The conference operator introduces the call and the speakers, Jen-Hsun Huang and Colette Kress. The call will be available for replay and the content is the property of NVIDIA. Forward-looking statements may be made during the call and non-GAAP financial measures will be discussed. The call is scheduled for February 21, 2024.

In the fourth quarter, data center revenue for NVIDIA was a record $18.4 billion, driven by the Hopper GPU computing platform and InfiniBand networking. Demand for Hopper remains strong and supply is improving, but the next-generation products may still be supply constrained. The growth was driven by both training and inference of generative AI and large language models across various industries and use cases. The versatility and performance of NVIDIA's data center platform allows for a high return on investment for many use cases, including AI training and inference, data processing, and CUDA accelerated workloads. Approximately 40% of data center revenue in the past year was for AI inference.

The use of AI solutions has become widespread across industries, with many companies deploying them through cloud providers or on-premise. Large cloud providers make up a significant portion of NVIDIA's data center revenue, and Microsoft's GitHub Copilot has seen rapid adoption. Consumer internet companies are also utilizing AI for recommendation systems, leading to improved engagement and ad performance. Additionally, generative AI is being used by both consumer internet and enterprise software companies to automate tasks and increase productivity.

Early customers partnering with NVIDIA for generative AI training and inference are seeing commercial success, such as ServiceNow's record-breaking net new annual contract value. Other leading AI and enterprise software platforms, including Adobe, Databricks, Getty Images, SAP, and Snowflake, are also utilizing NVIDIA's technology. The field of large-language models is thriving, with companies like Anthropic, Google, Inflection, Microsoft, OpenAI, and xAI making breakthroughs. Startups are also creating LLMs for specific languages, cultures, and industries. NVIDIA's AI infrastructure is in demand, with collaborations announced with Google to optimize language models on NVIDIA GPUs. There has been significant adoption of AI by enterprises in various industries, with NVIDIA offering application frameworks for vertical domains. The automotive vertical alone contributed over $1 billion in data center revenue last year.

NVIDIA's DRIVE infrastructure solutions are being used by almost 80 vehicle manufacturers and various companies in the automotive industry for the development of autonomous driving and AI applications. In healthcare, NVIDIA's platforms and services are being used for drug discovery and medical imaging. In financial services, AI is being used for various purposes such as fraud detection. However, the data center revenue in China has declined due to export control regulations imposed by the U.S. government.

NVIDIA has begun shipping alternative products to China due to government restrictions and expects China to make up a similar percentage of their data center revenue in the first quarter. Sovereign AI has also become a demand driver in regions outside of the U.S. and China. The majority of revenue was driven by the Hopper architecture and InfiniBand networking, which have become the standard for accelerated computing and AI infrastructure. The company is on track to release the H200 with improved performance in the second quarter. Their Quantum InfiniBand solutions have grown significantly and they are now entering the ethernet networking space with the launch of Spectrum-X, which offers higher networking performance for AI processing. Leading OEMs are partnering with NVIDIA to expand their AI solutions globally and Spectrum-X is set to be shipped this quarter.

NVIDIA has made significant progress in their software and services offerings, reaching a revenue of $1 billion in Q4. They have expanded their list of partners for NVIDIA DGX Cloud and announced the GeForce RTX 40 Super Series GPUs at CES. These GPUs offer high gaming performance and AI capabilities, with the ability to run AI up to 5x faster on RTX AI PCs. They also announced new RTX 40 Series AI laptops and the Avatar Cloud Engine microservices for developers to integrate generative AI into digital avatars.

At CES, ACE received several Best of CES 2024 awards and NVIDIA announced their end-to-end platform for generative AI applications on RTX PCs and workstations. The company saw a 11% sequential growth and 105% year-on-year growth in revenue for Pro Visualization, driven by the adoption of RTX Ada architecture GPUs and demand from industries such as manufacturing, automotive, and robotics. The automotive industry is also adopting NVIDIA Omniverse for digitizing workflows and enhancing the car buying experience. In Automotive, revenue was $281 million in the quarter and $1.09 billion for the fiscal year, with NVIDIA DRIVE Orin being the preferred AI car computer for software-defined AV fleets.

NVIDIA's successor to its DRIVE platform, NVIDIA DRIVE Thor, will offer more AI performance and a range of intelligent capabilities for autonomous driving. Several automotive customers have announced new vehicles built on NVIDIA. In Q4, GAAP gross margins expanded to 76% and non-GAAP gross margins to 76.7% due to strong data center growth. The company returned $2.8 billion to shareholders in Q4 and $9.9 billion in fiscal year '24. The outlook for Q1 includes expected revenue of $24 billion, with growth in data center and proviz offset by a seasonal decline in gaming. Gross margins are expected to be in the mid-70s percent range for the remainder of the year, and operating expenses are expected to grow in the mid-30% range for fiscal year 2025 as the company continues to invest in future opportunities.

The company expects GAAP and non-GAAP other income and expenses to be approximately $250 million, with a tax rate of 17%. Upcoming events for the financial community include attending conferences and hosting an annual event. The first question from a conference call is about the company's data center business and how their expectations for 2024 and 2025 have evolved. The caller also asks about newer areas of the data center business, such as software and sovereign AI, and whether the company will participate in the ASIC market.

