Nvidia reports record revenue in Q4 2025, strong data center growth and innovation
From Nasdaq: 2025-02-26 21:30:11
Nvidia held its Q4 2025 earnings call on Feb 26, 2025, reporting record revenue of $39.3 billion, up 12% sequentially and 78% year on year. Data center revenue for fiscal 2025 was $115.2 billion, doubling from the prior year. Blackwell sales exceeded expectations, with $11 billion in revenue meeting strong demand.
Blackwell, NVIDIA’s latest product, is experiencing great demand for inference and is addressing the entire AI market from pretraining to post-training. The architecture is revolutionary, offering higher token throughput and lower costs compared to previous models. Companies across industries are benefiting from NVIDIA’s STAG inference platform, boosting performance and reducing costs significantly. NVIDIA has seen a 200x reduction in inference costs in the past two years, delivering the lowest TCO and highest ROI. Large CSPs represented half of data center revenue in Q4, with sales doubling year on year. Regional cloud hosting of NVIDIA GPUs is increasing due to rising demand for AI reasoning models worldwide.
The company launched a 100,000 GB200 cluster-based incidents with NVLink Switch and Quantum 2 InfiniBand. Consumer Internet revenue grew 3x year on year driven by generative AI and deep learning use cases. NVIDIA’s Grace Hopper Superchip powers Meta’s Andromeda advertising engine, boosting inference throughput by 3x.
NVIDIA’s automotive vertical revenue is expected to reach $5 billion this fiscal year. The company is partnering with Hyundai Motor Group to accelerate AV and robotics development. The COSMO World Foundation model platform aims to revolutionize robotics. NVIDIA’s data center revenue growth was strongest in the U.S., with global demand for AI infrastructure surging.
France and the EU are making significant investments in AI infrastructure. Data center sales in China remain below previous levels due to export controls. NVIDIA is complying with export regulations while continuing to serve customers. Networking revenue declined 3% sequentially, but networking attached to GPU compute systems remains robust.
NVIDIA’s gaming revenue decreased 22% sequentially, but full-year revenue increased by 9% year on year. The new GeForce RTX 50 Series GPUs deliver a 2x performance leap and feature new AI-driven rendering capabilities. The company also announced GeForce Blackwell laptop GPUs with new NVIDIA Max-Q technology to extend battery life by up to 40%.
Professional visualization business revenue of $511 million was up 5% sequentially and 10% year on year. NVIDIA technologies and generative AI are transforming design, engineering, and simulation workloads in key industries such as automotive and healthcare. NVIDIA’s RTX workstations are in high demand due to their integration in leading software platforms from ANSYS, Cadence, and Siemens. Automotive revenue hit a record $570 million, driven by a 103% increase year on year, fueled by the ramp-up in autonomous vehicles. Toyota, Aurora, and Continental will deploy driverless vehicles powered by NVIDIA’s technology.
NVIDIA announced that Toyota will build its next-gen vehicles on NVIDIA Orin, while Aurora and Continental will deploy driverless trucks using NVIDIA Drive Thor. The end-to-end autonomous vehicle platform NVIDIA Drive Hyperion passed industry safety assessments, making it the first AV platform to receive such comprehensive third-party evaluations. NVIDIA’s Blackwell architecture is expected to ramp up in Q1, with improved gross margins anticipated in the low 70s.
NVIDIA reported a record $8.1 billion return to shareholders in Q4 through share repurchases and dividends. Revenue for the first quarter of fiscal year ’26 is projected to be $43 billion, with a significant ramp of Blackwell expected. Full-year operating expenses are expected to grow in the mid-30s, with other income expenses anticipated around $400 million. The company’s tax rates are expected to be around 17%.
Jensen Huang, NVIDIA’s CEO, discussed the increasing blurring of lines between training and inference in AI models, highlighting the need for test-time compute and reinforcement learning. The future of AI models may require architectures that can handle post-training scaling, reasoning, and inference on a much larger scale, potentially hundreds of thousands or even millions of times greater than current capabilities. Designing such architectures will be crucial for the future of AI technology. NVIDIA’s Blackwell architecture offers incredible performance for training and scaling AI models. With faster speeds and higher throughput, Blackwell is versatile for various data center configurations. The company has shipped 1.5 million components for Blackwell racks, generating $11 billion in revenue last quarter. Demand remains high, with partners successfully bringing Blackwell systems online.
