The latest financial results show that NVIDIA's Q2 revenue maintained double-digit growth, with new-generation Blackwell architecture chip revenue increasing 17% quarter-over-quarter, which CEO Jensen Huang described as a sign of "very strong demand." However, the company's core data center business revenue continues to underperform, partly due to reduced H20 chip revenue, with no H20 sales to China during the quarter.

During the subsequent earnings call, CEO Jensen Huang emphasized that "future opportunities remain enormous," stating that over the next few years, certainly by the end of this decade, we will see very rapid growth and very important growth opportunities.

Regarding China sales, Executive Vice President and CFO Colette Kress indicated that if geopolitical issues can be resolved, H20 chip sales to China in Q3 could reach $20-50 billion, and "some" Chinese customers have obtained licenses in recent weeks.

**China Presents $50 Billion Opportunity This Year with 50% YoY Growth Potential**

Huang stated that China could bring a $50 billion opportunity this year. If we can serve this market with competitive products, assuming this year is $50 billion, we expect this market to grow 50% annually, just like AI markets in other parts of the world are also growing.

He pointed out that China is a very large market, and NVIDIA is discussing with the US government the importance of serving the Chinese market, calling the opportunity to "bring Blackwell to the Chinese market" real, and will continue to pursue sales of Blackwell architecture GPUs to China.

**$3-4 Trillion AI Capital Expenditure Projection "Quite Reasonable"**

Huang mentioned that by the end of this decade, AI infrastructure spending will reach $3-4 trillion, with project scale and scope providing NVIDIA with significant long-term growth opportunities.

He stated that capital expenditure from just the top four hyperscale cloud service providers doubled within two years to $600 billion. Considering that cloud service providers globally are building AI infrastructure, the expectation of "$3-4 trillion" over the next five years is quite reasonable.

Kress noted that factors driving continued investment growth include: agentic AI requiring substantial increases in training and inference computing, global sovereign AI construction, enterprise AI adoption, and the arrival of physical AI and robotics.

**10x ROI per Token Based on Blackwell**

Kress stated that Blackwell has set the benchmark as the new standard for AI inference performance. As inference and agentic AI proliferate across industries, the AI inference market is rapidly expanding. Blackwell's rack-scale NVLink and CUDA full-stack architecture address this by redefining inference economics.

Kress explained that the new NVFP4 4-bit precision and NVLink72 on the GB300 platform deliver 50x improvement in per-token efficiency compared to Hopper, enabling unprecedented scale monetization of compute. For example, a $3 million GB200 infrastructure investment can generate $30 million in token revenue - a 10x return.

**Gross Margin Expected to Reach Mid-70% Level**

Kress stated that the company continues to expect non-GAAP gross margin to reach mid-70% level by year-end. The company is accelerating investments in the business to address enormous future growth opportunities.

Huang noted that "in many ways, the more you buy, the more you grow." Our products offer exceptional value proposition, and customers also achieve excellent margins - therefore growth opportunities and gross profit opportunities using NVIDIA architecture are absolutely the best.

**$20 Billion Sovereign AI Revenue Expected This Year**

Kress also indicated during the call that sovereign AI-related revenue this year could reach $20 billion, more than doubling from last year.

Kress explained that sovereign AI is rising, with countries developing sovereign AI capabilities using domestic infrastructure, data and talent, presenting significant opportunities for NVIDIA. For example, the EU plans to invest 20 billion euros to establish 20 AI factories in France, Germany, Italy and Spain, including 5 super factories, scaling AI compute infrastructure tenfold.

**NVIDIA Fiscal 2026 Second Quarter Earnings Call Transcript**

Event Date: August 27, 2025 Company: NVIDIA

**Investor Relations Toshiya Hari:** Thank you. Good afternoon everyone, and welcome to NVIDIA's fiscal 2026 second quarter call. Joining me today from NVIDIA management are President and CEO Jensen Huang and Executive Vice President and CFO Colette Kress.

**Executive Vice President and CFO Colette Kress:** Thank you, Toshiya. We delivered another record quarter while continuing to navigate a dynamic external environment. Total revenue was $46.7 billion, exceeding our outlook, with sequential growth across all market platforms. Data center revenue grew 56% year-over-year. Despite a $4 billion decline in H20 revenue, data center revenue still achieved sequential growth.

NVIDIA's Blackwell platform reached record levels, growing 17% sequentially. We began production shipments of GB300 in Q2. Our full-stack AI solutions for cloud service providers, emerging clouds, enterprise and sovereign customers are all contributing to our growth.

We are at the beginning of an industrial revolution that will transform every industry. We expect AI infrastructure spending to reach $3-4 trillion by the end of this decade. The scale and scope of this buildout provides NVIDIA with significant long-term growth opportunities.

GB200 NVL systems are being widely adopted and deployed across cloud service providers and consumer internet companies. Lighthouse model builders including OpenAI, Meta and Mistral are using GB200 NVL72 at datacenter scale for training next-generation models and serving inference models in production.

The new Blackwell Ultra platform is also performing strongly, generating hundreds of billions in revenue. Since GB300 shares architecture, software and physical footprint with GB200, the transition to GB300 is seamless for major cloud service providers, allowing them to easily build and deploy GB300 racks.

Production is currently running at full capacity, producing approximately 1,000 racks per week. This output is expected to accelerate further as additional capacity comes online in Q3.

**President and CEO Jensen Huang:** From the highest level, the growth drivers will be the evolution and introduction of agentic AI inference. Past chatbots were one-shot - you give it a prompt and it generates an answer. Now AI conducts research, thinks and makes plans, and can use tools. This is called long thinking - the longer it thinks, the better answers it usually produces.

The computational requirements between one-shot generation and what agentic AI models need could differ by 100x, 1000x, or even more, because it needs to do extensive research, reading and understanding. So the computational growth required for agentic AI is enormous.

Due to agentic AI and vision-language models, we're now seeing breakthroughs in physical AI, robotics and autonomous systems. AI has made tremendous progress over the past year. Agentic systems and reasoning systems are completely revolutionary.

We've built Blackwell NVLink72 systems - rack-scale computing systems born for this moment. We've been developing this for years. Last year, we moved from NVLink8 (node-level computing where each node is a computer) to NVLink72, where each rack is a computer.

In the next few years - you asked about the long term. Over the next five years, we will expand through Blackwell, Rubin and subsequent products into a $3-4 trillion AI infrastructure opportunity.

We're at the beginning of this buildout, and AI technological advances are truly enabling AI to be adopted and solve problems across many different industries.

The earnings call transcript continues with detailed Q&A sessions covering competition, market opportunities, technical specifications, and financial projections, demonstrating NVIDIA's strong position in the AI infrastructure market and optimistic outlook for continued growth across multiple segments including gaming, professional visualization, automotive, and data center operations.