A：Nvidia's first quarter fiscal 2027 financial results include record revenue, operating income, and free cash flow. Total revenue reached $82 billion, an increase of 85% year over year and 20% sequentially. Data center revenue grew to $75 billion, up 92% year over year and 21% sequentially, driven by the strength of the Blackwell architecture and demand for GPUs like GB 300 and BL 72. Notable achievements include a record sequential revenue increase of $13.5 billion, the fastest product ramp in company history, and the introduction of new reporting frameworks to reflect growth drivers.

A：Nvidia's new reporting framework consists of two market platforms: data center and edge computing. Within the data center platform, submarkets are hyperscale, AI/ML clouds, industrial, and enterprise. The hyperscale submarket includes revenue from public cloud and major consumer internet companies, while the AI/ML clouds submarket focuses on AI-specific data centers across industries and countries. The edge computing platform encompasses devices for physical AI, such as PCs, gaming consoles, workstations, and automotive applications.

A：The demand for AI infrastructure is expanding at an unprecedented pace, with a build-out of AI factories accelerating. The value of Nvidia AI infrastructure is rising, as evidenced by year-over-year increases of 20% in the price of renting H-100 and nearly 15% for a 100 cloud. The two primary drivers for this expansion are the transition of the largest hyperscale workloads from CPU to GPU-accelerated computing and the adoption of AI-native products and services. The shift is driven by the necessity of AI for enhancing productivity across all industries and roles.

A：Nvidia's Blackwell architecture is significant as it is adopted and deployed by every major hyperscaler, cloud provider, and every major model maker. The architecture is at the forefront of AI advancements, with the GPT 5.5 model being trained and served on Blackwell, positioned atop artificial analysis leaderboards. Notable collaborations include Microsoft's Fairwater data center, powered by Blackwell GPUs, and partnerships with companies like AWS, Google, Anthropic, OpenAI, SpaceX, Meta, and Microsoft to expand AI compute capacity and support their growth.

A：Nvidia's AI factory solutions deliver the industry's lowest token cost, highest token throughput, and the highest ROI. They also provide the most economic metric for AI factory operators, encompassing lifetime cost per intelligence token produced, efficiency metrics like tokens per watt or per dollar, and factors such as uptime, utilization, time to production, software, durability, and asset life.

A：Vera, built on custom Arm cores and designed with Nvidia GPUs and NV link, is set to deliver up to 1.5x faster performance per core, 2x performance per watt, and 4x density per rack compared to x 86 based alternatives. It is targeting a new $200 billion market segment for Nvidia, with partnerships from major hyperscale and system makers. Nvidia expects to generate visibility to nearly $20 billion in total CPU revenue for the year, potentially making it the world's leading CPU supplier.

A：Nvidia's edge computing market platform generated $6.4 billion in revenue, up 10% from the previous quarter and 29% year over year, with robust demand for Blackwell workstations contributing to the growth. Consumer demand fell modestly due to higher costs. Notable partnerships and growth areas include a deal with Uber to power a robo-taxi fleet and the use of Nvidia technology by leading companies across various industrial, surgical, and humanoid applications.

A：Nvidia is proactively securing supply to support customer growth by increasing total supply and inventory commitments. Their capital allocation strategy includes prioritizing R&D and strategic investments to drive ecosystem growth and maintain market position. This includes investments in AI technology to lower costs and increase throughput, as well as returning value to shareholders through an increased quarterly dividend and an 80 billion share repurchase authorization.

A：For the second quarter, Nvidia expects total revenue to be $91 billion, plus or -2%, with sequential growth primarily driven by the data center segment. GAAP and non-GAAP gross margins are expected to be 74.9% and 75%, respectively, with full-year expectations of being in the mid-70s for both. Full-year GAAP and non-GAAP operating expenses are forecasted to be approximately $8.5 billion and $8.3 billion, respectively, and Opex growth is expected to grow in the upper 40s year over year. The company also projects a GAAP and non-GAAP effective tax rate between 16 and 18% for the full year 2027.

A：AI is diverse, covering various fields such as languages, 3D graphics for manufacturing and industrial robotics, proteins for life sciences, small chemicals or life sciences, material sciences, physics for the physical sciences, energy sector, science labs, higher education, and others. The applications of AI are also diverse, including enterprise, the energy sector, manufacturing sector, hyper-scaled cloud, AI natives, on-premises solutions, industrial applications in factories and plants, supercomputing centers, edge computing including self-driving cars and robotics, and future AI-powered radio networks.

A：Nvidia is unique because it is the only company that builds all of the technology components for AI in an extreme co-design, end-to-end, and full-stack manner. They then open the platform for integration into various environments, catering to different market segments where an integrated, full-stack solution is required.

A：Nvidia's business is segmented into three large segments: hyperscale clouds, AI natives/enterprise on-premises and industrial on-premises, and the robotic edge. Within the hyperscale segment, there are three different ways of operation - accelerating data processing and AI processing inside public clouds, providing an end-to-end solution for AI integration, and supporting the growth of AI adoption across all industries and countries.

