NVIDIA Q2 FY2026: Surfing the AI Surge with Blackwell, Steering Through China’s Export Challenges

TradingKey - NVIDIA Corporation (NASDAQ: NVDA) reported strong Q2 FY2026 results on August 27, 2025, after market close, driven by robust AI-powered demand and adoption of its next-generation Blackwell GPU platform. The company delivered significant revenue and profit beats overall, with impressive operational leverage and strong free cash flow, reinforcing NVIDIA’s leadership in AI data centers and gaming.

However, its key Data Center revenue slightly missed analyst forecasts, reflecting regulatory restrictions on H20 product shipments to China. This shortfall, combined with cautious Q3 guidance excluding H20 sales to China, generated investor uncertainty. As a result, NVIDIA’s stock declined by about 3% in after-hours trading despite the solid quarter, reflecting market concern over near-term headwinds. Investors remain focused on the sustainability of AI growth amid evolving regulatory and market dynamics.

Source: TradingKey

Key Financial Results

Source: NVIDIA Q2 FY26 Earnings Release

Q3 FY2026 Guidance

• Revenue: $54.0 billion ± 2%, reflecting expected sequential growth but conservatively excluding H20 shipments to China.

• Gross Margins: GAAP 73.3%; Non-GAAP 73.5% ± 0.5%, indicating continued margin improvement.

•  Operating Expenses: GAAP $5.9 billion; Non-GAAP $4.2 billion, consistent with sustained R&D and AI ecosystem investments.

•  Tax Rate: Approximately 16.5% ± 1%, excluding discrete items.

Segment Performance

Data Center: The Growth Engine

NVIDIA’s Data Center revenue surged 56% YoY to $41.1 billion, accounting for nearly 88% of total revenue. This growth was powered by strong demand for the Blackwell platform, with Blackwell Data Center revenue growing by 17% sequentially. Production of Blackwell Ultra GPUs is running at roughly 1,000 racks per week and is expected to ramp further in Q3. Demand remains broad-based with about half of Data Center revenue from major cloud providers and significant contributions from enterprises and sovereign AI initiatives. Hopper GPUs continues to serve diverse AI workloads beyond advanced reasoning tasks.

CEO Jensen Huang highlighted Blackwell’s roughly 10x energy efficiency improvement for reasoning AI, positioning it as the foundation for a multi-trillion-dollar AI infrastructure investment over the next decade. However, no H20 chip sales occurred in China this quarter due to export restrictions, limiting near-term revenue visibility, though NVIDIA recognized $180 million from previously reserved H20 inventory sold outside China.

Gaming: Accelerating Momentum

The Gaming segment revenue increased 49% YoY to $4.3 billion, supported by strong adoption of the Blackwell-powered GeForce RTX 5060—the fastest ramping "x60" class GPU ever. Expansion of NVIDIA DLSS 4 technology into over 175 titles and enhancements to the GeForce NOW cloud gaming platform further expanded the gaming ecosystem’s reach and engagement.

Professional Visualization and Automotive: Expanding Ecosystem

Professional Visualization revenue grew 32% YoY to $601 million, fueled by new RTX PRO GPUs and partnerships like Siemens advancing industrial digitalization. Automotive revenue jumped 69% YoY to $586 million as the DRIVE AV software platform reached full production and newly launched Jetson AGX Thor AI supercomputers power robotics and autonomous transportation deployments.

Strategic Developments & Financial Discipline

Rubin Architecture and AI Leadership

CEO Jensen Huang highlighted NVIDIA’s ongoing innovation in AI infrastructure beyond Blackwell, introducing Rubin as the next-generation GPU architecture slated for release in early 2026. Rubin is expected to deliver over twice the AI inference performance of Blackwell, with approximately 50 petaflops of FP4 compute—a significant leap from Blackwell’s around 20 petaflops. Rubin GPUs will feature 288 GB of HBM4 memory per GPU and significantly increased memory bandwidth of about 13 TB/s.

Rubin Ultra, arriving in 2027, will further double this performance to 100 petaflops of FP4 compute and substantially increase memory capacity to 1 TB of HBM4e memory per GPU, using four GPU chiplets per package. Rubin Ultra will support extreme rack scaling with up to 576 GPUs per rack, delivering up to 15 exaflops of FP4 inference compute at the rack level.

While Blackwell currently powers NVIDIA’s leading AI platforms such as the RTX PRO 6000 Blackwell Server Edition, Rubin will drive the subsequent wave of AI model training and inference with cutting-edge improvements in compute performance, memory bandwidth, and interconnect speed.

These advancements position NVIDIA to maintain technological leadership in accelerating increasingly complex AI workloads. Collaborations with OpenAI, Novo Nordisk, and others underscore NVIDIA’s deepening role across AI innovation, spanning from model training to drug discovery and industrial AI applications.

Regulatory Impact: H20 Export Restrictions

NVIDIA’s H20 AI chip, primarily designed for the China market, recorded no shipments to Chinese customers in Q2 due to U.S. export license restrictions. CFO Colette Kress explained that the company’s data center compute revenue declined sequentially, mainly because of a $4 billion reduction in H20 sales. NVIDIA’s Q3 FY26 revenue guidance of $54 billion excludes any China-based H20 sales but acknowledges potential shipments if geopolitical conditions permit. Depending on how export licenses evolve, NVIDIA could realize between $2 billion and $5 billion in H20 revenue in Q3.

CEO Jensen Huang emphasized the real possibility of bringing the next-generation Blackwell AI chip to China if export restrictions are lifted and reiterated the strategic importance of U.S. technology leading the Chinese AI market, projected to grow approximately 50% annually with an estimated $50 billion opportunity.

Competitive Landscape: ASICs and Broadcom

NVIDIA’s CEO Jensen Huang highlighted the company’s unique approach, distinct from ASIC-focused competitors. Unlike ASICs, which often struggle to scale due to the complexity of accelerated computing, NVIDIA’s full-stack co-design supports evolving AI models, from generative to multi-modal transformers, across cloud, edge, and robotics. The Blackwell and Rubin platforms integrate fast memory, low-power caches, CPUs, GPUs, and advanced interconnects like NVLink and Spectrum XGS, delivering unmatched performance-per-watt for power-constrained data centers. NVIDIA’s unified programming model and robust software ecosystem solidify its leadership in AI computing, setting it apart from ASIC-only solutions.

Capital Returns and Share Buybacks

In H1 FY26, NVIDIA returned $24.3 billion to shareholders via repurchases and dividends. With $14.7 billion remaining from the previous buyback authorization and a newly approved $60 billion (roughly 1.5% of NVIDIA's current market capitalization) no-expiration repurchase program, NVIDIA signals significant confidence in sustained cash flow and shareholder value generation.

Margin Recovery and Investments

GAAP gross margins rebounded sharply to 72.4%, recovering nearly 12 points sequentially due to easing supply constraints and product mix improvements. The company targets mid-70% non-GAAP gross margins by year-end. Operating expenses increased 38% YoY, reflecting continued investments in R&D, infrastructure, and AI software ecosystems to fuel long-term growth.

Conclusion and Outlook

NVIDIA's Q2 FY2026 results confirm its dominant position in AI, driven by record revenue growth and strong Blackwell platform adoption amid a $3-4 trillion AI infrastructure buildout. The $60 billion share buyback program signals financial strength and shareholder confidence. However, cautious Q3 guidance, due to ongoing China export restrictions on H20 chips, tempers upside, projecting about 15% sequential growth versus a possible 20-25%. Margins are recovering but shifts in product mix and rising R&D require attention.

Looking ahead, NVIDIA's product roadmap through Blackwell and Rubin, combined with its scalable ecosystem including NVLink and Spectrum networking, supports capturing over 50% of the rapidly expanding AI Accelerator Chip Market. Key risks include regulatory hurdles in China, supply chain constraints, and broader macroeconomic pressures. NVIDIA’s forward P/E of around 34-40x is reasonable given its robust growth and strategic AI investments. Stock dips may present buying opportunities in this AI infrastructure leader, though investors should watch geopolitical and market risks that could affect near-term momentum.