From Nasdaq: 2024-11-20 21:00:11

Nvidia held its Q3 2025 earnings call, reporting record revenue of $35.1 billion, up 94% year over year. Data center revenue hit $30.8 billion, with exceptional demand for the H200 model. Cloud service providers deployed H200 infrastructure, driving significant growth in AI training and inference workloads.

Nvidia continues to dominate the inference platform market, with rapid advancements in software algorithms boosting Hopper inference performance by 5x in one year. The upcoming release of NVIDIA NIM is expected to further increase Hopper inference performance by 2.4x. Customers are eager to deploy Blackwell, a full-stack AI data center system, with Oracle announcing Zettascale AI cloud computing clusters. Microsoft takes the lead by offering Blackwell-based cloud instances with NVIDIA GB200 and Quantum InfiniBand. Blackwell outperformed Hopper in recent MLPerf training results, showing a 2.2x performance increase. This advancement aims to reduce the cost of compute, with 64 Blackwell GPUs needed for GPT-3 benchmark instead of 256 H100s.

AI-native companies like Anthropic, Perplexity, Mistral, and more are thriving in the market. NVIDIA AI Enterprise is driving industry leaders to develop Copilots and agents. Consulting giants like Accenture have embraced NVIDIA AI, with projected full-year revenue growth. The industrial AI sector is rapidly evolving, leveraging solutions like NVIDIA Omniverse for enhanced efficiency.

In India, CSPs are ramping up NVIDIA GPU deployments, while SoftBank in Japan is building a powerful AI supercomputer. Networking revenue is on the rise, with InfiniBand and Ethernet switches experiencing sequential growth. NVIDIA’s Spectrum-X Ethernet for AI saw a significant revenue increase, attracting interest from CSPs and consumer Internet companies.

Gaming revenue surged to $3.3 billion, driven by strong demand for GeForce RTX GPUs. New AI PC models from ASUS and MSI offer enhanced AI capabilities for gaming, photo editing, and more. GeForce 256’s 25th anniversary marks its legacy in transforming graphics and fueling the AI revolution. NVIDIA’s GPUs continue to power technologies like ProViz and Automotive, with revenues up in both sectors. GAAP and non-GAAP gross margins declined due to shift in data center systems. Fourth-quarter outlook includes expected revenue of $37.5 billion, supply constraints impacting Gaming revenue, and focus on Blackwell AI infrastructure ramp-up.

CEO Jensen Huang discusses scaling for large language models, highlighting pretraining, post-training, and inference time scaling. Demand for NVIDIA’s infrastructure is high, with new Blackwell products set to meet industry needs. The company is also seeing growth in inference scaling due to its large installed base. NVIDIA is experiencing a rise in demand for inference as well as increased scaling of inference time. The number of AI native companies continues to grow, and enterprise adoption of agentic AI is on the rise. Blackwell production is in full steam, with more systems being delivered this quarter than previously estimated.

The supply chain is working to increase Blackwell production, with seven custom chips being built for the systems. The integration of various components into data centers is complex but necessary for the scale of Blackwell ramping up. NVIDIA is working with numerous partners to ramp up production, with the goal of executing on their annual roadmap.

Increasing performance while reducing costs is a key focus for NVIDIA, as it makes AI more accessible. Power constraints in data centers mean that high performance per watt is crucial for maximizing revenues for partners. NVIDIA’s Blackwell systems are being integrated into data centers worldwide, with numerous companies involved in the ramp-up process. NVIDIA’s annual road map aims to cut costs and boost revenues for customers. Blackwell is expected to surpass Hopper in shipments, with gross margins starting in the low 70s and increasing to the mid-70s. The company anticipates a shift from CPU coding to GPU-driven machine learning, leading to the rise of generative AI and AI factories. Recovery to mid-70s gross margin may be possible by the second half of next year, depending on ramp mix. No digestion phase is expected until $1 trillion worth of data centers are modernized. Businesses are encouraged to build data centers for the future of machine learning and generative AI, as the world’s data centers are set to modernize at a cost of a couple trillion dollars by 2030. New AI-native companies are emerging, creating a new market segment with unique capabilities and opportunities.

In the next quarter, total data center revenues are expected to increase by several billion dollars. Blackwell is ramping up strongly, while Hopper remains a fast-growing product. China’s market is also showing strength. The company aims to improve gross margins to the mid-70s range through yield and product enhancements.

As the demand for inference grows, AI companies hope that businesses will integrate it into various departments for marketing, forecasting, and legal purposes. The goal is for every company to engage in inference activities 24/7, prompting the emergence of thousands of AI-native startups focused on generating tokens and AI. In the world of AI, the goal is to train models so people use them effectively. Physical AI, a new genre, understands human language and the physical world. This capability is valuable for industrial AI and robotics, sparking innovation in AI-native and robotics companies. Omniverse was created to enable physical AI through synthetic data generation and reinforcement learning.

NVIDIA’s ecosystem allows for rapid innovation on top of their architecture, ensuring compatibility and ease of deployment. The growth of large models with long context lengths requires innovative solutions. NVIDIA’s ability to innovate in every direction while having a large installed base ensures that creations can be deployed worldwide, from data centers to robotic systems.

NVIDIA’s focus on networking growth has been strong, especially since acquiring Mellanox. The slight dip in the networking business is temporary, with a strong outlook for future growth. NVIDIA’s networking is a critical part of their data center solutions, and more systems are expected to incorporate their networking technology, especially with Blackwell on the horizon.

Sovereign AI is a key driver of growth, with models being built in various countries. The pipeline for sovereign AI is strong, with models being developed in local languages and cultures. Regarding gaming supply constraints, NVIDIA is shifting focus towards data centers, potentially impacting gaming supply availability. NVIDIA’s regional cloud growth opportunities are expanding, particularly in Europe and Asia Pac. Gaming supply is in high demand, with products selling quickly. Challenges include ramping up supply for the current quarter, but more is expected in the new year. Sequential growth is strong, with a 7% increase projected. Post-training technologies are advancing, with a focus on pretraining, reinforcement learning, and inference. The AI revolution is driving NVIDIA’s growth, with a shift towards machine learning and neural networks. The company anticipates significant demand for new products like Blackwell. Foundation model makers have increased in number over the past year, with a growing scale of pretraining and post-training in computing. AI-native start-ups are on the rise, along with successful inference services. The introduction of ChatGPT o1 and OpenAI o1 has led to the emergence of a new scaling law known as test time scaling. AI is transforming industries, companies, and countries, with enterprises adopting agentic AI to revolutionize workflows. Investments in industrial robotics are increasing due to breakthroughs in physical AI, driving demand for new training infrastructure. The age of robotics is approaching, and countries worldwide are recognizing the importance of developing national AI infrastructure. NVIDIA is well-positioned to serve the multitrillion-dollar AI and robotics opportunities ahead. 1. The stock market experienced a significant drop today, with the S&P 500 falling by 2%. This was due to concerns over rising inflation rates and the potential impact on interest rates. The Dow Jones Industrial Average also saw a decrease of 400 points, marking a challenging day for investors.

2. In international news, tensions between Russia and Ukraine have escalated, leading to fears of a potential military conflict. The United States has expressed support for Ukraine, urging Russia to de-escalate the situation. This comes amidst ongoing geopolitical tensions in the region.

3. The latest unemployment report shows a decrease in jobless claims, with 260,000 fewer claims filed last week. This marks a positive trend in the labor market, as businesses continue to hire and the economy shows signs of recovery. Analysts are optimistic about the future of job growth in the coming months.



Read more at Nasdaq: Nvidia (NVDA) Q3 2025 Earnings Call Transcript