The conference operator welcomes everyone to NVIDIA's Second Quarter Earnings Call and reminds them that the call is being recorded. The operator then gives instructions for the question-and-answer session. Simona Jankowski then introduces Jensen Huang, President and Chief Executive Officer, and Colette Kress, Executive Vice President and Chief Financial Officer. The call is being webcast live on NVIDIA's Investor Relations website and will be available for replay. The operator also reminds the audience that any forward-looking statements made during the call are subject to risks and uncertainties. Non-GAAP financial measures will also be discussed and can be found in the CFO commentary on the website.

NVIDIA had an exceptional quarter with record revenue of $13.51 billion, up 88% sequentially and 101% year-on-year. Data Center revenue nearly tripled year-on-year due to strong demand from cloud service providers and large consumer Internet companies. Networking revenue almost doubled year-on-year. The U.S. was the strongest region for data center growth, while China demand was within the historical range. Reports of increased regulations on exports to China have been addressed, with NVIDIA believing the current regulation is achieving intended results.

NVIDIA is anticipating that additional export restrictions on their Data Center GPUs may not have an immediate financial impact, but over the long term, it could result in a loss of opportunity in the Chinese market. The HGX platform is being used by cloud service providers and consumer internet companies to upgrade their data center infrastructure for AI, and enterprises are heavily investing in this technology as well. NVIDIA is partnering with Snowflake to provide enterprises with access to acceleration libraries, pre-trained models, and APIs to create customized generative AI applications.

NVIDIA NeMo platform allows enterprises to create custom LLMs for AI services such as chatbots, search, and summarization. AI Copilot and assistants are set to create new market opportunities for customers. WPP and Shutterstock are using NVIDIA Picasso to create images and 3D backgrounds with generative AI. NVIDIA has partnered with ServiceNow and Accenture to launch AI Lighthouse to simplify the development of enterprise AI capabilities, and with Hugging Face to create new and custom AI models for enterprises.

VMware and NVIDIA have announced a new enterprise offering called VMware Private AI Foundation with NVIDIA, which provides customers access to infrastructure, AI and cloud management software for customizing models and running AI applications. Additionally, NVIDIA is releasing AI enterprise-ready servers featuring the L40S GPU and BlueField-3 DPU data center infrastructure processor, as well as the GH200 Grace Hopper Superchip which combines an ARM-based Grace CPU with Hopper GPU. The second generation of the Grace Hopper Superchip with HBM3e memory will be available in Q2 2024.

NVIDIA announced the DGX GH200, a new AI supercomputer with 256 Grace Hopper Superchips connected together with NVIDIA's new switch system. The system is expected to be available by the end of the year, with Google Cloud, Meta, and Microsoft being the first to gain access. InfiniBand infrastructure was used to connect HGX GPU systems, providing more than double the performance of traditional Ethernet for AI. The BlueField-3 DPU was also announced, which is in qualification with major OEMs and CSPs. Gaming revenue was up 11% sequentially and 22% year-on-year, thanks to GeForce RTX 40 Series GPUs. End customer demand was solid and consistent with seasonality, and there is a large upgrade opportunity ahead.

NVIDIA's GPUs have seen strong growth in the back-to-school season, with laptop shipments outpacing desktops in several regions. RTX and DLSS games are expanding, with 35 new games added to DLSS support. NVIDIA has also launched the GeForce RTX 4060 and the GeForce RTX 4060 TI GPUs, as well as three new desktop workstation GPUs based on the Ada generation, the RTX 5000, 4500 and 4000. These GPUs offer up to 2x the RT core throughput and up to 2x faster AI training performance compared to the previous generation.

NVIDIA announced a partnership with MediaTek to bring new experiences to drivers and passengers inside the car, and announced new cloud APIs, including RunUSD and ChatUSD, to bring generative AI to OpenUSD workloads. Revenue was $253 million, down 15% sequentially and up 15% year-on-year. GAAP and non-GAAP gross margins expanded due to higher data center sales, and GAAP and non-GAAP operating expenses increased due to increased compensation and benefits.

The company has returned $3.4 billion to shareholders in the form of share repurchases and cash dividends, and has approved an additional $25 billion in stock repurchases. For the third quarter of fiscal 2024, total revenue is expected to be $16 billion, plus or minus 2%, with Data Center, gaming, and ProViz driving growth. Additionally, the new L40S GPU will help address the growing demand for many types of workloads. GAAP and non-GAAP gross margins are expected to be 71.5% and 72.5%, respectively, plus or minus 50 basis points. The company will attend several financial conferences in August and September.

Jensen Huang is discussing the emergence of large model inference, which is being accelerated by products such as Grace Hopper Superchip. He explains that large language models are able to understand unstructured language and compress a large amount of human knowledge into them. He also mentions the process of distillation, which is used to create smaller versions of the model.

Colette was asked to clarify how much incremental supply she expects to come online in the next year, to which she responded that it is growing every quarter. Jensen was asked for his confidence that hyperscalers can continue to carve out more of the pie for generative AI, to which he responded that the Data Center outlook of $12-13 billion implies that many servers are already AI accelerated, and that this demand is sustainable over the next one to two years.

