11/20 Nvidia earnings results and conference call
(Future outlook, AI virtuous cycle, depreciation, mentioned companies, Q&A)
Nvidia earnings results
EPS: Estimated at $1.25, above $1.30
Revenue: Estimated $55.1 billion, above $57.01 billion
Data Center Revenue: Market Forecast of $49.3 Billion, Earnings of $51.2 Billion
Revenue Outlook: Estimated $62 billion, up from $65 billion +/- 2%
Margin Rate Outlook: Expected 74.6% and above 75.0%
a presentation
Blackwell led the sales, and there was little cost related to H20, a chip for export to China. In particular, Jensen Huang said, “Blackwell chip sales are on the verge of breaking through the charts, and cloud GPUs are completely sold out.” Noteworthy is the mention of the AI virtuous cycle structure, which is said to have entered a stage where AI learning and demand for reasoning accelerate each other and grow exponentially. This is cheered by the market in that it talks about continued growth related to the AI bubble.
Conference call contents
It mentions that $500 billion in sales will be made in Blackwell and Rubin by the end of 2026. Furthermore, it claims that it will account for more than half of the AI infrastructure construction market, which is expected to be worth between $3 and $4 trillion.
The virtuous cycle structure and agent AI in the AI industry
Nvidia emphasized that the current AI industry has gone beyond a simple boom and entered a logical virtuous cycle structure. The Scaling Law, which improves the intelligence of AI models as the computing power increases in the pre-learning, post-learning, and inference stages, is still in effect. In particular, with the creation of smarter AI models with more computing investments, this creates new apps and services, and again forms a structure that leads to computing reinvestment. Furthermore, it was announced that AI, an agent that does its own work beyond just chatbots, has appeared. For example, Salesforce’s agent increased work productivity by 30%, and Cursor, a coding tool, drastically reduced the speed of development. It claims that this is a solid ROI basis for companies to open their wallets to AI.
Status of major customers
Nvidia’s customer base is diversifying beyond cloud big tech to national, enterprise, and robotics
Hyperscaler: Major cloud players such as Amazon, MS, Google, and Oracle are aggressively expanding their infrastructure based on Nvidia GPUs.
Foundation Model Developer: Model performance of OpenAI, Ant-Phobic, xAI, Mistral, Google DeepMine, etc. Advanced
Enterprise and software: Palantir is the first to apply Nvidia CUDA to its AI platform. ServiceNow, Salesforce, SAP integrate AI agents into business software. Foxconn, Lenova are AI factory building partners.
Robotics and Physical AI: Toyota, TSMC, and Amazon Robotics introduced factory automation and digital twins.
Tesla uses large GPU clusters to learn autonomous driving
Depreciation, Total Cost of Ownership (TCO) and Equipment Lifetime
Responding to the fast aging (depreciation burden) of existing chips, which is a concern for investors, Jensen Huang uses CUDA software and TCO reduction logic. Notably, Nvidia’s GPUs are not just hardware but work on the software ecosystem called CUDA. Jensen Huang emphasized that the A100 GPU, which was launched six years ago, is still being utilized 100% by the optimization of the latest library. In other words, it claims that the economic useful life of software support is much longer than physical depreciation. In addition, Blackwell is expensive but several times to several tens of times superior to its predecessor, which rather reduces costs in terms of cost per performance or per electricity.
the outlook for the next quarter
Nvidia is expected to maintain solid profitability at around 75% (±0.5%) with an increase of about 14% from the previous quarter to $65 billion (±2%) on the back of Blackwell’s strong growth momentum. CEO Jensen Huang emphasized that the current market situation is not an AI bubble but a fundamental change in industrial structure, arguing that three huge platform transitions are underway at the same time: “Transition to Accelerated Computing (CPU→GPU) under Moore’s Law limits, “evolution to Generative AI” that increases the company’s actual revenue (ROI), and “emergence of agents and physical AI such as autonomous driving and robotics.”
Key Questions Answering
Question: Can you update me on Blackwell and Rubin’s market opportunities worth $500 billion from 2025 to 2026, which I mentioned in GTC in the past?
Answer (Jensen Huang):
By the end of 2026, the roadmap for Blackwell and Rubin will be carried out without a hitch, based on a $500 billion projection. Data center sales have surpassed $50 billion this quarter, but this is only the beginning. We are confident that we will be able to meet the additional computing demand that extends into fiscal 2026 and continue our growth.
Question: AI investment seems reasonable in terms of ROI at the moment, but the industry as a whole doesn’t seem to have fully felt the effect yet. Next-generation models will continue to come out, so do you expect supply to catch up with demand in the next 12 to 18 months
Answer (Jensen Huang):
Supply chain management is expanding its production capacity by working closely with supply chains around the world such as TSMC, memory companies, and packaging partners. Currently, the market is not just a boom, but three huge changes are taking place at the same time: ① the transition from general-purpose computing to accelerated computing, ② the introduction of generative AI, and ③ the rise of agent AI. Demand is diversifying, and the search and recommendation systems that CPU used to do are now moving to GPU-based Generative AI Through this, new applications such as coding and medical AI are exploding.
Question: When calculating the $500 billion market size, what do you assume is Nvidia’s revenue per data center power capacity? Also, do you think there’s enough customer base to raise money when you forecast a $3-4 trillion market by 2030?
Answer (Jensen Huang):
The performance of Blackwell and Rubin is improving dramatically. Blackwell performs significantly better than Hopper, which dramatically lowers the total cost of ownership (TCO) for customers. The market is hyperscaler (Big)