Nvidia Q3 revenue hits $57b, forecasts higher Q4

Nvidia reported record revenue of US$57 billion for Q3 FY26, up 22% from Q2 and 62% year-on-year, ending October 26, 2025.

Data center revenue reached US$51.2 billion, a 25% rise from the previous quarter and 66% up from a year earlier.

Quarterly GAAP gross margin was 73.4%, and diluted earnings per share were US$1.3.

Nvidia returned US$37 billion to shareholders in the first nine months of FY26 through share buybacks and dividends, with US$62.2 billion remaining for future repurchases.

The company expects Q4 revenue to reach about US$65 billion, plus or minus 2%.

Gaming revenue for the quarter was US$4.3 billion, down 1% from Q2 but up 30% year-on-year, while professional visualization revenue rose 26% quarter-on-quarter to US$760 million.

Automotive revenue reached US$592 million, up 1% from Q2 and 32% year-on-year.

🔗 Source: Nvidia

🧠 Food for thought

Implications, context, and why it matters.

Blackwell demand outpaces Nvidia’s supply

Revenue rose 22% to $57 billion, yet supply is tight. Blackwell systems (the company’s latest AI accelerator platform that succeeds the prior generation Hopper architecture) sell out faster than Nvidia can make them, and management called demand ‘insane’ with shortages likely for several quarters 1.
The move from Hopper to Blackwell nudged gross margin down from 74.6% to 73.4% 2. Multi-year cloud service agreements more than doubled to $26 billion, and inventory reached $19.8 billion to lock in long lead-time parts 2. Booked demand still outruns what Nvidia can ship, which is the risk to watch in coming quarters.

800G optics makers have a short window to win AI deployments

Network upgrades are pulling forward with the move from 400G to 800G modules (optical transceivers moving 400 or 800 gigabits per second), with the 800G transceiver market rising from $1.2 billion in 2024 to $6.5 billion by 2033 3 4.
Pluggable optics (swappable optical modules used in switches and servers) demand in North America may nearly double from $2.09 billion in 2023 to $4.99 billion by 2031 5. AWS, Microsoft Azure and Google Cloud drive much of this spend, which also feeds Nvidia’s growth 5 1. Vendors should focus on QSFP-DD (Quad Small Form-Factor Pluggable Double Density) and OSFP (Octal Small Form-Factor Pluggable) while tracking co-packaged optics (CPO), which can cut power from about 15 pJ/bit to about 5 pJ/bit (picojoules per bit, a measure of energy per data moved) 4 3. Those gains could decide AI infrastructure awards over the next 18 to 24 months.