Colette Kress and Jensen Huang discussed the growth of their supply and the transition of data centers from classical computing to accelerated computing and generative AI. They expect to continue increasing their supply over the next quarters and into the next fiscal year. The world has approximately $1 trillion worth of data centers, and this transition is expected to require $0.25 trillion of capital spend each year.

Colette Kress and Jensen Huang of Nvidia discuss the growth of the company's Data Center in the quarter, with HGX systems and Ampere architecture driving the revenue increases. DGXs are sold with accompanying software, and the L40S GPUs will add to the growth going forward. Huang adds that the H100 is 35,000 parts, 70 pounds, and nearly 1 trillion transistors.

NVIDIA has invested a lot of person years into creating a software ecosystem for their hardware platform, AI Enterprise, which is used for data processing, training, inference, deployment, and scaling out to a data center. It can be used on any of their GPUs and is used for both hyperscale and enterprise data centers.

NVIDIA AI Enterprise has 4,500 software packages, 10,000 dependencies, and has been optimized for two decades of use. Its architecture is flexible, versatile, and high-performing, allowing for data processing, training, inference, pre- and post-processing, and tokenizing of languages. This allows for the lowest cost of ownership, as it accelerates many different tasks, and is why so many software developers use the platform.

NVIDIA's platform is in high demand due to its large installed base, reach, scale, and velocity. This allows software developers to build businesses and get returns on their investments, and for customers to use the platform for internal consumption and training. NVIDIA is also working with VMware to bring generative AI to the world's enterprises, and has a broad distribution from all of the world's OEMs and ODMs. Furthermore, they are introducing a new architecture or product every six months, making it possible for the ecosystem to build their companies and businesses on top of NVIDIA.

Jensen Huang answers Atif Malik's question about the L40S, explaining that it is designed for a different type of application than the H100, such as fine-tuning pretrained models and being easy to install into hyperscale data centers. It is also designed for the world's enterprise IT systems and is being adopted by HPE, Dell, Lenovo, and other system makers.

Jensen Huang explains that the demand for accelerated computing and generative AI is increasing as companies recognize the benefits of offloading workloads from CPUs. He states that this is not being driven by a single application, but instead is the result of a new computing transition happening in data centers all over the world.

Toshiya asked Colette and Jensen questions regarding the types of customers in their Data Center business and whether there would be enough applications or use cases to generate a reasonable return on their investments. Colette revealed that CSPs contributed more than 50% of their revenue in Q2, followed by consumer Internet companies and enterprise and high performance computing. Jensen explained that general purpose computing is no longer the best way to go forward due to its energy costs, expense, and slow performance, and that there is a new way of doing it.

Accelerated computing and generative AI have enabled data centers to increase throughput while saving money and energy. Investing in these technologies is the best way to deploy resources in the data center, and companies such as CoreWeave are taking advantage of this. Enterprises also need to support the management system, operating system, security, and software-defined data center approach in order to take advantage of accelerated computing and generative AI.

Jensen Huang explains that they have been working with VMware to enable virtualization of GPUs and distributed computing capabilities, and they are offering a new SKU called VMware Private AI Foundation to enterprise customers. He also states that they are not using the attach rate of their networking solutions to prioritize the allocation of GPUs, and that InfiniBand is a popular choice for customers building large infrastructure due to its efficiency and cost savings.

Jensen Huang discussed the reception of DGX Cloud, which is designed to create a close partnership between NVIDIA and cloud service providers. He also discussed the software business, which is already helping margins. Finally, he mentioned BlueField-3, which is part of the Spectrum-X solution that enables generative AI capabilities within an Ethernet environment.

NVIDIA has created DGX Cloud to work with AI partners and improve the performance of hyperscale clouds. It also allows them to use large infrastructures for their self-driving car team, research team, generative AI team, language model team, and robotics team. It has been a success with both their CSPs and their own internal engineers, who are asking for more of it.

NVIDIA is seeing tremendous demand for their products due to two platform transitions, accelerated computing and generative AI. Data centers are transitioning to accelerated computing to achieve better performance, energy efficiency and cost. Generative AI is driving a platform shift in software and enabling new applications. NVIDIA is significantly expanding their production capacity to meet the demand.

NVIDIA has developed a new computing platform to increase supply for the rest of this year and next. It is special due to its architecture, large installed base, reach, scale, and velocity. It accelerates various AI models, has hundreds of millions of CUDA-compatible GPUs, is in clouds, enterprise data centers, etc., and is investing in accelerated computing and generative AI. It is upgrading and adding new products every six months to address the expanding universe of generative AI.

NVIDIA's Spectrum-X, consisting of an Ethernet switch, BlueField-3 Super NIC and software, helps customers achieve the best AI performance on Ethernet infrastructures. NVIDIA is partnering with leading enterprise IT companies and system partners to bring generative AI to the world's enterprises, and they are building NVIDIA Omniverse to digitalize and enable multi-trillion dollar heavy industries to automate their physical assets. This shift in computing has the potential to realize trillions of dollars of productivity gains.

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