As NVIDIA continues to ramp up production of Blackwell systems, gross margins are expected to improve from the low 70s to the mid-70s later this year. Customizable features like liquid-cooling and multiple networking options offer opportunities to enhance gross margins further. The focus now is on expedited manufacturing to meet customer demand promptly.
NVIDIA remains confident in sustained strong demand for AI computing in data centers. The shift towards machine learning-based software and accelerated computing for generative and reasoning AI drives the need for NVIDIA’s architecture. Partner forecasts, capital investments, and innovative start-ups signal continued growth in the AI computing market, with new breakthroughs on the horizon. NVIDIA’s next-generation Blackwell Ultra set to launch in the second half of the year, following the successful recovery and ramping of the current generation Blackwell solutions. The company is working closely with partners to manage the transition between the two products seamlessly.
Jensen Huang discusses the demand dynamics for Blackwell Ultra and emphasizes the company’s general architecture, end-to-end capabilities, and performance advantages. The fast ROI and monetizable nature of AI factories make NVIDIA’s architecture a valuable target for companies looking to generate revenue.
The complexity of building an ASIC and the rapidly advancing software ecosystem are key challenges in deploying custom chips. Despite the design of a chip, successful deployment is not guaranteed. NVIDIA’s focus on general architecture, end-to-end capabilities, and performance advantages make its products attractive to a wide range of customers. Nvidia is deploying advanced processors into AI factories, boasting superior performance and software capabilities. The surge in U.S. demand raises concerns about regulations in other geographies but China remains a key market. AI is prevalent in consumer services, education, finance, and more, indicating its mainstream integration. The company sees AI as the future of computing, with vast potential to impact global GDP. Enterprise and cloud service provider spending on Nvidia solutions is growing rapidly. NVIDIA is working closely with enterprise companies to optimize workloads and infrastructure for AI, video processing, and data processing, potentially reducing total cost of ownership. The growth of enterprise AI and physical AI is expected to revolutionize industries like manufacturing and transportation, creating new opportunities for agentic and physical AI applications. These advancements could significantly impact global GDP and industrial sectors.
As NVIDIA approaches the two-year anniversary of the Hopper inflection, the company continues to see demand for older GPU architectures like Voltas and Pascals due to their programmability and versatility for data processing and machine learning tasks. By leveraging existing infrastructure and deploying new technologies like Hopper systems, NVIDIA aims to enhance data processing, curation, and training capabilities across a range of industries.
The implementation of AI models for data processing and curation, along with the training of Hopper systems, is expected to drive further advancements in AI applications and improve overall efficiency and productivity. By utilizing CUDA-compatible architectures and leveraging existing infrastructure, NVIDIA aims to optimize workloads and maximize the use of its GPU hardware across various industries. Nvidia’s Blackwell system, NVLink 72, and Ethernet mix are complex but offer opportunities to improve gross margins. Confidence in revenue trajectory in the second half of the year is strong despite uncertainties around tariff impacts. Demand for Blackwell is high, with AI evolving into reasoning models that require more computation and drive the need for AI computing. Blackwell is in full production and designed to meet the demands of reasoning AI and inference time scaling. AI is advancing rapidly, and data centers will increasingly focus on accelerated computing and AI. The future of AI includes multimodal, enterprise, sovereign, and physical AI. 1. The stock market experienced a sharp decline today, with the S&P 500 dropping by 3% and the Dow Jones falling by 400 points. Experts attribute this drop to concerns over rising inflation and interest rates, as well as uncertainty surrounding global trade tensions.
2. In other news, a new study has found that over 50% of Americans are now working from home at least part of the time. This shift in work habits has been driven by the COVID-19 pandemic, with many companies embracing remote work as a permanent option for employees.
3. The city of Los Angeles has announced a new initiative to combat homelessness, allocating $1 billion in funding towards building affordable housing and providing support services for those in need. This initiative comes as the city grapples with a growing homeless population and a shortage of affordable housing options.
4. Tesla has announced plans to build a new Gigafactory in Texas, with construction set to begin later this year. The factory will produce batteries and electric vehicles, creating thousands of new jobs in the area. This move is part of Tesla’s strategy to expand its production capacity and meet growing demand for electric vehicles.
Read more at Nasdaq: Nvidia (NVDA) Q4 2025 Earnings Call Transcript