Nvidia’s October guide appeared light versus the whisper, but adjusting for China shows it was actually higher.

The October guide called for $54B in revenue, plus or minus 2%. That compared to the $55B whisper expectation but was above the Street’s published estimate of $53.4B. In the near term, the whisper carried more weight, which pressured the stock.

I believe about half of Nvidia’s post-earnings sell-off was due to disappointment around the October guide. At first glance, the guide was 2% below the whisper. Adjusting for noise around China, however, the guide would have been between 2% and 7% higher than the whisper.

The $54B guidance did not factor in any contribution from China. In mid July, when Nvidia was cleared to sell the H20 in China, the Street was looking for $51.5B in sales for October. Following that announcement and hyperscaler results, estimates increased by $2B, some of which was tied to China.

On the call, management said China could contribute between $2B and $5B if licensing comes through as expected. At the high end, this would bring revenue to $59B, or 7% above the $55B whisper that already included China.

Nvidia’s revenue is increasingly concentrated among a small group of hyperscalers.

The top two customers accounted for 39% of sales in July 2025 compared to 25% a year earlier. According to Nvidia’s latest 10-Q, “For the second quarter of fiscal year 2026, sales to one direct customer, Customer A, represented 23% of total revenue; and sales to a second direct customer, Customer B, represented 16% of total revenue.” For reference, I believe “Customer A” accounting for 23% of Nvidia’s revenue is Meta.

Furthermore, I estimate the top six customers (Meta, Microsoft, Google, Amazon, Oracle, and xAI) accounted for 63% of sales versus 55% in July 2024.  Removing China revenue last year (which was likely 10% of sales), the top six represented 61% of sales. That means concentration rose from 61% to 63% year-over-year on an ex China apples-to-apples basis.

This growing concentration raises concern because in theory it becomes harder for the hyperscalers alone to sustain Nvidia’s growth. The biggest offset is sovereign AI spending, which is still emerging. In the near term, however, Nvidia’s trajectory will be driven by the same six hyperscalers, four of which will update capex guidance when they report September earnings.

Meta’s aggressive capex contrasts with cautious peers, setting up a critical September read.

Following June earnings, Meta guided for CY26 capex to increase 47% from CY25. By comparison, Microsoft, Google, and Amazon (which account for about 32% of Nvidia’s overall revenue) guided for an average increase of 7%. Given Meta is unlikely to raise further, all eyes are on whether the other three hyperscalers revise higher. Below are four potential outcomes of increased Capex from Microsoft, Google, and Amazon following the September quarter, and what it means for overall revenue growth. This excludes China and assumes the rest of the business grows at 35% next year versus 45% growth in CY25.

In the scenarios below, the baseline is set by the Street’s CY26 revenue growth estimate of 31%. To get there we have to assume that all other Nvidia customers, excluding the hyperscalers, will increase their capex by 39%.

The bottom line, by looking at capex from Microsoft, Google and Amazon, we get a sense that Nvidia’s 2026 growth will likely fall in a 30–40% range, with best-case upside to 43%. Consensus is at 31%, which based on this exercise looks slightly conservative, though a blowout year above 40% appears unlikely. Even if Microsoft, Google, and Amazon all matched Meta’s 47% growth in capex, Nvidia’s overall growth would reach 43%, which I view as a best-case outcome.

While some may see this as limiting 2026 upside, I believe by then investor attention will have shifted to CY27. I expect Nvidia’s growth in CY27 will be in the 15–20% range, which is higher than Street expectations for low single-digit growth. This suggests growth durability is under-appreciated once the market looks beyond 2026.

Disclaimer