Nvidia Earnings Review: The AI Bubble That Wasn't

– Revenue: Revenue came in at $57 billion, up 63% year-over-year and 22% quarter-over-quarter. This figure was significantly higher than the consensus estimate of $54.95 billion and the company's prior guidance of $54 billion.

– Gross Margin: GAAP gross margin was 73.4%, down 1.2 percentage points year-over-year but up 1.6 percentage points quarter-over-quarter, in line with the consensus estimate of 73.4% (prior guidance was 73.3%). Non-GAAP gross margin came in at 73.6%, down 1.4 percentage points year-over-year but up 0.9 percentage points quarter-over-quarter, matching the consensus estimate of 73.6% (prior guidance was 73.5%).

– Net Income: GAAP net income reached $31.91 billion, a 65% increase year-over-year and 21% quarter-over-quarter. This significantly beat the consensus estimate of $29.38 billion and prior guidance of $28.54 billion. Non-GAAP net income was $31.78 billion, up 59% year-over-year and 23% quarter-over-quarter, surpassing the consensus estimate of $30.74 billion and prior guidance of $30.05 billion.

– Data Center: Represented by AI computing and networking chips, Data Center revenue reached $51.215 billion, growing 66% year-over-year and 25% quarter-over-quarter. This segment accounted for 90% of total revenue. Within this, AI computing chip revenue was $43.028 billion (up 56% YoY, 27% QoQ), while networking chip revenue was $8.187 billion (up 162% YoY, 13% QoQ).

– Gaming: Represented by desktop/laptop GPUs and Switch console chips, Gaming revenue was $4.265 billion, up 30% year-over-year but down 1% quarter-over-quarter, accounting for 7% of revenue.

– Professional Visualization: Represented by workstation GPUs and AI software, revenue was $0.76 billion, up 56% year-over-year and 26% quarter-over-quarter, accounting for 1% of revenue.

– Automotive: Represented by autonomous driving chips and robotics edge AI chips, revenue grew 32% year-over-year and 1% quarter-over-quarter, accounting for 1% of revenue.

The keynote from this year's GTCDC remains fresh in memory. Calculating from FY26 Q1 (calendar Q1 2025) as per the keynote, Nvidia's Data Center business has totaled $131.4 billion over the first three quarters of this year (including Hopper architecture products). This implies a potential revenue space of $368.6 billion for the Data Center business from FY26 Q4 through FY27 Q4.

This quarter, the Data Center business saw an acceleration in year-over-year growth for the first time, with both networking and AI computing segments accelerating. Management explicitly stated that a series of recently announced large orders, including those from Anthropic, can begin delivery next year. Importantly, these are not included in the prior $500 billion revenue guidance.

Management clearly stated that Blackwell Ultra has become the primary architecture for customers across the board. However, demand for the earlier Blackwell version remains strong. While the incremental growth is driven more by the Ultra platform, this signals to the market that both new and old Blackwell models are selling simultaneously. Management also noted that Rubin remains on track for large-scale mass production in the second half of next year.

Addressing market concerns about depreciation of older cards, management noted that A100 products shipped six years ago are still operating at full load today. Furthermore, Cloud GPUs continue to sell as fast as they can be produced. AI server chips are essentially sold out, and supply continues to trail demand.

Regarding customer concentration concerns, management emphasized that growth is not driven solely by the four major cloud providers (hyperscalers). Demand also comes from AI startups, large internet companies, major enterprises building their own AI factories, and sovereign AI projects worldwide. The customer structure is becoming increasingly diversified rather than being dictated by a few hyperscalers.

Regarding China, the CFO reaffirmed that H20 revenue for this quarter was negligible. The guidance assumes no new orders from China. However, the company emphasized it will continue communicating with US and Chinese regulators to find compliant shipping opportunities.

Over the past few quarters, the market has closely monitored Nvidia's gross margin. Due to the delay and slow ramp of Blackwell products last year, as well as the drag from H20 product volumes, Nvidia's Non-GAAP gross margin fell from a high of 78.9% in Q1 2024 to 71.3% in Q1 2025. It recovered to 72.7% in Q2 2025, and the guidance for Q3 2025 was 73.5%. The market was anxious to see if it could return to the 75% level in Q4 2025. This quarter, management finally provided guidance for a Non-GAAP gross margin of 75% for Q4 2025 and emphasized that this level would be maintained next year.

– Revenue: Management expects Q4 revenue of $65 billion (excluding China Data Center revenue), significantly exceeding the consensus estimate of $62 billion. This represents a 65% year-over-year increase, driven primarily by the Blackwell volume ramp.

– GAAP Gross Margin: Expected to be 74.8%, an increase of 1.8 percentage points year-over-year and 1.4 percentage points quarter-over-quarter.

– Non-GAAP Gross Margin: Expected to be 75%, an increase of 1.5 percentage points year-over-year and 1.4 percentage points quarter-over-quarter. This marks a return to 75% after four quarters, with expectations to maintain this level next year.

– GAAP Net Income: Expected to be $35.21 billion, up 59% year-over-year and 10% quarter-over-quarter.

– Non-GAAP Net Income: Expected to be $36.73 billion, up 66% year-over-year and 16% quarter-over-quarter. This is poised to set a global record for single-quarter net profit.

In summary, this earnings report powerfully refutes the "AI bubble" theory. It also addresses market concerns regarding the depreciation of older cards and declining gross margins, providing renewed support for the AI narrative in the US stock market.

Check out moomoo's past insights on Nvidia: