NVIDIA Corporation Q2 FY2026 earnings call

• NVIDIA delivered a record quarter with total revenue of $46.7 billion, exceeding outlook. Data center revenue grew 56% Y/Y, with Blackwell platform at record levels and GB300 in production. - Geopolitically, US government reviewing licenses for H20 sales to China; no H20 shipments to China included in Q3 outlook but potential $2-$5B in Q3 if issues resolve. - Growth fueled by capital expenditures from cloud to enterprises, with $600B expected in data center infrastructure and compute this year. - Blackwell sets benchmark for AI inference performance, with NVFP44 bit precision and NVLink 72 on GB300 platform delivering 50x energy efficiency per token vs Hopper. - Robotics computing platform Justin Thor available, with over 2 million developers and 1,000+ partners adopting it. - Sovereign AI on the rise, with NVIDIA at forefront of initiatives in EU and UK, aiming for over $20B in Sovereign AI revenue this year.

Data Center: Total revenue was $46.7 billion, with data center revenue growing 56% year over year. Blackwell platform reached record levels, growing sequentially by 17%. GB300 began production shipments in Q2. Networking: Record revenue of $7.3 billion, a 46% sequential and 98% year-on-year increase, driven by Spectrum X Ethernet, InfiniBand, and NVLink. Gaming: Record $4.3 billion, a 14% sequential increase and 49% year-on-year jump, driven by Blackwell GeForce GPUs and GeForce NOW expansion. Professional Visualization: $601 million, a 32% year-on-year increase, driven by high-end RTX workstation GPUs. Automotive: $586 million, up 69% year on year, primarily driven by self-driving solutions and Thor SoC shipments.

• Total revenue expected to be $54 billion plus or minus 2% for Q3. - GAAP and non-GAAP gross margins expected to be 73.3% and 73.5%, respectively, plus or minus 50 basis points. - Expect operating expenses to be approximately $5.9 billion (GAAP) and $4.2 billion (non-GAAP) for Q3. - Board approved a $60 billion share repurchase authorization to add to remaining $14.7 billion authorization at end of Q2.

• Geopolitical uncertainties regarding licenses for H20 sales to China, which could impact revenue if not resolved. - Power and cooling limitations may pose challenges to data center build-outs, affecting AI infrastructure growth.

Q: With wafer-in to rack-out lead times of twelve months, and Rubin on track for ramp in the second half, speak to vision for growth into 2026, especially between network and data center.

A: Drivers include evolution of reasoning agentic AI, where computation needed for such models has grown tremendously. Over next five years, NVIDIA aims to scale into a $3 to $4 trillion AI infrastructure opportunity, with CapEx of top CSPs doubling to $600B this year. NVIDIA's platform complexity and perf per watt drive revenue growth in data centers.

Q: Clarify $2 billion to $5 billion in China H20 shipments, what needs to happen and sustainable pace of China business; also competitive landscape with ASIC projects.

A: Still waiting on geopolitical issues for H20 shipments to China; if more licenses and interest arrive, could ship more. NVIDIA builds full stack solutions unlike ASICs, with advantages in full stack co-design, availability across cloud to edge, and accelerating entire AI pipeline. NVIDIA's perf per watt and gross margins make its platform superior.

Q: $3 to $4 trillion in data center infrastructure spend by end of decade, share of that and bottlenecks like power.

A: CapEx of top four hyperscalers doubled to $600B/year. NVIDIA represents about 35% of a gigawatt data center's $50-60B cost. Power limitations are a key limiter, and NVIDIA's perf per watt drives revenue growth in power-limited data centers.

Q: Long-term prospects in China market, importance of Blackwell architecture licensed there.

A: China market has over $50B opportunity this year, expected to grow 50% per year. 50% of world's AI researchers are in China, and open-source models from China fuel global AI adoption. Advocating for American tech companies to address China market and license Blackwell there.

Q: Spectrum XGS announcement, opportunity set and sizing of Ethernet portfolio.

A: NVIDIA offers three networking technologies: NVLink for scale-up, Spectrum X for scale-out, and Spectrum XGS for scale across. Spectrum X Ethernet is a sizable business, and Spectrum XGS connects multiple data centers into super factories. Networking is critical for AI factories, with perf improvements directly translating to revenue growth.