NVIDIA's financial report shows impressive revenue: AI data center business explodes, Jen-Hsun Huang: Blackwell sells out.

NVIDIA ( released its Q3 FY2026 financial report as of 10/26/2025 this morning, 11/20, with overall revenue and data center revenue hitting all-time highs. CEO Jen-Hsun Huang stated that the global demand for AI training and inference is in an exponential surge phase, with their own Blackwell GPUs even selling out. At the same time, NVIDIA announced a series of large collaborations this quarter, including AI infrastructure projects in Europe, America, and Asia, and reiterated that AI factories are spreading globally.

Blackwell sold out completely, NVIDIA's Q3 revenue hits a new high.

NVIDIA's revenue this season reached 57 billion USD, a year-on-year increase of 62%. Among them, the data center business revenue reached 51.2 billion USD, a year-on-year increase of 66%, continuing to be the main growth driver. The earnings per share )EPS( is 1.30 USD. At the same time, NVIDIA also provided a strong outlook for the next quarter:

“Q4 estimated revenue is $65 billion )±2%(, clearly reflecting the market's continued explosive demand for AI training and inference.”

Jen-Hsun Huang specifically stated that the Blackwell GPUs are in short supply, and the cloud GPUs are also sold out. He mentioned that the world is entering a positive cycle of AI, not just in the United States, but more countries, startups, and industries are also increasing their investments simultaneously. After the earnings report was released, Nvidia's stock price rose over 5% in after-hours trading, indicating market confidence in the company.

Large clients collectively place orders, AI supercomputing infrastructure construction accelerates comprehensively.

This season, NVIDIA's strongest business momentum is definitely the “data center”. Not only has revenue soared to new highs, but Blackwell also achieved the best performance in the InferenceMAX test by the well-known semiconductor research firm SemiAnalysis, with a throughput increase of 10 times per million watts compared to the previous generation. Amidst the surge in demand, several large technology companies and research institutions have initiated a new round of construction:

OpenAI: Building the next generation AI infrastructure, with a scale of at least 10 GW.

Google Cloud, Microsoft, Oracle, xAI: Joining forces to build AI infrastructure in the United States, with a scale of hundreds of thousands of NVIDIA GPUs.

Anthropic: First comprehensive adoption of NVIDIA GPUs, initial deployment of 1 GW.

Oracle and the U.S. Department of Energy )DOE(: Building the supercomputing system Solstice, equipped with 100,000 Blackwell GPUs.

Equinox Supercomputing: An additional 10,000 Blackwell GPUs.

In addition, TSMC )TSMC( has successfully produced the first batch of Blackwell wafers at its Arizona facility, symbolizing the official mass production of Blackwell in the United States, marking another significant milestone this season.

From new GPUs to AI-RAN, technology has been fully updated.

NVIDIA has simultaneously launched several new technologies this season, covering everything from GPUs, system architecture to AI factory blueprints, including:

Rubin CPX: A GPU specialized in “ultra-large context” processing.

NVQLink: Connecting GPUs with quantum processors, already adopted by more than 12 supercomputing centers.

NVLink Fusion × Arm Neoverse: Strengthening AI server architecture.

NVIDIA Omniverse DSX: Complete Blueprint for AI Factory Design and Operations.

BlueField-4: The operating system processor for AI factories.

The image shows the NVIDIA Rubin CPX GPU, designed to handle inference tasks with ultra-long contexts of up to millions of tokens.

In the communications field, NVIDIA and Nokia launched the AI-native wireless network, which evolves from 5G to 6G called “AI-RAN”, and announced the AI-RAN technology stack across the United States, with partners including Booz Allen, Cisco, MITRE, ODC, and T-Mobile.

Global AI infrastructure investment is also heating up:

UK: Nvidia invests about $2.6 billion to build AI infrastructure.

Germany: Launched the world's first “industrial-grade AI cloud” in collaboration with Deutsche Telekom.

South Korea: Hyundai, Samsung, SK Hynix, and NAVER Cloud have procured over 250,000 NVIDIA GPUs.

The gaming and professional visualization sectors are expanding synchronously, while the self-driving and robotics businesses are accelerating forward.

In the gaming and AI PC sector, this quarter's revenue reached $4.3 billion, down 1% quarter-over-quarter, but still up 30% year-over-year. Several major games such as Borderlands 4, Battlefield 6, and ARC Raiders are equipped with DLSS 4 and NVIDIA Reflex. The 25th anniversary of GeForce has also driven updates to RTX-related tools, including RTX Remix and 3D generation tools.

Professional Visualization ) ProViz ( revenue reached 760 million USD, a year-on-year increase of 56%, and began shipping personal AI supercomputer DGX Spark.

Revenue in the automotive and robotics sectors reached $592 million, an annual increase of 32%. This quarter's highlights include:

DRIVE AGX Hyperion 10: Level 4 autonomous vehicle development platform.

Collaboration with Uber: Building a fleet of 100,000 Level 4 vehicles starting in 2027.

U.S. Reindustrialization Partners: including TSMC, Foxconn, Toyota, Agility Robotics, Amazon Robotics, etc.

PTC and Siemens introduce Omniverse Digital Twin.

Introducing IGX Thor: An industrial-grade, edge instant AI platform.

The image shows IGX Thor, specially designed for industrial, robotic, and medical environments.

) Note: DLSS 4 is a technology designed specifically for the RTX 50 series GPUs, capable of generating additional frames for each traditionally rendered image, significantly improving frame rates and helping to achieve higher frame rates in ray tracing games. RTX technology is used in graphics cards to enhance 3D design, simulation, and visual effects through AI and ray tracing technology, commonly found in gaming laptops and creator laptops. (

Blackwell's increase in volume could improve the gross margin, the Chinese market was not mentioned.

NVIDIA's gross margin for this quarter is 73.4%, slightly lower than the market expectation of 74%. The company explained that the short-term costs are higher than last year due to the successive launches of new chips and systems, putting pressure on the gross margin. However, as the Blackwell product line is ramping up production quarter by quarter, the gross margin is expected to improve.

In addition, this quarter's financial report completely omits any mention of the Chinese market, perhaps continuing the stance from the August financial forecast, meaning that NVIDIA has not included any shipments of H20 chips to China in its revenue expectations.

An increase in inventory is to prepare in advance, locking in long lead times and future structures.

In the face of external doubts regarding the “AI perpetual motion machine” and the cyclical nature of customer investments, Nvidia's financial structure has become even more robust. At the end of the last quarter, the company held $60.6 billion in cash and cash equivalents, providing ample resources to continuously drive more industries to adopt AI.

Although inventory has clearly increased this season, which is a warning signal for chip manufacturers, NVIDIA emphasizes that this inventory rise is in response to the explosive demand for Blackwell, as well as early orders to secure long-lead-time components, while also preparing for the ramp-up of products based on the next-generation architecture. This is considered strategic stocking rather than hoarding due to a decline in demand.

) Note: The delivery period refers to the overall time required from when the customer places the order to when the goods are shipped, including various stages such as material procurement, production manufacturing cycle, and logistics distribution. (

This article highlights NVIDIA's impressive revenue report: AI data center business surges, Jen-Hsun Huang: Blackwell sold out. First appeared in Chain News ABMedia.