confidence in the AI buildout. I see it as Zuckerberg being smart about how he is investing, and the absolute dollars of investment will continue to grow over the next few years. There is no doubt he is all in on AI.

Rewinding a few weeks, the mega cap capex outlook is a read on how early we are in AI and the strength of Nvidia’s tailwind. What we learned from June mega cap tech earnings is that we are still earlier than most believe.

The June quarter message from Big Tech: investment in AI will exceed expectations for the next year, a sign we are still early in the buildout. A closer look at the comments suggests that when Amazon, Microsoft, and Alphabet report September and December results, they should all guide for a measurable increase in capex in CY26 compared to current expectations as of August 14.

The table below outlines the change to capex. Focus on the two columns on the far right, “CY26 Change” and “Y/Y Growth.” This is an easy way to see how quickly these companies are ramping spend.

A screenshot of a white table AI-generated content may be incorrect. — Source: FactSet

Meta’s capex expectations for next year will likely remain in line with comments from the June earnings call. Putting it together, if three of the four hyperscalers meaningfully increase capex next year in the September and December earnings calls, the market should view that as a sign numbers for most AI infrastructure companies, including Nvidia, will go higher. The fact that Meta’s expectations will not go higher may be viewed by some as a negative. Deepwater’s read is that investors will not be concerned with an unchanged Meta 2026 capex outlook because it is already well above the other hyperscalers.

What’s most important: Meta guided to about 47% capex growth for next year compared to Amazon, Microsoft, and Alphabet, which are now expected to be up by an average of 7%. Eventually, we estimate Amazon, Microsoft, and Alphabet will grow capex next year by an average of 25%. An increase in next year’s outlook in the quarters to come should be a positive for shares of NVDA.

Meta steps out. The June quarter revealed that Meta has significantly increased its optimism around AI compared to the other mega cap tech companies. For next year, Meta suggested capex would be 70% higher than the Street had previously expected. That compares to Microsoft, Amazon, and Alphabet all guiding next year’s capex up by an average of 13%. We believe the biggest reason is Zuckerberg’s determination to be first to artificial superintelligence. The AI industry as a whole is racing to reach artificial general intelligence, which means machines will think on par with humans. By contrast, artificial superintelligence means machines will vastly surpass human intelligence. That means the machine will give a response that humans will not understand how it got there, and largely trust that the result is better than what humans could come up with. This focus on superintelligence is something Zuckerberg went public with this spring. If he is right that we eventually reach superintelligence, then we are just scratching the surface in terms of building out AI, and demand for infrastructure will be elevated for 5 to 10 more years.

The networking wildcard

Networking revenue falls into Data Center revenue, which accounts for about 90% of total revenue. The segment’s long-term potential is underappreciated by investors for two reasons. First, it’s small today, accounting for about 11% of total revenue. Second, it’s been a lumpy business. In the January 2025 quarter, Networking was down 9% y/y, while the Data Center segment was up 94%. In the recent April quarter, Networking was up 56% y/y, while the Data Center segment was up 65%.

Looking back at the last four quarters, here is how Networking (estimate) and Data Center growth rates compare.

The takeaway is Networking is an important part of Nvidia’s overall business, accounting for just over 10% of revenue, and long term will likely grow faster than GPU given the law of large numbers. In five years, I believe this will account for about 15% of revenue, which implies Networking will grow at an average of 20% per year for the next five years compared to GPU growing at an average of 10% per year.

Disclaimer