The speaker discusses the evolution of expectations for data centers and states that conditions are favorable for continued growth in the coming years. This is due to two industry-wide transitions: a shift from general to accelerated computing, and the emergence of generative AI. These transitions offer improved energy efficiency, cost savings, and speed, and are enabling a new industry focused on generative AI. The data center is no longer solely for computing and storing data, but is now also serving a larger purpose in the development of AI.

NVIDIA has created a new type of data center called an AI generation factory which uses data and AI supercomputers to produce valuable tokens. These tokens are used in various applications such as ChatGPT, Midjourney, and digital biology. The company's business has seen growth in both inference and training, as well as diversification into new industries, including specialized CSPs and enterprise software platforms.

The speaker discusses the rise of generative AI and its impact on various industries, including consumer internet services, industrial AI, and sovereign AI. They mention the increasing importance of inference, which now accounts for 40% of their company's revenues. The speaker predicts that generative AI will continue to grow and become a major industry in the future.

Jensen Huang discusses the growth of LLMs from inference and how it is being measured. He explains that the estimate of this growth is likely understated due to the recent migration of recommender systems to deep learning and generative AI, which require GPU acceleration. This has led to GPUs being used in every step of a recommender system, making them a crucial component in many major companies. Huang also mentions the various applications of generative models, such as ChatGPT, Midjourney, Getty, and Firefly, which did not exist a year ago.

Jensen Huang, the CEO of NVIDIA, is discussing the company's supply chain and the challenges they face with their new products. They expect the next generation of products, specifically the Blackwell, to be supply constrained due to the complexity and large number of components. However, overall, the supply chain is improving and they are doing their best to meet the strong demand for their products. The cycle times are also improving, but it takes time to ramp up production for new products. Currently, they are ramping up production for their H200 products.

The speaker discusses the current demand for their new product, Spectrum-X, and how it is greater than their supply. They have optimized ethernet for AI and InfiniBand will be their AI-dedicated infrastructure. The company is working to capture the demand, but overall supply is increasing. The question is then posed to Colette and Jensen about how they are allocating product to customers, considering factors such as customer readiness and competition among industries.

Colette Kress and Jensen Huang discuss how they work with their customers to fairly allocate their unique GPU technology. They have transparent relationships with their customers and communicate about product roadmaps and transitions, which helps with allocation. They strive to allocate fairly and avoid unnecessary allocations.

The speaker mentions the importance of allocating resources fairly and avoiding waste in the data center. They also highlight the strong ecosystem of partners and end markets that NVIDIA has, and how they are constantly looking for opportunities to connect these partners and end users. The speaker also addresses a question about converting backlog into revenue and mentions that lead times for their products have decreased. They did not discuss inventory purchase commitments, but the aggregate of their supply was slightly down.

Colette Kress, the CFO of Nvidia, discusses the company's inventory and purchase commitments, as well as their prepaids. She clarifies that these commitments have different lengths and are not indicative of a decrease in financial commitment to suppliers. She also mentions that the company's gross margins should return to the mid-70s, potentially due to the inclusion of HBM content in new products.

The speaker discusses the unique gross margin in Q4 and Q1, driven by favorable component costs and manufacturing processes. They expect a mid-70s gross margin for the rest of the fiscal year. The questioner asks about the long-term usability of NVIDIA's investments, and the speaker explains that their accelerated and programmable platform has allowed for significant performance improvements, making it the only architecture to have evolved to support various versions of deep learning.

NVIDIA has been able to support and optimize every version and species of AI that has emerged, such as vision transformers and multi-modality transformers. They have also been able to simultaneously invent new architectures and technologies, like their Tensor cores and transformer engine, while still supporting their installed base. This allows them to bring software to the installed base and make it better over time, while also creating revolutionary capabilities for future generations. NVIDIA also plans to make breakthroughs in large-language models available to their installed base. This approach has also been successful in their China business.

The speaker discusses the impact of US government restrictions on NVIDIA's business in China and their efforts to comply with these restrictions while still competing in the market. They mention a decline in business due to pausing shipments, but hope to resume competing after reconfiguring their products. A question is asked about the extent of products currently being shipped and the possibility of expanding into other alternative solutions in the future.

The speaker discusses the success of NVIDIA's software business, which has reached over $1 billion. They explain that the company's success in software is due to their close collaboration with cloud service providers and their extensive engineering teams. They also mention the importance of software in enabling new markets and applications for accelerated computing, and highlight the difference between accelerated computing and general-purpose computing in terms of software needs.

NVIDIA is working closely with CSPs and using generative AI to enable enterprises to embrace accelerated computing. This is necessary because general-purpose computing is no longer sufficient to sustain improved throughput. NVIDIA is offering NVIDIA AI Enterprise, an operating system for AI, for $4,500 per GPU per year. This is expected to be a significant business for the company and has already reached a $1 billion run rate. The computer industry is currently undergoing two platform shifts, with data centers transitioning from general purpose to accelerated computing.

NVIDIA is accelerating the development of data centers to meet the increasing demand for computing power and manage costs and energy. This is enabling a new computing paradigm, generative AI, which has the potential to revolutionize industries and drive a doubling of the world's data center infrastructure in the next five years. NVIDIA's full stack computing platform and partnerships with various industries position them to help companies become AI companies. They will share more updates at next month's GTC conference.

The paragraph states that the speaker has finished their message and the listener is now free to disconnect from the conversation.

This summary was generated with AI and may contain some inaccuracies.