A：Nvidia's data center business excluding China grew about 120% in the quarter, and the company is guiding for about $100 billion in hyperscale CapEx growth this year. The goal is to grow faster than hyperscaler CapEx, which is forecasted to grow 90% to 100% and reach $3 to $4 trillion by the end of the decade. Nvidia aims to grow faster by focusing on its two large segments: hyperscale data centers and other AI applications such as AI natives, enterprise on-premises, and the robotic edge.

A：The second category consists of AI native clouds that are regional and global, supporting startups and 250,000 enterprise companies worldwide. Many industrial companies need to build AI factories on-site to operate because the data cannot be moved to the cloud, necessitating local, reliable, and quick responses. This category also includes data centers that want to buy systems and operate them without designing or building them themselves, which is a rapidly growing segment.

A：The second category of AI native clouds is extremely diverse, with hundreds and potentially thousands of companies, as opposed to the first category which is represented by a handful of major companies. The second category includes many small companies with smaller installations, and this category is projected to continue growing at an incredible pace.

A：The speaker's company has a unique platform that is vertically integrated to ensure everything works seamlessly but can also be disassembled for customers to customize and assemble according to their needs. This allows the company to effectively serve the physical AI industry, which is anticipated to grow rapidly.

A：Partnerships with companies like Anthropic are expected to significantly boost the speaker's company's share of the inference market. The addition of Anthropic as a partner and the substantial computing capacity being brought online for them will lead to a considerable growth in the company's inference market share.

A：Vera Rubin is anticipated to be even more successful than Grace Blackwell, with the potential for rapid uptake across all frontier model companies, which wasn't the case with Grace Blackwell.

A：The Lpx is designed for low latency, high token rate applications, but its model size capacity is limited and its context processing ability is not as robust as for software coding or agentic workloads. It is a niche product suited for customers with a large portfolio of different types of token services and high token rates. It is not expected to have broad use cases but can complement services that already exist, potentially contributing to a small percentage of the market.

A：Grace Blackwell supports the entire life cycle of AI, including data processing for training and preparation, post-training reinforcement learning, and inference. It is considered the best platform for these purposes and can potentially enhance services when combined with low latency, high token rate services like Lpx.

A：Vera has four different use cases: Vera Rubin, Vera standalone CPU, Vera with Cx 9 for storage, and Vera in A with Cx 9 for security and compute isolation. The 20 billion figure refers to the standalone CPU use case. The speaker suggests that each use case is built on Vera and implies that supply constraints may exist throughout the life of Vera, due to the four different use cases.

A：Agents are compared to harnesses that orchestrate tasks and manage resources like memory and tool use. They run on CPUs, and the tools they use also run on CPUs. However, the 'thinking' or inference processes within agents happen on GPUs. Sub-agents spawned by agents also use GPUs for inference, while the agents and sub-agents may use simulators on either CPUs or GPUs.

A：The speaker suggests that the world is going to have billions of agents in the future, not just today. These agents will use tools similar to how humans use PCs today. The agents will spawn sub-agents, each requiring inference processes that are expected to happen on GPUs. This indicates a projected growth in the use of agents and their computing needs.

A：The company is accelerating all of the world's tools, data processing engines, and database engines to run on CPUs, due to the high tolerance for performance by agents and the need for rapid processing. This strategy is in line with the projected growth in the use of agents and their computing needs.

A：While traditional computing's economics are measured in dollars per core, the economics of AI are described as tokens per dollar or dollars per token. This reflects the need to generate and process tokens quickly, which Vera is designed to do efficiently.

A：Vera was built to provide the necessary infrastructure for AI, which includes great storage (via Stx), networking (via Spectrum X), Gpus for inferencing, and security features such as confidential computing. It is the first platform to offer end-to-end confidential computing.

A：AI native clouds, which do not design their own chips or assemble unrelated parts, are expected to grow faster than hyperscale in the future. However, the growth may not be exclusively faster as the speaker also suggests that both segments could experience similar growth. Nvidia's architecture is well-suited for AI native clouds and the computing platform is highly rentable, offering a range of benefits to AI native companies.

A：The second category includes enterprises and industrial applications that are similar to Oems and large enterprises. This category started growing after the AI ecosystem developed in the hyperscale, primarily due to their focus on consumer applications and excellent data center capabilities. However, these applications only gained momentum once AI could enhance service delivery, perform productive work, and generate impact and income safely.

A：The second category, which encompasses industrial and enterprise applications, is expected to develop slower initially than hyperscale because AI had not yet proven its capability to be impactful and profitable in these sectors. Nonetheless, it is anticipated to grow significantly over time, becoming the larger segment in the future due to its potential to contribute to the world's economy, which is estimated to be around 50 to 80 trillion dollars and could be larger with the advent of AI.

A：The physical AI and robotic segment is forecasted to grow incredibly fast within the next five years, which is considered a foregone conclusion by the speaker.

A：Growth above the trillion-dollar mark is expected to be driven by the continued growth of share in frontier AI models, the inclusion of Vera CPU which was not factored into the initial trillion-dollar estimate, and the combination of Vera, Rubin, and Lpx to address the entire spectrum of AI from pre-training to post-training to inference and agentic systems.

A：Vera is scheduled to be launched in the second half of the year, starting in Q3. The initial pieces will be available in Q3, and the ramp-up is expected to continue into Q4 and Q1 of the following year.